¿Que Determina La Productividad?

18/2/2018 ¿Qué determina la productividad?
Página 1
Journal of Economic Literature 2011, 49: 2, 326-365

http: www.aeaweb.org/articles.php? doi = 10.1257 / jel.49.2.326
¿Qué determina la productividad?
Chad Syverson *
Los economistas han demostrado que las diferencias grandes y persistentes en los niveles de productividad
en todos los negocios son omnipresentes. Este hallazgo ha dado forma a las agendas de investigación en un número
de campos, incluidos (entre otros) macroeconomía, organización industrial,
trabajo y comercio. Este documento estudia y evalúa direcciones de trabajos empíricos recientes
la pregunta de por qué las empresas difieren en sus niveles de productividad medidos. los
las causas son múltiples y difieren según la configuración particular. Incluyen ele-
productos originados en las prácticas de producción y, por lo tanto, sobre qué productores tienen
cierto control directo, al menos en teoría, así como de las operaciones externas de los productores
entornos ambientales. Después de evaluar el estado actual del conocimiento, expongo lo que
ver son las principales preguntas que la investigación en el área debe abordar en el futuro.
( JEL D24, G31, L11, M10, O30, O47)
1. Introducción persistentes diferencias de productividad medidas

a través de los productores, incluso dentro de
L a la infusión masiva de detalles industrias definidas.
datos de actividad de producción en Las magnitudes involucradas son sorprendentes.
estudio en las últimas dos décadas, Chad Syverson (2004b) encuentra que dentro de cuatro
los investigadores en muchos campos han aprendido dígitos
unade las industrias SIC en los Estados Unidos.
gran oferta sobre cómo las empresas convierten los sector
insumos
privado,
en la diferencia promedio en
salidas. Productividad, la eficiencia con productividad total de los factores (TFP) entre
que esta conversión ocurre, ha sido un plantas del percentil 90 y 10 de la industria
tema de particular interés. Los detalles es 0.651. Esto corresponde a una proporción de TFP de
de estos estudios han variado dependiendo de e 0.651 = 1.92. Para enfatizar lo que esto
los intereses específicos de los investigadores, pero número
hay implica, dice que la planta en el
es un hilo común. Ellos han documentado, Percentil 90 de la distribución de productividad
prácticamente sin excepción, enorme y ción hace casi el doble de salida con
las mismas entradas medidas que el décimo por
* Universidad de Chicago y Oficina Nacional de Eco- Planta de centile Tenga en cuenta que este es el promedio
Investigación nómica Agradezco a Eric Bartelsman, Rango
Nick90-10.
Bloom,La desviación estándar de la gama
Roger Gordon, John Haltiwanger, Chang-Tai Hsieh, Ariel
en las industrias de cuatro dígitos es 0.173, por lo que
Pakes, Amil Petrin, John Van Reenen y anónimos
las industrias
árbitros para comentarios útiles. Este trabajo es apoyado por generales ven productos mucho
NSF (SES-0519062 y SES-0820307), y tanto el diferencias entre sus productores. NOS
Stigler Center y la Fundación Centel / Robert P. Reuss la fabricación no es excepcional en términos de
Fondo de investigación de la facultad en el puestodispersión
de la Universidad de ChicagoDe hecho, si algo,
de la productividad
Escuela de Negocios.
326
Página 2
Syverson: ¿Qué determina la productividad? 327
es pequeño en relación con la variación de productividad

de ninguna manera tiene la intención de ser un
observación observada en otra parte. Chang-Tai Hsieh
sive contabilidad. Ellos hablan a la amplitud
y Peter J. Klenow (2009), por ejemplo, del impacto que responde al título de este trabajo
encontrar diferencias de productividad aún mayores
pregunta
en tendría.
https://translate.googleusercontent.com/translate_f 1/30
China e India, con un promedio de 90-10 TFP Macroeconomistas están disecando agregado
relaciones superiores a 5: 1. 1 crecimiento de la productividad: la fuente de casi todos
Estas diferencias de productividad a través de las diferencias de ingreso per cápita en todos los países
los ducers no son fugaces, tampoco. Regressing intenta-en varios microcomponentes, con
la PTF actual de un productor en su año la intención de comprender mejor las fuentes
la PTF retrasada produce coeficientes autorregresivos de tal crecimiento. Foster, Haltiwanger y C.
del orden de 0.6 a 0.8 (ver, por ejemplo, Árpád J. Krizan (2001), por ejemplo, visión general del
Ábrahám y Kirk White 2006 y Foster, papel sustancial de reasignaciones de
Haltiwanger y Syverson 2008). Poner sim- actividad hacia productos de mayor productividad
De hecho, algunos productores parecen haber descubierto (tanto entre las plantas existentes como a través de
su negocio (o al menos están en camino), entrada y salida) para explicar los programas agregados
mientras que otros carecen lamentablemente. Mucho crecimiento
más de ductividad. Hsieh y Klenow (2009)
que los derechos de fanfarronear están en juego aquí: pregunta
otro cuánto más grande es el chino y el indio
hallazgo robusto en la literatura, virtualmente economías serían si lograron el
invariante a país, período de tiempo o industria misma eficiencia en la asignación de entradas a través de
intentar-es que los productores de mayor productividad unidadessonde producción como lo hace Estados Unidos.
más probabilidades de sobrevivir que sus menos eficientes Modelos de fluctuaciones económicas impulsadas por
competidores de la industria. La productividad está losbastante
choquesiluminada
de productividad son cada vez más
generalmente una cuestión de supervivencia para enriquecido las empresas. para tener en cuenta los patrones de micro niveles,
y se estiman y prueban usando plantas
1.1 Cómo la productividad de Micro-Level
datos de productividad a nivel de empresa en lugar de
Variación y persistencia tiene
agregados (por ejemplo, Jeffrey R. Campbell y
Investigación influenciada
Jonas DM Fisher 2004, Eric J. Bartelsman,
El descubrimiento de ubicuo, grande y per- Haltiwanger y Stefano Scarpetta 2009,
las diferencias de productividad constante ha dadoy forma Marcelo Veracierto 2008). Microprocesador
agendas de investigación en una serie de campos.datos aquí de ductividad también se han llevado a cabo
son algunos ejemplos de esta influencia, aunque sobre cuestiones de crecimiento a largo plazo, la conversión de ingresos
gence y derrames de tecnología. Ellos ofrecen
un nivel de resolución inalcanzable con
1 Estas cifras son para medir la productividad basada en los ingresos
sures; es decir, donde la producción se mide utilizando datos bloqueados
los ingresos de la planta
(deflactado a través de los años usando índices de precios En la organización
específicos de industrial,
la industria).
la investigación tiene
Medidas de TFP que usan cantidades físicas comoniveles medidas de salida
de productividad vinculados a una serie de
las sures en lugar de los ingresos en realidad exhiben aún más
variación que las medidas basadas en los ingresoscaracterísticas de la tecnología, la demanda y el mercado
según lo documentado
en Lucia Foster, John Haltiwanger y Syverson (2008). estructura. Los ejemplos incluyen el efecto de
Hsieh y Klenow (2009) también encuentran mayor productividad
competencia (Syverson 2004a y James A.
dispersión en sus medidas de TFP que utilizan proxies de cantidad
para medir la producción (las cantidades físicas reales no están Schmitz 2005),disponibles)
el tamaño de los costos hundidos (Allan
capaz para la mayoría de los productores en sus datos). Collard-Wexler
Aunque es2010), y la interacción de
solo un componente de la PTF basada en los ingresos (siendo
rivalidad del el otro de productos y derrame de tecnología
mercado
precio promedio del productor), la TFP basada en cantidad puede ser
más disperso porque tiende a ser negativamente correlativo overs (Nicholas Bloom, Mark Schankerman,
con precios, ya que los productores más eficientesy venden John Vana menor
Reenen 2007). Otra línea de
precios. Por lo tanto, las medidas de productividad basadas
estudio ha en los ingresos,
analizado que
la interacción de las empresas
combinar la productividad basada en la cantidad y los precios, tienden a
subestimar la variación en las eficiencias físicas de estructuras organizativas con productividad
los productores.
Página 3
328 Revista de Literatura Económica, vol. XLIX (junio de 2011)
niveles (por ejemplo, Vojislav Maksimovic y Gordon

literaturas, los hechos anteriores plantean obvio y
Phillips 2002, Antoinette Schoar 2002, y preguntas cruciales. ¿Por qué las empresas (o fábricas,
Ali Hortaçsu y Syverson 2007, 2011). tiendas, oficinas o incluso producción individual
Los economistas del trabajo han explorado líneas, para el caso) difieren tanto en su
importancia del capital humano de los trabajadores
habilidades
en para convertir entradas en salida? Lo es
explicando las diferencias de productividad (Johnsuerte
M. tonta o en su lugar algo, o muchos
Abowd et al. 2005 y Jeremy T. Fox y cosas más sistemáticas? Pueden productores
Valérie Smeets 2011), los efectos de productividad
controlar los factores que influyen en la productividad
de pago de incentivo (Edward P. Lazear 2000), o son productos puramente externos de la
otras diversas prácticas de recursos humanos ¿entorno operativo? Lo que apoya tal
(Casey Ichniowski y Kathryn Shaw 2003), grandes diferencias de productividad en equilibrio?
talento y prácticas gerenciales (Bloom y Hace una década, cuando Bartelsman y Mark
Van Reenen 2007), forma organizativa Doms (2000) escribió la primera encuesta de la
(Luis Garicano y Paul Heaton 2007), y literatura de productividad de microdatos para esto
conexiones sociales entre compañeros de trabajo diario,
(Orianalos investigadores estaban empezando a
Bandiera, Iwan Barankay e Imran Rasul pregunta "¿Por qué?" Gran parte del trabajo
2009). También ha habido un enfoque en el a ese punto se había centrado en establecer hechos
papel de la reasignación impulsada por la productividad
como losende arriba, el "¿Qué?" de la productividad
dinámica del mercado de trabajo a través de la creación
dispersión
de empleo
de la naturaleza
y Desde entonces, la literatura tiene
destrucción (Haltiwanger, Scarpetta, y enfocado más intensamente en las razones por las cuales
Helena Schweiger 2008). los niveles de productividad son muy diferentes en
Quizás en ningún otro campo tengamos la producción
negocios. Definitivamente ha habido progreso
patrones de dispersión de actividad mencionados ress.
anteriormente
Pero también hemos aprendido más sobre qué
una mayor influencia en la trayectoria de la que no sabemos, y esto está guiando las formas
agenda de investigación que en la literatura comercial.
en el cual la literatura de productividad será
Marcos teóricos utilizando heterogénea emocionante. Este artículo está destinado a ser una guía para
empresas de productividad como Jonathan Eaton yy comenta sobre esta investigación.
Samuel Kortum (2002) y Marc J. Melitz Comienzo por establecer algunos límites. yo tengo
(2003) son ahora el concepto conceptual dominante a. Una descripción completa de micro-
lentes a través de los cuales los economistas ven el
investigación
comercio de productividad fundada no es ni
impactos. En estos modelos, los impactos comerciales
posible en este formato ni deseable. Ahí
varían entre productores y dependen de su son simplemente demasiados estudios para permitir
niveles de productividad en particular. Agregar quate cobertura de cada uno. Primero, me enfocaré en
las ganancias de productividad provienen de selecciones
trabajo empírico
mejoradasEsto no es porque veo
ción y una mayor competencia que es más importante que la teoría. Más bien,
trae. Una multitud de estudios empíricos tienen ofrece una cobertura más profunda de esta importante
acompañado y estimulado por estos pequeña faceta de una literatura gigante y es mejor
teorías (por ejemplo, Nina Pavcnik 2002, Andrewrefleja mi experiencia. Dicho eso, voy a esbozar
B. Bernard, J. Bradford Jensen y Peter K. una simple productividad heterogénea
Schott 2006 y Eric A. Verhoogen 2008). modelo de industria a continuación para enfocar la discusión
Han confirmado muchas de las predicciones sion, y ocasionalmente mencionaré
patrones y preguntas planteadas por los suyos. trabajo teórico específico con particularmente
vínculos cercanos a los problemas empíricos discutidos.
1.2 La pregunta de "¿Por qué?"
Además, por razones obvias, lo haré
Dado el importante papel que desempeñan centrarse en la investigación que se ha hecho desde
las diferencias de actividad juegan en estos dispares
Bartelsman y Doms (2000) fueron escritos.
Página 4
Incluso dentro de estos límites, hay La Sección 5 discute lo que veo como el gran
hay más estudios de los que pueden ser satisfactoriamente
preguntas sobre la productividad a nivel empresarial
descrito individualmente Veo el papel de este artículo
patrones que aún necesitan ser respondidos UN
como filtrar las lecciones más amplias de la lit- breve conclusión de la sección siguiente.
erature a través de la lente de un subconjunto de clave
estudios. Los documentos en los que me enfoco no son
2. Productividad: qué es, cómo es
elegidos necesariamente porque son los primeros
Medido y cómo es su dispersión
o solo un buen trabajo en su materia,
Es sostenido
sino más bien porque tenían un arquetipo
calidad que me permite tejer una narración de Esta sección revisa brevemente qué productos
la literatura. Insto a los lectores cuya interacción la actividad es conceptualmente, cómo se mide en
Ests se han despertado con más intensidad práctica, y cómo las diferencias de productividad
explorar las literaturas relevantes. Hay lejos entre los productores de bienes similares podría ser
más que aprender de lo que puedo transmitir aquí.apoyado en equilibrio. Discusión más profunda
Un descargo de responsabilidad: algunos de missionesdebates
sobre la teoría de los índices de productividad
contiene elementos de comentario. Estas opiniones se puede encontrar en Douglas W. Caves, Laurits
iones son solo míos y pueden no ser R. Christensen y W. Erwin Diewert
consenso de investigadores en el campo. (1982) y las referencias en el mismo. Más
Organizo este artículo de la siguiente manera. losdetalles sobre problemas de medición se pueden encontrar
la siguiente sección esboza el respaldo conceptualen la gran literatura sobre el tema; ver por
terreno: qué es la productividad, cómo es a menudo ejemplo, G. Steven Olley y Ariel Pakes
medido en la práctica, y cómo las diferencias (1996), Zvi Griliches y Jacques Mairesse
en la productividad entre los productores de similares
(1998), Richard Blundell y Stephen R.
los bienes pueden mantenerse en equilibrio. Bond (2000), James Levinsohn y Amil
La Sección 3 analiza las influencias en la productividad
Petrin (2003) y Daniel C. Ackerberg et
que operan principalmente dentro del negocio. Alabama. (2007). Ejemplos de modelos que derivan
Esto puede ser a nivel de empresa, nivel de plantaEquilibrios
o de la industria con productos heterogéneos
incluso en procesos específicos dentro de la empresa.productores de ductividad incluyen Boyan Jovanovic
Muchas de estas influencias pueden potencialmente (1982), Hugo A. Hopenhayn (1992), Richard
estar bajo el control de los actores económicos Ericson y Pakes (1995), Melitz (2003),
dentro del negocio. En otras palabras, pueden Marcus Asplund y Volker Nocke (2006),
ser "palancas" que la administración u otros tienen y Foster, Haltiwanger y Syverson
disponible para impactar la productividad. Sección (2008)
4
se centra en la interacción de los productores
2.1 Productividad en concepto
niveles de conductividad y los mercados en los que
ellos operan Estos son elementos de En pocas palabras, la productividad es eficiencia en
entornos externos que pueden afectar producción: cuánto producto se obtiene
niveles de productividad. Este impacto podría no de un conjunto dado de entradas. Como tal, es
ser siempre directo, pero pueden inducir típicamente expresado como una entrada de salida
Ducers para tirar de algunas de las palancas discutidas
proporción. Medidas de productividad de un solo factor
en la sección 3, influencia indirecta observada reflejar las unidades de producción producidas por unidad de
niveles de productividad en el proceso. Que puedeuna entrada particular. La productividad laboral es la
también ser factores que afectan la cantidad de pro- la medida más común de este tipo, aunque
dispersión
en deylalaductividad
equilibrio influenciaque puedepro
observó ser sostenida
ocasionalmente
tivity se usan Porcapital o incluso
supuesto, solo producción de materiales
diferencias de conductividad a través de ese canal.los niveles de productividad de los factores se ven afectados por
Página 5
intensidad de uso de las entradas excluidas. Dos TFP es, en esencia, un residuo. Como con todo
los productores pueden tener una mano de obra bastante residuos, diferente
de alguna manera es una medida de nuestro
niveles de productividad a pesar de que tienen el ignorancia: es la variación en la producción lo que
misma tecnología de producción si sucede una no se puede explicar en base a lo observable
usar capital mucho más intensamente, dicen entradas. Entonces, es justo interpretar el trabajo dis-
porque enfrentan diferentes precios de factores. cundió en esta encuesta como un intento de "poner un
Debido a esto, los investigadores a menudo usan se un
enfrentan en "ese residual, o más exactamente,
concepto de productividad que es invariante para "Poner el caras", dadas las múltiples fuentes
intensidad de uso de las entradas de factores observables. de la variación de productividad La literatura tiene
Esta medida se llama factor total progresó cuando puede explicar el sistema-
(TFP) (también se llama a veces mul- influencias áticas en la producción en
productividad tifactor). Conceptualmente, TFP unidades de unidad que no provienen de cambios en
las diferencias reflejan los cambios en las isocuantas entradas
de unobservables como trabajo estándar o capital
función de producción: variación en pro- Tal medidas.
producido a partir de un conjunto fijo de entradas. Mayor-TFP
2.2 Medición de la productividad
los productores producirán mayores cantidades de
salida con el mismo conjunto de entradas observables Si bien la productividad es relativamente directa
que las empresas de menor PTF y, por lo tanto, tienen reenviar en concepto, una gran cantidad de mediciones
isocuantas que se desplazan hacia arriba y hacia elsurgen problemas cuando se construye
derecho. La variación del precio del factor que impulsa medidas fac-
de producción de datos reales.
para las diferencias de intensidad no afecta a TFPIrónicamente, mientras que la investigación con micro pro-
porque induce cambios a lo largo de isocuantas datos de ducción amplía en gran medida el conjunto de
en lugar de cambios en isocuantas. preguntas que responden y mueve el nivel de
La TFP se ve más fácilmente en el uso frecuente análisis más cerca de donde las decisiones económicas
formulación de una función de producción dondeestán hechos que los datos agregados, también
salida es el producto de una función de observación plantea problemas de medición y calidad de datos
insumos capaces y un factor neutro (alternativa más frecuentemente.
Shifter tively, Hicks-neutral): El primer conjunto de cuestiones se refiere a la salida
medida. Muchas empresas producen más
Y t = A t F ( K t , L t , M t ), de una salida. ¿Deberían ser agregados?
a una sola medida de salida, y cómo si es así?
donde se produce Y t , F (·) es una función de Además, incluso microdatos detallados del productor
capital de entradas observables K t , trabajo L t , yno suelen contener medidas de salida
materiales intermedios M t , y A t es el cantidades. Los ingresos son típicamente observados
palanca de cambios factor-neutral. En este tipo deenformulario lugar. Dada esta limitación de los datos, la
lation, TFP es A t . Captura variaciones en el enfoque estándar ha sido utilizar los ingresos
salida no explicada por los cambios en el observador (deflactado a los valores reales de un año común
entradas capaces que actúan a través de F (·). 2 utilizando la serie de deflactores de precios) para medir
poner. Si bien esto puede ser aceptable, e incluso
deseable, si las diferencias de calidad del producto son
2 Utilizo un cambio de tecnología separable multiplicativa plenamente reflejado en los precios, puede ser problemático
para facilitar la exposición, pero se puede extraer la TFP
a partir de una función de producción variable en el tiempo general Y = t
G ( A , K , L , M ). Totalmente diferenciando Sin

t t t t t estapérdida
producción
de generalidad, podemos elegir unidades para nor-
la función da: malize ∂ G / ∂ A = 1. Por lo tanto, cuando las entradas observadas son fijas
_ _ _ _
d Y =∂ A
t Gd A + ∂ K
t Gd K + ∂ LGd L + ∂∂ M
t t t Gd M . comió cambiosdM
t ( dK
t = dL
t = = 0), cambios diferenciales en TFP, dA , cre-
en la salida d Y .
t
t
Página 6
siempre que la variación de precios en vez encarna El tercer conjunto de problemas de medición
diferencias en el poder de mercado en la producción
implica agregar múltiples entradas en un
ers. En ese caso, los productores midieron pro- Medida TFP. Como se describió anteriormente, la diferencia de TFP
los niveles de ductividad pueden reflejar menos acerca
Las ferencias
de cómoreflejan cambios en la salida mientras se sostiene
eficientes son y más sobre el estado de entradas constantes. Para construir la salida-
su mercado de producción local. El trabajo reciente
relación
tiene de entrada que mide TFP, un investigador
comenzó a profundizar en las consecuencias debe ponderar las entradas individuales apropiadas
de asumir productores
utilizando los ingresos de unmedir
para solo producto y
la producción. cada
índicevez
Discutiréde que construya
entrada. una sola dimensión
La ponderación correcta es fácil
esto más abajo. Mientras tanto, iré est a ver cuando la función de producción es
adelante suponiendo ingresos deflactados Cobb-Douglas:
reflejar de forma clara la producción del productor.
El segundo conjunto de problemas de medición _ Yt
TF P t = A t =
considera entradas. Para el trabajo, existe la opción K t α k L t α l M t α m.
de si usar la cantidad de empleados,
horas-empleado, o alguna calidad ajustada En este caso, las entradas se agregan por tak-
medida laboral (la nómina salarial se utiliza a menudo
el exponente
en de cada factor a su respec-
este último papel, basado en la noción de que los elasticidad
salarios de salida activa. Resulta que esto
capturar productos marginales de heterogénea es más general como un primer orden aproximado
unidades de trabajo). El capital generalmente se mide
mación a cualquier función de producción. los
utilizando el establecimiento o el valor contable deíndice
la empresa
de entrada en el denominador de TFP puede ser
de su capital social. Esto plantea varias preguntasconstruido de manera similar para producción general
ciones. Qué tan bueno es un proxy para el capitalfunciones.
social 3
para el flujo de servicios de capital? Si el Incluso después de determinar cómo con-

stock sea simplemente el libro reportado por el productor
estructurar el índice de entrada, uno debe
valor, y ¿cuáles son los deflactores? O deberia seguro las elasticidades de salida α j , j ∈ { k , l , m }.
el stock se construye utilizando Varios enfoques son comunes en la lit-
inversiones y el inventario perpetuo erature. Uno se basa en suposiciones de
método-y qué asumir acerca de la depreciación minimización de costos para construir la elasticidad
ación? Al medir materiales intermedios vincula directamente con los datos de producción observados.
als, un problema similar a los ingresos como salida Un productor que minimice los costos equiparará un
la cuestión anterior surge, porque típicamente soloelasticidad de salida de entrada con el producto de
los gastos totales del productor en insumos el costo compartido de esa entrada y la escala elástica-
están disponibles, no cantidades de entrada. Más itydiversión-
Si se pueden medir los costos compartidos (obtener
damentalmente, ¿cómo deberían las entradas intermedias
Los costos de capital usualmente son los más prácticos
¿ser manejado? ¿Debería uno usar una producciónpunto bruta?de fricción aquí) y la elasticidad de la escala
función de producción e incluyen intermediarios ya sea estimado o supuesto, entonces la salida
comió los insumos directamente, o debería ser intermedios
simplemente se resta de la salida para
3 Si bien los enfoques al estilo Cobb-Douglas son probablemente
lidiar con una función de producción de valor agregado?
el más común en la literatura, muchos investigadores también
Además de todas estas consideraciones, uno haceuse la forma translog (vea Caves, Christensen y Diewert)
estas opciones de medición de entrada en el con- 1982), que es una aproximación de segundo orden al general
funciones de producción y, como tal, es más flexible, aunque
texto de saber que cualquier salida impulsada pormás exigente de los datos. También hay un no completamente
variaciones de entrada no medidas (debido a la entrada
enfoque paramétrico, análisis envolvente de datos (DEA),
diferencias de calidad o capital intangible, para que se usa en ciertos círculos, algo distintos del
literatura. Ver William W. Cooper, Lawrence M. Seiford,
ejemplo) se mostrará como productividad. y Kaoru Tone (2006) para una descripción general de los métodos DEA.
Página 7
las elasticidades α j pueden construirse directamente.Este

Si un
enfoque plantea problemas econométricos.
investigador está dispuesto a hacer algo de Como lo señaló por primera vez Jacob Marschak y
pero no asunciones inocuas, a saber, William H. Andrews (1944), las opciones de entrada son
competencia perfecta y retornos constantes a probable que esté correlacionado con los programas del productor
escala, entonces las elasticidades equivalen a la proporción
ductividadde ω t : productores más eficientes son, todos
ingresos pagados a cada entrada. Esto hace que de lo contrario, es probable que contrate más insumos. Ahí
estructurando el α j simple. Materiales y mano de también
obra es posible sesgo de selección cuando un panel
las acciones suelen ser fáciles de cobrar se usa, ya que los productores menos eficientes-aquellos
con la factura de los salarios y los gastos de materiales
con bajo ω t - es más probable que salga de
datos a la mano. La parte del capital puede ser la muestra. (Como se discutirá a continuación, el
estructurado como el residual, obviando la necesidad
correlación positiva entre productividad y
para medidas de costo de capital. (Aunque hay unla supervivencia es uno de los hallazgos más sólidos en
problema conceptual ya que, como el modelo quela literatura.) Luego está la cuestión de pro-
sigue a continuación señala, no está claro qué variación de precios a nivel ducer mencionada anteriormente.
hace que el tamaño del productor sea finito de forma
Unaperfecta
literatura sustancial ha surgido para abordar
competitivo, mundo de devoluciones constantes).estos asuntos; ver Griliches y Mairesse
advertencia importante es que el enfoque de índice (1998), Ackerberg et al. (2007) y Johannes
asume los costos de ajuste de los factores de distancia.
Van Biesebroeck
Si (2008) para los resúmenes.
están presentes, las condiciones de primer orden Existe un debate sobre cuál de los muchos
vinculando las partes observadas del factor con lalos
salida
métodos
elas- disponibles es lo mejor. Al final, como veo
las ticities no se mantendrán. Esto puede ser mitigado
elegir un método es cuestión de preguntar
en parte (pero al costo) mediante el uso de costosacompartidos
sí mismo qué quesuposiciones uno es confortante
han sido promediados en cualquier momento o pro- capaz de hacer. Ciertamente uno no puede escapar
ducers para suavizar idiosincrásico el hecho de que algunas suposiciones deben hacerse
desajustes impulsados por el costo de ajuste en al estimar la función de producción.
niveles de entrada reales y óptimos, pero algunos Afortunadamente, a pesar de estos muchos
la mala medición podría permanecer. cerns, muchos de los resultados descritos en este
Un enfoque separado es estimar el es probable que el papel sea bastante robusto para
elasticidades α j mediante la estimación de la producción
peculiaridades de seguridad. Cuando los estudios tienen
función. En este caso, la TFP (registrada) es simplemente
prueba robustez directamente, típicamente encuentran
la suma estimada de la constante y la poca sensibilidad a las elecciones de medición.
residual. En el caso Cobb-Douglas (que La variación inherente en el establecimiento o
de nuevo, recordemos, es una aproximación de primer microdatos
orden a nivel de empresa es generalmente tan grande como
a tecnologías más generales), la estimación para pantano cualquier pequeña medida inducida
la ecuación es: diferencias en las métricas de productividad. Simplemente
poner, los productores de alta productividad tenderán a
En Y t = α 0 + α k ln K t + α l ln L t mirar eficiente independientemente de la forma específica
que su productividad se mide Yo usualmente
+ α m ln M t + ω t . utilizar números de índice de TFP basados en el costo compartido
un primer pase en mi propio trabajo; son fáciles de
Por lo tanto, la estimación de TFP sería α 0 + ω t ,construir y ofrecer la solidez de ser
donde el primer término es común a través de una aproximación de primer orden no paramétrica
unidades de producción en la muestra (generalmente a unael función de producción general. Dicho esto,
la tecnología se estima a nivel industrial), siempre es aconsejable verificar los resultados de uno
y el segundo es idiosincrásico para un particular solidez a los detalles de la medición
productor. enfoque.
Página 8
2.3 Un modelo de dentro de la industria tomar como dado. En imperfectamente competitivo

Dispersión de productividad mercados, las suposiciones sobre el lugar R (·)
restricciones sobre la forma de interconexión competitiva
Dadas las grandes diferencias en productos acción (ya sea monopolísticamente competitiva o
dentro de una industria que discutí anteriormente,oligopolístico) y a través de esto las formas de
una pregunta natural es preguntar cómo podrían las curvas de demanda residual. Los contenidos
mantenerse en equilibrio. La ubicuidad de de D también dependerá de los detalles de
esta dispersión sugiere que debe haber algunos la estructura competitiva. Por ejemplo, en
fuerza económica real en el trabajo, en lugar de un oligopolio de Cournot de costo heterogéneo, D
simplemente ser un artefacto de medición contendrá los parámetros de la industria
o una oportunidad extraña. Aquí, esbozo un simple curva de demanda y los niveles de productividad
modelo que muestra cómo eso es posible. los productores de la industria, ya que estos son
modelo también será útil para facilitar ficiente para determinar el equilibrio de Nash
discusión a lo largo de esta encuesta. resultados y por lo tanto los ingresos de cada pro-
Los productores de la industria, indexados por Ducer
i , ganani . A pesar de estas restricciones, esta configuración
beneficios dados por es razonablemente general.
Las suposiciones sobre la forma de R (·)
πi=R(Ai,Li,D)-wLi-f. implica que, dado el estado industrial D , cada
productor tiene un empleo óptimo único
R (·) es una función de ingresos generales. A i es el
nivel de nivel *Lque i está aumentando en su pro-
nivel de productividad del productor, y L i es su mano
niveldedeobra
ductividad. Intuitivamente, el productor
entrada. (Supongo que el trabajo es la única entradaniveldeóptimo
la empresa
de empleo (al que me refiero
por el bien de la simplicidad.) Niveles de productividad
de aquí en adelante como su tamaño), que se establece
difieren entre los productores. La forma específica para
de equiparar los ingresos marginales y marginales
R (·) depende de la estructura de la salida costos, se limita al aumentar marginal
mercado y la función de producción. Ingresos costos en mercados perfectamente competitivos y
también puede depender de un estado industrial D. unaEsta
curva de demanda residual con pendiente descendente
puede ser un vector o un escalar y, dependiendo de (y posiblemente aumentando los costos marginales
la estructura de la competencia del mercado de salida,
bien) en mercados imperfectamente competitivos.
puede incluir shocks de demanda en toda la industria, Denote las ganancias del productor en su opti-
la cantidad de productores industriales, sus pro- mal tamaño por
niveles de ductilidad, y / o momentos de la pro-
distribución de la ductividad. Tanto el salario w π ( A i , *L ,i D ) = R ( A i , L i* , D )
y el costo fijo f son comunes a través, y tomados
según lo dado por, todos los productores. -wLi *-f.
Supongo que R (·) es dos veces diferenciable con
∂ R / ∂ L > 0, ∂ 2 R / ∂ L 2 <0, ∂ R / ∂ A > 0, y Por el teorema de la envolvente y las condiciones
∂ 2 R / ∂ A ∂ L > 0. Si la industria está perfectamente
ciones sobre la función de ingresos, los beneficios son
competitiva, estas condiciones son satisfactorias aumentando en la productividad del productor
fied dado una función de producción que es nivel A i . Esto implica que habrá una crítica
similarmente diferenciable, cóncavo en L , y nivel de productividad cal A tal que para A i < A ,
donde la productividad y el trabajo son comple- los beneficios serán negativos. A dependerá en general
mentos. Además, bajo competencia perfecta, eral en w , f , y el estado industrial D. Ya que
toda la información contenida en D se refleja en D puede depender de la distribución de
el precio de mercado P que iguala la demanda total niveles de productividad en la industria, lo haremos
y suministro, que los productores por supuesto necesita una condición adicional para determinar
Página 9
el equilibrio de la industria. Esto viene de that all producers make nonnegative oper-
una estructura de entrada de la siguiente manera. ating profits and that entry occurs until
Un gran grupo de potencial idéntico ex ante the expected value of taking a productivity
los participantes deciden si entrar en la industria draw is zero. By pinning down the equilib-
tratar. Primero eligen si pagar o no rium distribution of productivity levels in
costo de entrada s para recibir una the industry through determining A , it also
sacar de una distribución con probabili- determines the equilibrium industry state
función densidad densidad g ( A ) durante el intervalo D . The particular values of A and D depend
[ A l , A u ]. 4 Si un posible participante elige on the exogenous components of the model:
recibir un sorteo, determina después de observar g ( A ), w , f , and s , and the functional form of
si comenzar la producción en su óptimo R (·).
tamaño y ganar el correspondiente operativo The equilibrium productivity distribution
beneficios π ( A * ,i ,DL).i will be a truncation of the underlying pro-
Solo los posibles participantes con productos productivos ductivity distribution g ( A ). Específicamente, el
atrae lo suficiente como para hacer que no sea negativo equilibrium distribution (denoted γ ( A )) is:
las ganancias operativas elegirán producir en
equilibrio. De ahí el valor esperado de _
if A ≥ _UN
pagar s es el valor esperado de π ( A , L * , D )
sobre g ( A ), condicional al dibujo A i ≥ A.
Este valor esperado obviamente se ve afectado por
{
γ(A)=
1 − GA( )_
0
g(A) de otra manera.
el nivel de costo de corte A. Una condición de entrada There

libre
are two notable features of this dis-
ción clava este valor: A debe establecer la red tribution. First, it is not trivially degenerate;
valor esperado de entrada en la industria V e the model supports productivity heterogene-
igual a cero. Por lo tanto, A satisface ity under general conditions. Esto es porque
A ti high-productivity producers are limited in
π ( A , L * , D ) g ( A ) dA - s = 0 . their ability to sell to the industry's entire
Ve= ∫ _ UN mercado. This finite optimal producer size is a
consequence of the concavity of the revenue
Esta expresión resume la industria función. In perfectly competitive markets,
equilibrio. 5 Combina las dos condiciones this concavity comes from increasing mar-
ginal costs. In industries with imperfectly
competitive output markets, the concavity
4 Estos límites son esencialmente arbitrarios siempre que arises from downward-sloping demand
span A para cualquier posible D. Es decir, un productor con producción
nivel de actividad A no es rentable (es decir, no puede
l
curvescubrir
(due tosuproduct differentiation from
costos) en cualquier estado industrial posible, y uno anycon
source)
producción
and, possibly, from increasing
tivity A siempre es rentable.
u
marginal costs as well. In either case, one
5 He hecho dos suposiciones implícitas en esta ecuación.
First, V is exactly zero only in industries with a large
e
can interpret productivity A as a factor of
number of producers. I will assume there is a continuum production that differs in quantity or quality
of producers for the remainder of the discussion. Esto acrossesproducers. A higher level of A loosens
consistent with an assumption of perfect competition or
monopolistic competition in the product market, though the size constraint but does not eliminate it.
obviously rules out strategic oligopolistic interactions. Second,
los the average productivity level
model's logic applies to industries with a discrete in number
the industry will vary as the exogenous
of firms, however. In that case, free entry condition will
imply a number of producers N such that the expected parameters change. Increases in the aver-
value of entry with N − 1 firms is positive but is negative age productivity level across plants (com-
with N firms. The other assumption is that the productivity ing from parameter changes that increase
distribution g ( A ) is continuous, but the model can be mod-
ified to accommodate discrete productivity distributions. A ) will thus expectedly translate into higher
Página 10
Syverson: What Determines Productivity? 335
aggregate industry productivity—the ratio of may be favorable enough to induce some of

total industry output to total industry inputs. 6 them to want to do so. Any adjustment to a
Therefore what happens at the micro level new, postshock equilibrium will therefore
feeds upwards into aggregates. This feature require reallocation of inputs from their ini-
reflects a major thrust behind the research tial locations. Favorably shocked producers
agenda of understanding micro productivity: will grow, unfavorably shocked producers
it teaches us more about aggregate produc- will shrink or exit, and new producers may
tivity movements. enter the industry at a productivity level
Of course, this model is very simple and above exiters. These patterns of reallocation
leaves out many features observed in empiri- are robust features of the data.
cal work on productivity. I will quickly dis- A greater limitation of the model is that
cuss two such features. a producer's productivity is exogenous. los
As a two-stage entry and production model, equilibrium productivity distribution is
the model abstracts from dynamics. Puede endogenized only through a selection effect:
therefore be interpreted as characterizing the determination of who produces in equi-
long-run industry equilibria. That said, ver- librium via A . While I discuss below that
sions of this model's type with more complex selection is an empirically important mecha-
dynamics have been worked out by, among nism, it is abundantly clear that producers
others, Hopenhayn (1992) and Asplund and often take actions to try to raise their pro-
Nocke (2006). Further, even this simple ductivity level. In this case, the equilibrium
structure hints at how the dynamics of real- sketched out above will not directly apply,
location—a focus of some of the literature though many of its basic elements will.
discussed below—might work. Suppose the Despite the model's simplicity and limited
industry is initially in equilibrium and then scope, it can form a useful conceptual base
each producer is hit with a persistent, inde- upon which to build the discussion below.
pendent productivity shock. Those receiv-
ing favorable shocks will see an increase in
3. Productivity and the Plant or Firm
their optimal size, while those hit by negative
shocks will want to shrink. Indeed, some may This section discusses factors that directly
be hit by shocks so adverse that they will no impact productivity at the micro level by
longer be profitable. And if we imagine there operating within the plant or firm. Ellos
are still potential entrants who could pay the are “levers” that management or others can
sunk cost to take a productivity draw, the potentially use to impact the productivity of
environment after the productivity shocks their business. They are akin to forces that
would allow firms in the model of the pre-
vious section to raise their A i draw, though
6 For differentiated product industries, relating an
most likely at a cost. Section 4 below will
industry's aggregate productivity level to the productivity
levels of its component firms requires constructing focus
a quan-
on influences external to the firm: ele-
tity index that adds up firms' disparate outputs. The proper
ments of the industry or market environment
index depends on how the product varieties enter final
that can induce productivity changes or sup-
demanders' utility functions. Under standard aggregators,
increases in the average firm-level productivity translate
port productivity dispersion.
into increases in aggregate industry productivity (see, eg, broken up the discussion of direct
I have
Melitz 2003). However, there are complications involved
in empirically mapping back-and-forth between changes productivity impacts by category for the sake
in micro-level productivity distributions within anofindustry
exposition. However, it's good to keep in
and changes in aggregate industry productivity (see, mindeg,that some forces can overlap these cat-
Paul Schreyer 2001, Petrin and Levinsohn 2005, Susanto
Basu et al. 2009, and Charles R. Hulten 2009). egories, and multiple mechanisms can act in
Página 11
336 Journal of Economic Literature, Vol. XLIX (June 2011)
concert. I will point out many of these across- A recent set of papers has made consider-
category links as the discussion goes along. able efforts to close this measurement gap.
Some have focused on single-industry or even
3.1 Managerial Practice/Talent
single-firm case studies by necessity, given the
Researchers have long proposed that man- detail required in the data. More comprehen-
agers drive productivity differences. 7 Whether sive efforts that cover a broader cross section
sourced in the talents of the managers them- of economic activity are underway, however.
selves or the quality of their practices, this is Bloom and Van Reenen (2007) offer one
an appealing argument. Managers are con- of the most comprehensive studies relating
ductors of an input orchestra. They coor- management practices (though not managers
dinate the application of labor, capital, and per se) to productivity. They and their team
intermediate inputs. Just as a poor conductor surveyed managers from over 700 medium-
can lead to a cacophony rather than a sym- sized firms in the United States, United
phony, one might expect poor management to Kingdom, France, and Germany. They sur-
lead to discordant production operations. veyed plant managers, so the measured prac-
Still, perhaps no potential driver of pro- tices revolve around day-to-day and close-up
ductivity differences has seen a higher ratio operations rather than the broader strategic
of speculation to actual empirical study. Datos choices made at the executive level.
limitations have been the stumbling block. Surveys were conducted over the phone
The proliferation of production microdata by a questioner who shared the respondent’s
has afforded a great increase in detail, but native language. Information was probed
such data rarely contains detailed informa- on eighteen specific management practices
tion on any aspect of managerial inputs. in four broad areas: operations, monitor-
Sometimes there may be a distinction made ing, targets, and incentives. The interview-
between blue- and white-collar or producers scored the firm on its practices based
tion and nonproduction employees, but on these responses. Given the inherently
that is usually it. The identity, much less subjective element of this measurement pro-
the characteristics, practices, or time alloca- cess, Bloom and Van Reenen took several
tion of individual managers is rarely known. steps to enhance accuracy and consistency.
Furthermore, managerial inputs can be very Managers were not told they were being
abstracto.
ters It's the
but what not manager
just time does
allocation that mat-
with their scored.(eg,
ended Questions on tell
“Can you practices
me howwere
youopen-
pro-
time, like how they incentivize workers or mote your employees?” rather than “Do you
deal with suppliers. promote your employees based on tenure?”).
Financial performance was not discussed.
7 I mean long proposed: Francis A. Walker (1887) pos-

its that managerial ability is the source of differences employers
in of labor, in the matter of ability to meet these
surplus across businesses: “The excess of produceexacting which conditions of business success, that we have the
we are contemplating comes from directing forcephenomenon to its in every community and in every trade, in
proper object by the simplest and shortest ways; from whatever
sav- state of the market, of some employers realiz-
ing all unnecessary waste of materials and machinery; ing nodeprofits at all, while others are making fair profits;
boldly incurring the expense—the often large expense—of others, again, large profits; others, still, colossal profits.”
improved processes and appliances, while closelyItscruti- is impressive how Walker's description closely matches
nizing outgo and practising a thousand petty economies (albeit with the flowing prose typical of the time) the
in unessential matters; from meeting the demandsviewpoints of the of researchers over 120 years later. We finally
market most aptly and instantly; and, lastly, from are exercis-
becoming able, with the growing availability of broad-
ing a sound judgment as to the time of sale and the based
terms production microdata, to test such hypotheses on a
of payment. It is on account of the wide range among comprehensive
the basis.
Pagina 12
The firms were small enough so that the CEO's succession—ie, he is the eldest son
interviewers would not already be aware of of the firm's founder. (I will discuss the com-
the performance of the firms they surveyed. petition–productivity link more extensively
Each interviewer conducted dozens of inter- in section 4. Interestingly, primogeniture’s
views, allowing Bloom and Van Reenen to tie to productivity is not about family own-
control for interviewer fixed effects when ership per se—in fact, family ownership in
relating management scores to outcomes. isolation is positively correlated with good
Further, over sixty firms were surveyed management.) These two factors are respon-
twice, by different interviewers; la corre- sible for explaining most of the difference
lation between the separate management between the country-level average manage-
practice scores for the same firms was 0.73. ment scores in the sample. The variation
Much of what was scored as “best practice” in these averages is largely the result of the
management in the interviews was based on United Kingdom and France having a left
the recommendations of the management tail of poorly managed firms. Both countries
consulting industry. This raises concerns have traditionally favored primogeniture by
about whether these practices are actually tradition and family-firm exemptions in their
related to performance, or just the manage- inheritance tax laws.
ment fad of the moment. Importantly, there- Disentangling whether these correlations
fore, Bloom and Van Reenen document that are causal is more challenging. Quizás
higher-quality management practices (and management consultancies base their rec-
higher scores) are correlated with several ommendations on the practices observed at
measures of productivity and firm perfor- successful firms, but some excluded factor
mance, including labor productivity, TFP, drives both management practice and per-
return on capital, Tobin's Q , sales growth, and rendimiento. Bloom and Van Reenen, aware of
the probability of survival. 8 The correlation this issue, estimated a specification in an ear-
between a firm's management practice score lier working paper version of the article that
and its total factor productivity is statistically used competition and primogeniture mea-
strong and economically nontrivial. Spanning sures to instrument for management scores.
the interquartile range of the management The notion is that the competitive and legal
score distribution, for example, corresponds environments are orthogonal to other factors
to a productivity change of between 3.2 and that drive management practices, at least
7.5 percent. This is between 10 and 23 per- in the short run. The estimated effect of
cent of TFP's 32 percent interquartile range management practices on TFP remains sta-
in their sample. tistically significant and is in fact larger than
Bloom and Van Reenen show two fac- the ordinary least squares case. Esto puede
tors are important predictors of the qual- suggest that unobserved third factors have a
ity of management practice in a firm. Más modest role, if any, and that Bloom and Van
intense competition in the firm's market, Reenen's management practice scores reflect
measured in several ways, is positively cor- (albeit noisily) true managerial acumen.
related with best-practice management. Bloom and Van Reenen have since
Additionally, management practice scores expanded their management practice sur-
are lower when the firm is family-owned vey program to gain greater coverage of
and primogeniture determined the current business practices across economies. Bloom
and Van Reenen (2010) and Bloom et al.
(2010) review results from an extension of
8 The data from this paper is available online at http://
cep.lse.ac.uk/_new/publications/abstract.asp?index=2313. this survey program to nearly 6,000 firms in
Página 13
seventeen countries, including fast-growing effects (particularly for CEOs) have signifi-
China, India, and Brazil. The broader results cant explanatory power over firms' returns on
echo those above. A particularly interesting bienes. Adding these fixed effects to a regres-
pattern emerging from the early analysis is sion of returns on firm and year fixed effects
that the much lower average management raises the adjusted R 2 from 0.72 to 0.77.
practice scores in China, India, and Brazil These results reflect performance differ-
are driven not so much by lower productiv- ences that can be explained by the identity
ity across the board (though this is present to of the managers. This still leaves open the
some extent), but in particular by a large left question of what the managers do or know
tail of very poorly managed firms. Esto tiene that affects performance. Bertrand and
obvious implications for how trade growth Schoar don't have the sort of detailed man-
and its assorted competitive pressures might agement practice data of Bloom and Van
impact productivity evolution in these and Reenen, but they do regress their estimated
otros países (More about Chinese and manager fixed effects on two variables they
Indian firms' TFP levels below.) Bloom and observe for the executives in their data: age
Van Reenen are now further expanding the and MBA attainment. They find that while
survey program to incorporate a panel ele- age is not a significant factor, managers with
ment. This will be extremely useful, as it will MBAs have significantly higher return on
allow one to control for unobservable fixed assets effects (by roughly 1 percent, as com-
heterogeneity across firms as well as to see pared to a mean of 16 percent). Esto podría
how firms' management practices change be due to their more aggressive behavior as
when their external environment does. reflected in investment, leverage, and divi-
Other work in this vein includes James dend-paying (or lack thereof) choices. Más
B. Bushnell and Catherine Wolfram (2009), recent work (eg, Steven N. Kaplan, Mark
who find that power plant operators have M. Klebanov, and Morten Sorensen 2008
nontrivial impacts on the thermal efficiency and Ulrike Malmendier and Geoffrey Tate
of power plants. The best can boost their 2009) has started to dig deeper into how par-
plant's fuel efficiency by over three percent, ticular CEO practices and philosophies are
saving millions of dollars of fuel costs per tied to firm performance.
año. Unfortunately, the data are less clear Other within-firm work has suggested
about what particular actions or attributes that the human resources components
predict good plant management. of management, in particular, can affect
These research lines study managerial productividad. This research—see for example
actions and policies at levels below the exec- Ichniowski, Shaw, and Giovanna Prennushi
utive suite. Other work has focused on how (1997), Lazear (2000), Barton H. Hamilton,
those at the apexes of corporate hierarchies Jack A. Nickerson, and Hideo Owan (2003),
influence performance. Marianne Bertrand the papers cited in Ichniowski and Shaw
and Schoar (2003) study top executives (eg, (2003), Bruce Shearer (2004), and Bandiera,
CEOs, CFOs, Presidents, etc.) who manage Barankay, and Rasul (2007 and 2009)—uses
at least two firms for three years each dur- highly detailed, production-line-specific data
ing their 1969–99 sample period. Siguiendo to tie nonstandard human resource man-
managers across multiple firms lets them test agement practices like pay-for-performance
if individual executives can explain variation schemes, work teams, cross-training, and
in firms' performance measures. While they routinized labor–management communica-
don't measure productivity specifically, they tion to productivity growth. These papers
do find that the individual manager fixed have elucidated some interesting details
Página 14
about the productivity effects of these prac- long way toward establishing whether or not
tices. For instance, these practices may be a causal link exists. Any such link would raise
complements: while they may have only additional questions. First, even if the inter-
modest impact on productivity when imple- ventions raised productivity, were they cost
mented in isolation, their total impact is effective? That is, would they pay for them-
larger than the sum of its parts when used selves in a market setting? Second, given
in conjunction. Further, these practices are what we know about Indian firms in general,
likely to have heterogeneous effects across particularly for the left tail of the productiv-
production lines, even in the same plant, ity distribution, if management consulting
if different lines produce product vari- were to be effective anywhere, it would be
ants of varying complexity. Brent Boning, in India. Should the experiment therefore be
Ichniowski, and Shaw (2007), for example, thought of as measuring the upper bound of
find an interaction between the complexity the causal effect of management practices?
of the production process and the ability of
3.2 Higher-Quality General Labor and
innovative human resource management in Capital Inputs
raising productivity.
Alexandre Mas (2008) shows in a vivid Management is an unmeasured input
case study how poor management–labor in most production functions, and hence
relations can have productivity effects. Él is embodied in the productivity measure.
looks at the resale values of equipment made Similarly, the productive effects of inputs like
at plants and times where Caterpillar was (nonmanagement) labor and capital can also
experiencing labor strife during the 1990s. enter productivity if there are input quality
Compared to otherwise identical products differences that standard input measures do
made at plants or times without unrest, these not capture. 9
products had about 5 percent lower resale There is of course an enormous literature
valores. This substantial productivity impact on human capital, far too large to cover here,
due to the implied reduction in the equip- that has tied several factors to labor quality,
ment's quality-adjusted service flows totaled including education, training, overall experi-
$400 million. ence, and tenure at a firm. Much of this work
With these and other studies, the evi- in labor economics has focused on wages as
dence that management and productivity are the outcome of interest. A smaller set of
related is starting to pile up. Further, some work has looked at human capital's impact
of this work strongly suggests that this rela- on productivity.
tionship is causal. Still, establishing causality
definitively remains a key issue for research.
Bloom, Benn Eifert, Aprajit Mahajan, David 9 Attempts to capture labor quality differences in labor
measures rather than productivity are the impetus behind
McKenzie, and John Roberts (2011) are using the wage bill to measure labor inputs rather than
attempting to establish as much by using the number of employees or employee-hours. The notion
what many consider to be the gold stan- is that market wages reflect variations in workers' contri-
butions to production; firms with more productive work-
dard for establishing causality: a random- ers will have a higher wage bill per employee. Por supuesto,
ized field experiment. They are providing there are problems with this approach: wage variation
management consulting to a random set of might reflect the realities of local labor markets, or cau-
sation could be in the other direction, if more productive
Indian firms and will compare productiv- producers earn rents that are shared with or captured by
ity growth in this treatment group to that employees (Van Reenen 1996; Abowd, Francis Kramarz,
observed in a set of control firms not receiv- and David N. Margolis 1999). Hence, more direct labor-
quality measures are needed to definitively pin down labor
ing the intervention. This study could go a quality's productivity contribution.
Página 15
Newer work using matched employer- productivity dispersion as a metric of a newly

employee datasets, which allow individual measured factor's importance in explain-
workers to be tracked across plants or firms ing productivity—or an R 2 -type measure as
over time, has offered evidence on the Bertrand and Schoar use—is a good idea.
importance of labor quality. Abowd et al. Studies seeking to explain productivity dis-
(2005) offer a broad survey of the early evi- persion should strive to conduct and report
dence from these types of datasets, which similar exercises.)
tend to be newly constructed and therefore Capital can also vary in quality in ways not
still have short panel histories. Their applica- captured with standard measures. If capi-
bility for studying productivity, while limited tal vintages differ from one another in how
now, will greatly increase over time. Todavía, much technological progress they embody,
some progress has been made with such the common book-value-based capital
datos. Pekka Ilmakunnas, Mika Maliranta, stock measures will tend to miss variations
and Jari Vainiomäki (2004), for example, use in average capital vintages across produc-
Finnish matched worker–plant data to show ers. Several studies have tried to measure
that (not surprisingly) productivity is increas- the rate of capital-embodied technological
ing in workers' education as well as age. progress by carefully constructing measures
As great a potential as such data may hold, of the distribution of capital vintages within
the results in Fox and Smeets (2011) sug- plants or firms. Plutarchos Sakellaris and
gest that matched employer–employee data Daniel J. Wilson (2004) do exactly this using
will not answer all of the literature's burn- the annual investment histories of plants in
ing questions. They use matched employer– the US Annual Survey of Manufactures
employee records from the Danish economy and industry-year-specific depreciation mea-
to control for worker education, gender, sures. They estimate a production function
experience, and industry tenure in produc- that is standard in all respects except that,
tion function estimation. While these labor rather than measuring capital inputs with sin-
quality measures have significant coefficients gle dollar-valued stock, they use a weighted
in the production function, accounting for sum of the plant's past investments. los
their influence only decreases the average weights combine the cumulative deprecia-
within-industry 90–10 percentile productiv- tion of a particular vintage's investment and
ity ratio from 3.74 to 3.36. There is plenty of a technological progress multiplier that they
productivity variation left to be explained. En estimar. They assume that capital efficiency
a somewhat encouraging find for researchers units grow at a constant rate per year, which
using more limited datasets, they find that they estimate to be between 8 to 17 percent
including the wage bill alone as a measure per year, depending on the specification.
of labor inputs—data that is almost always These numbers are striking in their implica-
available—does almost as well as including tions about how much productivity growth
the full array of their human capital mea- can come from investment alone. (Note that,
sures, though they caution that wage bills unlike the standard capital deepening effects
are subject to endogeneity concerns, as dis- of investment that serve only to shift labor
cussed above. This finding of only a modest productivity, capital-embodied technologi-
role for finer labor skills measures in explain- cal progress also raises TFP.) Other studies
ing productivity differences is echoed in using different methodologies (eg, Jason G.
Fernando Galindo-Rueda and Jonathan E. Cummins and Giovanni L. Violante 2002)
Haskel's (2005) investigation with similar have found somewhat smaller values, on the
UK data. (Incidentally, using the decline in order of five percent per year. This seems to
Página 16
be an area desperate for further evidence, productivity growth in the mid-1990s after
given its potential importance. twenty years of sluggish performance, and
Van Biesebroeck (2003) measures the that IT has more generally influenced pro-
productivity impact of auto assembly plants ductivity patterns across multiple industries
shifting to “lean” technologies, which in that and countries. Given the sheer size of GDP
context involves new capital plus a host of per capita variation that can be driven by
complementary practices (teamwork, just- even a modest change in trend productivity
in-time ordering, etc.). This is also clearly growth over a sustained period, it is not sur-
related to the managerial practice discus- prising that sources of such changes receive
sion earlier. He finds that both the entry of considerable research attention. Porque
new lean plants and the transformation of this attention, I discuss the work done on this
earlier vintage plants are responsible for the particular capital type separately here.
industry's acceleration of labor productiv- An overview of IT capital's broad pro-
ity growth during the late 1980s and early ductivity impacts, particularly in driving the
1990s Interestingly, his estimates of each growth resurgence, can be found in Dale W.
technology's parameters suggest that capi- Jorgenson, Mun S. Ho, and Kevin J. Stiroh
tal-augmenting productivity is the primary (2005, 2008) and Stephen D. Oliner, Daniel
driver of labor productivity growth under E. Sichel, and Stiroh (2007). These stud-
lean processes, while Hicks-neutral TFP- ies document that IT-related productivity
type productivity drives growth in the tradi- gains—both spectacular productivity growth
tional technology plants. in IT- producing industries and more modest
Of course, not just physical capital can changes in IT- using industries—play an
have unobservable quality differences. important role in explaining aggregate US
Certain types of capital may be themselves productivity growth over the past couple of
invisible—that is, intangible capital. Tal décadas
capital can include any of a number of con- At the same time, Bart van Ark, Mary
cepts, like a firm's reputation, know-how, or O'Mahony, and Marcel P. Timmer (2008)
its loyal customer base, just to name a few. show that the European Union's compara-
Despite the difficulty in quantifying these bly sluggish productivity growth over the
types of capital, they can have very real out- same period can be explained in large part
put effects that, as such, will result in mea- by the later emergence and smaller size of
sured productivity differences. I will discuss IT investment in European economies.
some specific cases of intangible capital in Bloom, Sadun, and Van Reenen (forthcom-
operation below, but the full breadth and ing) suggest that it is not geography per se
depth of intangibles' role in explaining pro- that matters, but rather the location of the
ductivity differences are still very much open owning firm. They show US-based multi-
preguntas nationals operating in the European Union
are more productive than their EU coun-
3.3 Information Technology and R & D
terparts, and this productivity advantage is
While the research described above indi- primarily derived from IT capital. They link
cates that input heterogeneity matters, the their management practices data discussed
productivity effects of a particular type of above to data on IT usage to test for particu-
capital—information technology (IT)—have lar mechanisms through which this produc-
been the subject of intense study. Esto es tivity advantage arises. Their evidence points
rightly so; many have hypothesized that IT to a complementarity between IT capital and
was behind the resurgence in US aggregate human resources practices, explaining US
Página 17
multinationals' productivity advantage in the firms to better match their production capa-
European Union. bilities to their customers' desires, increasing
These broad patterns raise the question the surplus of their sales.
of which specific micro mechanisms actually Such a gain in surplus from product spe-
underlie the aggregate relationship between cialization raises an important broader point
IT and productivity growth. Varios estudios about productivity measurement. Mejor
have explored this issue with detailed pro- customization from IT can raise firms aver-
duction data. Thomas N. Hubbard (2003) age product prices. Measures of productiv-
shows how on-board computers raise aver- ity in physical units of output (eg, number
age utilization rates of trucks that they are of valves per unit input) may therefore not
installed in. The computers provide dis- fully capture the surplus gained. This is one
patchers real-time information on a truck’s case where the limit of most producer-level
locations and load status, allowing them to datasets to revenue-based output measures
better match the available cartage capacity to does not pose a measurement problem
innovations in demand. 10 because this sort of productivity gain would
Ann Bartel, Ichniowski, and Shaw (2007) be reflected in revenues but not physical
show how better computer numerically con- cantidades. (That said, the concern about
trolled (CNC) machining centers—auto- price variations due to local market power
mated devices that shape parts from raw or demand shocks creating productivity
material stock—raise productivity in the mismeasurement still applies in differenti-
valve manufacturing industry by shortening ated product settings.)
setup times, raising speeds of production Erik Brynjolfsson et al. (2008), Bartelsman,
runs, and even allowing quicker inspections. Pieter A. Gautier, and Joris de Wind (2010),
The appealing element of the study's empiri- and Giulia Faggio, Kjell G. Salvanes, and
cal approach is that both the products and Van Reenen (2010) each draw, in related
the production process, except for the partic- but distinct ways, broader lines connecting
ular pieces of IT capital whose contribution IT and productivity. Brynjolfsson et al. docu-
is of interest, remain constant across observa- ment case studies where IT enhances the
ciones. The paper also shows that IT-intensive speed with which firms can replicate prac-
product design tools like computer-aided tices they find productive in one of their lines
design packages make it easier to design of business across the entire organization.
customized parts, and lower setup times This ability to lever-up a productivity advan-
make multiple production runs less costly. tage means successfully innovating firms
Offering a broader array of parts allows the displace less productive competitors more
quickly. IT thus raises the volatility of firm
actuación. Brynjolfsson et al. test for and
10 Adopting any new technology, IT or otherwise, obvi- find this heightened volatility in a sample
ously has its own costs. A new technology's net productivity
of Compustat firms in sixty-one industries.
benefit to the adopter depends on the difference between In the context of the model in section 2,
the increased production the new technology facilitates
and its acquisition cost. For the marginal adoptingBrynjolfsson
pro- et al. essentially argue that IT
ducer, this net gain will be zero. However, inframarginalreduces the concavity of the firm's revenue
producers experience positive productivity gains.function,los allowing them to better lever-
aggregate productivity gains that any technology will offer
will therefore also depend on the competitivenessage (and in a dynamic world, do so more
of the
technology-producing sector. A lower markup andquickly) price any inherent productivity advan-
for the technology raises both the number of inframar- tages (increases in A i ) that they develop or
ginal adopters and the net productivity gain that each
experiencias. stumble upon.
Página 18
Bartelsman, Gautier, and de Wind (2010) There is a long literature linking R&D and
further develop the notion that IT shifts not productivity, and recent additions to it have
just the mean of the distribution of inno- focused on exploring the ties at the micro
vation outcomes but its variance as well. nivel. As with many input-based stories of
Because poor outcomes are truncated by productivity differences, the difficulty is in
the option to exit—again in the parlance of separating correlation from causation. Ahí
the model above, firms drawing a produc- are many reasons why more productive firms
tivity level below A don't need to produce might do more R&D, suggesting that some
at a loss—greater variance raises the value of the causation may go the other way.
of making risky innovations. Bartelsman, Ulrich Doraszelski and Jordi Jaumandreu
Gautier, and de Wind note, however, that exit (2009) model firm productivity growth as
costs (absent in the model in section 2) will the consequence of R&D expenditures with
stifle firms' willingness to innovate because uncertain outcomes. Estimating their model
they make it harder to dismiss unsuccessful using a panel of Spanish firms, they find that
resultados. They argue that employment- R&D does appear to explain a substantial
protection legislation like firing costs makes amount of productivity growth. Sin embargo,
exit more expensive and therefore reduces and picking up the theme of increased vari-
firms'
that willingnesssectors
IT-intensive to adopt
areIT.
in They show
fact smaller ancefind
also tied that
to ITfirm-level
capital discussed above,
uncertainty they
in the
in countries with greater legal restrictions on outcome of R&D is considerable, much
firms' abilities to close unsuccessful lines of more so than with respect to the return on
negocio. They cite employment protection physical capital investment. In fact, their
legislation as a major contributor to the IT estimates suggest that engaging in R&D
gap documented by van Ark, O'Mahony, and roughly doubles the degree of uncertainty
Timmer (2008). (I will further discuss the in the evolution of a producer's productivity
role of flexibility in input markets further in nivel.
section 4 below.) Bee Yan Aw, Mark J. Roberts, and Daniel
Faggio, Salvanes, and Van Reenen (2010) Yi Xu (2008) highlight the bidirectional cau-
document that within-industry productiv- sality between R&D and productivity in
ity dispersion in the United Kingdom has their study of Taiwanese electronics export-
trended upwards over the past couple of ers. They find that firms that select into
décadas They relate this increased disper- exporting tend to already be more produc-
sion to the growth in wage dispersion that has tive than their domestic counterparts (more
occurred over the same period in the United on this in the trade section below), but the
Kingdom and almost every other developed decision to export is often accompanied by
economía. It would be interesting to see if large R&D investments. These investments
similar productivity spreading is occurring in raise exporters' productivity levels further in
concert with wage dispersion growth in these turn, highlighting both selection and causal
other economies. More directly applicable effects tying productivity to R&D. The tim-
to the theme of this section, however, is that ing of this R&D blitz is consistent with a
Faggio, Salvanes, and Van Reenen show world where the exporters are more willing
that industries that experienced the greatest to innovate on the margin because they can
growth in productivity dispersion also saw spread the potential gains of productivity
the largest increases in IT capital intensity. growth across a larger market.
This is yet more evidence tying IT to greater Of course, R&D is simply one of the more
productivity variance. observable components of firms' overall
Página 19
innovative efforts. Many firms undertake a temporary but substantial increase in labor
both process and product innovation without requisitos.
formally reporting R&D spending. (I will Rebecca Achee Thornton and Peter
discuss product innovation's ties to produc- Thompson (2001) investigate what types of
tivity differences in further detail below.) experience matter in productivity growth
This limits the literature's ability to give a from learning by doing. Their data includes
comprehensive look into the relationships unit labor requirements for several design
between productivity and innovation. Still, it variants of 4,000 Liberty ships produced by
is a very useful start, and the mechanisms the multiple shipyards during World War II.
R&D literature highlights are likely to often The multidesign/multiyard nature of the
overlap with the effects of unmeasured inno- data lets them estimate the relative pro-
vative spending. ductivity contributions of four different
measures of past production experience:
3.4 Learning-by-Doing
the yard's past production experience with
The very act of operating can increase a particular design, the same yard's past
productividad. Experience allows producers to production of other designs, other yards'
identify opportunities for process improve- experience with the particular design, and
mentos. This productivity growth, often called other yards' production of other designs.
learning-by-doing, has a long and rich history Not surprisingly, a yard's past production
of study in the literature but has recently of a particular model matters most for pro-
been investigated in more detail given newly ductivity growth in that same model. Después
available micro-level production data. that comes the yard's experience with other
C. Lanier Benkard (2000) studies the pre- ship designs, at about 60 percent the size of
cipitous drop in the labor hours Lockheed the own-design effect. Cross-yard spillovers
needed to assemble its L-1011 TriStar wide- are considerably smaller—only about five
body aircraft. The first few units off the line to ten percent of the own-yard, own-design
required more than one million person hours learning impact. These cross-plant learn-
(equivalent to three shifts a day of 2,500 working effects, while relatively modest here, do
ers each for fifty work days). This was cut in show that producers may become more pro-
half by the 30th plane, and halved again by ductive by learning from other businesses. yo
the 100th. Benkard estimates both the learn- will discuss cross-business spillovers more
ing rate—how fast past production increases abajo.
productivity (decreases unit labor require- Steven D. Levitt, John A. List, and
ments)—and the “forgetting” rate, which is Syverson (2011) find more limited cross-
how fast the knowledge stock built by learn- model learning spillovers within an auto
ing depreciates. Forgetting is quantitatively assembly plant. Using detailed data on hun-
important in this setting: Benkard estimates dreds of individual operations during assem-
that almost 40 percent of the knowledge bly of thousands of cars, they studied the
stock depreciates each year. This may not be causes and effects of manufacturing defects.
literal forgetting but could instead primarily This particular plant began production of
reflect labor turnover. An additional factor three model variants (nameplates) of a com-
in “forgetting” was the shift to a new variant mon platform at staggered times during a
of the plane after about 130 units. Esta nueva production year. Each time a new model
variant was different enough that the imper- ramped up, the plant began a new learn-
fect substitutability of the knowledge stock ing curve. An interesting contrast was seen
between the original and new variants led to when looking at what happened to defect
Página 20
rates when a new shift started producing a price variation due to differences in mar-
given model. In that case, relearning was not ket power across producers exist.) Product
necesario. The new shift began operating at innovation can be aimed at entering new
defect rates at about the same level as the markets or at refocusing a firm's efforts
previous shift had achieved after it already toward growing demand segments as doc-
had run down much of the learning curve. umented in Daron Acemoglu and Joshua
Ryan Kellogg (2009) looks at oil and gas Linn (2004).
drilling in Texas to study how learning occurs Product innovation's productivity effects
when an upstream and downstream producer have been studied in several recent papers.
work together over time. He follows the As touched on above, one of the mechanisms
efforts of pairs of producers and drillers. los behind IT-based productivity growth that
former are companies actively involved in Bartel, Ichniowski, and Shaw (2007) point to
exploring for, extracting, and selling oil, while is an improved ability to customize products.
the latter firms specialize in boring out the Other inputs mentioned above, like R&D
wells that the producers hope will yield oil. and higher-quality employees, can also spur
Since producers typically work with multiple innovación.
drillers and vice versa, and work in different Rasmus Lentz and Dale T. Mortensen
fields, Kellogg is able to separately measure (2008) use Danish firm-level data to esti-
the productivity impacts of the experience mate a model of firms' product innovation
of producers alone (ie, regardless of the efforts in the vertical-quality-ladder style of
drilling firms they work with), drillers alone, Tor Jakob Klette and Kortum (2004). Ellos
and the joint experience of producer–driller find that about 75 percent of aggregate pro-
pairs. He finds that accumulated experience ductivity growth comes from reallocation of
between a producer–driller pair increases inputs (employment in their setup) to inno-
productivity above and beyond that of each vating firms. About one-third of this comes
of the firms' overall experience levels. Esta from entry and exit channels. The other two-
relationship-specific experience is a type of thirds occurs as inputs move toward growing
capital that is lost if the firms split up, giving firms (and hence innovating firms as seen
them incentives to preserve their contracting through the lens of their model) from firms
ambiente. that lose market share when they fall behind
the quality frontier.
3.5 Product Innovation
Natarajan Balasubramanian and Jagadeesh
Innovations in product quality may not Sivadasan (2011) link detailed and broad-
necessarily raise the quantity of output based data on firms' patenting and produc-
(measured in some physical unit) per unit tion activities (they merge the NBER patent
input, but they can increase the product database with the US Census Business
price and, therefore, the firm's revenue Register) to see what happens when a firm
per unit input. If one thinks about produc- patents. They find clear evidence that new
tivity as units of quality delivered per unit patent grants are associated with increases in
input, product innovation can enhance firm size (by any one of a number of mea-
productividad. This is captured in standard sures), scope (the number of products it
revenue-based productivity measures since makes), and TFP (though the evidence is
they reflect price variations across an indus- weaker here). Whether these correlations
try's plants or firms. (Though as mentioned reflect the causal effects of patents is not
above and discussed further below, revenue clear; patenting activity could be just one
productivity can also be misleading when part of a firm's coordinated push into new
Página 21
mercados. Nevertheless, given the breadth of encima). It is also the subject of Acemoglu
the study's coverage and its result that cor- et al. (2007). The evidence tends to be sug-
relations exist, more research in this area gestive but indirect, however, and this is an
would be worthwhile. area where careful work in measuring firm
Bernard, Stephen J. Redding, and Peter structures (not an easy task) could pay big
Schott (2010) show that a firm's TFP is dividendos.
positively correlated with the number of Silke J. Forbes and Mara Lederman
products it produces. This holds both in the (2011) look at how vertical integration
cross section and within firms over time. A affects airline performance. They find
the very least, these results indicate that that, among flights departing from a given
productivity growth accompanies expan- airport on a given day, airlines that own
sion of the variety of products a firm offers. their regional affiliates experience shorter
It is less clear whether innovative activity delays and fewer cancellations than those
drives both productivity and product-vari- contracting with affiliated regionals at
ety growth or whether firms experiencing arm's length. This performance advantage
general productivity shocks “strike while appears to come largely from differen-
the iron is hot,” expanding their product tial performance on adverse weather days.
offerings in response. The role of changes Forbes and Lederman posit that contracts
in product scope in firm size and produc- are limited in their ability to fully specify
tivity growth is one that is just beginning contingent actions necessary to react most
to get the attention it deserves in research effectively to short-horizon logistical prob-
agendas. lems. Vertical integration, by clearly setting
out the decision rights within the organiza-
3.6 Firm Structure Decisions
tion, allows airlines to more nimbly respond
A lot of the micro productivity litera- to unexpected scheduling issues. This flex-
ture uses the establishment (eg, factory, ibility comes at a cost, however: primarily
store, or office) as the unit of analysis. Esta in higher wage costs for integrated airlines.
is in part data driven; many surveys are This could explain why not every mainline
conducted at this level. Plus, plants often carrier has integrated.
embody the smallest indivisible unit of a Hortaçsu and Syverson (2011) use the
production process and, as such, are a natu- Longitudinal Business Database, which con-
ral level at which to study technologies. Pero tains most private nonagricultural establish-
it is also clear that firm-level factors and, in ments in the United States, to examine the
particular, the organizational structure of productivity of plants in vertically structured
the firm's production units—the industries empresas. They find that vertically integrated
they operate in, their vertical and horizontal plants have higher productivity levels than
linkages, their relative sizes, and so on—will their nonintegrated industry cohorts, but
sometimes be related to the productiv- most of this difference reflects selection of
ity levels of the firm's component business high-productivity plants into vertical struc-
unidades. tures rather than a causal impact of inte-
Some have suggested there is a link gration on productivity. Surprisingly, these
between firm decentralization and how eas- productivity differences—and indeed the
ily productive new technologies are adopted. firm's choice to have a vertical structure at
Bloom, Sadun, and Van Reenen (2009) favor all—usually are not related to the move-
this explanation for European firms' recent ments of goods along the production chain.
laggard productivity growth (as mentioned Vertically integrated firms' upstream plants
Página 22
ship a surprisingly small amount to down- It simply reflects the optimal allocation of
stream plants in their firm (small relative to resources within a business given the firm’s
both the firms' total upstream production inherent abilities. They support their effi-
and their downstream needs). Roughly one- cient allocation argument by showing that
third of upstream plants report no shipments conglomerate firms' most productive plants
to their firms' downstream units; half ship are in their largest segments, and segments
less than three percent of their output inter- of a given rank are more productive in larger
nally. This suggests that rather than moderat- empresas. Furthermore, conglomerates expand
ing goods transfers along production chains, on their strongest margins: their largest,
integration instead allows more efficient most productive segments are more sensi-
transfers of intangible inputs (eg, manage- tive to demand shifts than their smaller, less
rial oversight) within the firm. efficient lines of business.
Maksimovic and Phillips (2002) and Schoar (2002) notes that, in her sample,
Schoar (2002) both investigate the produc- plants in conglomerates have, if anything,
tivity of plants within conglomerate firms (in higher permanent productivity levels. los
their setting, those that operate in multiple observed discount reflects the temporary
two- or three-digit SIC industries). Su adjustment costs resulting from the very
work was spurred on in part by the exten- act of diversifying into new businesses. Ella
sive finance literature on the “diversification shows that when a conglomerate diversifies,
discount,” the term for the oft-measured the plants it buys actually experience pro-
negative correlation between a firm's finan- ductivity growth, suggesting that they are in
cial returns and the number of business lines fact being reallocated to more capable man-
it operates. Both papers leverage US manu- agement (there will be more on the realloca-
facturer microdata to convincingly argue that tion of productive inputs below). Al mismo
the diversification discount is not about low time, however, the conglomerate's exist-
productivity (or even, in one case, any sort of ing plants suffer productivity losses. Ya que
underperformance). They differ, however, in conglomerates have on average many more
their explanations. existing plants than acquired ones, average
Maksimovic and Phillips (2002) make a productivity in the firm falls for a period.
selection argument. Firms that choose to Schoar attributes these productivity changes
specialize are likely to have idiosyncratically to a “new toy” effect: managers (over-) con-
high productivity draws in a particular line centrate their efforts on integrating the new
of business but considerably weaker draws plants and business lines at the expense of
outside this segment. Firms that choose con- existing ones. She also finds evidence that
glomerate structures, on the other hand, are the firms' wages absorb any performance
likely to have high draws in several indus- rents, also leading to a bifurcation between
tries but not exceptionally high draws in performance as measured by productivity
any particular industry. Thus if one simply and by stock market returns.
compares the performance of a conglom-
erate's segments to the focused and highly
4. External Drivers of
productive segments of a specialist, the lat-
Productivity Differences
ter would expectedly be higher. Este resultado
does not rely on the previous literature’s The previous section discussed factors
favored explanations of management over- that operate within the firm to determine
reach, cross subsidization of weak segments, niveles de productividad. Producers have, at least
or other agency problems at conglomerates. in theory, some degree of control over these
Página 23
factores. This section focuses instead on how observed productivity levels. The most basic
producers' operating environments can producer theory, after all, says any profit-
influence productivity levels and growth. minimizing firm minimizes its cost of pro-
These environmental factors may not oper- ducing its chosen quantity. This prediction
ate directly on productivity, but they can is invariant to the structure of the market in
affect producers' incentives to apply the fac- which the firm operates.
tors discussed in the previous section. Ellos The presence of spillovers is one possible
can also influence the extent to which such channel through which the external environ-
efforts are successful at moving producers to ment affects productivity levels. I discuss
a higher position within their industry's pro- situations where other firms' production
ductivity distribution, and how responsive practices influence another business's pro-
market share and survival are to productiv- ductivity level first in this section.
ity differences. That is, these external drivers A second possibility is that external drivers
can impact both the so-called “within” and influence the extent of Darwinian selection
“between” components of aggregate produc- in the firm's market. This force is highlighted
crecimiento de tivity The within component comes by the model in section 2. Environmental fac-
from individual producers becoming more tors that shift the model's exogenous param-
eficiente. The between component arises eters or the shape of the revenue function
when more efficient producers grow faster will change the minimum productivity level
than less efficient ones, or when more effi- necessary for profitable operation, A , and the
cient entrants replace less efficient exiting responsiveness of market share to productiv-
negocios. 11 ity differences. This will shift the observed
By their nature, these environmental ele- productivity distribution among the market’s
ments are also the most closely tied to gov- productores.
ernment policy. Therefore understanding Even in the absence of spillovers or selec-
these drivers merits special attention when tion, external factors can affect producers'
considering the productivity implications of incentives to raise their own productivity
market interventions. nivel. How can this be if theory says firms
Before discussing the specific external minimize costs? Well, the standard, static
drivers, it is worth taking a minute to discuss cost-minimizing firm model is an inadequate
why the operating environment should affect description of the technology adoption pro-
Cesses. A richer model like that in Thomas
J. Holmes, David K. Levine, and Schmitz
11 Many studies attempt to quantify the relative con- (2008)—who build off Kenneth J. Arrow’s
tributions of within and between effects by decompos- (1962) seminal work—points out additional
ing aggregate productivity growth into terms thatchannels reflect through which a firm's market
the separate effects. Petrin and Levinsohn (2005) have
recently raised caveats about the robustness of these environment
com- (and the competitive structure
monly used “accounting decompositions.” They advocate in particular) shifts producers' incentives
a method that focuses on measuring the gaps between thetheir productivity level. Holmes,
to raise
estimated social marginal benefits and costs of produc-
ers' inputs. Aggregate productivity grows when inputs areand Schmitz suppose that adopting
Levine,
distinct in theory
reallocated in a way
andthat
empirical
reducesimplementation
the average gap.
from
aMientras
productivity-enhancing
the costs: a temporary
disruption practice involves
period where
accounting decompositions, such “gap methods” have the
same conceptual goal: to separately measure howcostsmuchare actually higher than before any
aggregate productivity growth comes from businesses
technological change was made. Ruptura
becoming more efficient themselves and how much comes
could be due to installation issues, fine-
from reallocation of economic activity to more efficient
productores. tuning new technology, retraining workers,
Página 24
y así. With such adoption costs, produc- in their own and closely related industries,
ers facing less competition have less incen- regardless of whether they share a common
tive to adopt the new technology because the input market.
higher per-unit profits that monopoly power Any empirical search for spillovers must
brings raise the opportunity cost of changing face the classic “reflection problem” famil-
production practices. In the context of the iar to the peer effects literature: correlated
model in section 2, less competition means productivity levels among cohorts of pro-
a more concave revenue function due to ducers can be a sign of spillovers, but the
steeper residual demand curves. Esto podría correlation might also reflect the impact
arise from, for example, less scope for con- of common shocks from unobserved third
sumers to substitute across producers in the factores. Obviously, if one can observe exog-
output market. 12 enous productivity shocks for a subset of
The reality of production is also much more producers and track how related produc-
complex than even in these augmented mod- ers' productivity levels evolve in response,
els. Most technologies, even if routinized, this goes a great way towards identifying
are intricate, multifaceted processes that causalidad. Such instances can be difficult
require considerable coordination. Son to observe generally, however, and such an
constantly being buffeted by shocks to input approach cannot be used in a single cross
costs and demand-driven shifts in capacity sección. An alternative strategy is to test
requisitos. Cost-minimizing production whether the intensity of the productivity
practice is really therefore a moving target, correlation is related to some measure of
a constantly shifting ideal combination of between-producer distance, be it in geo-
operaciones. Elements of a firm's market envi- graphic, technological, or product-market
ronment can affect the firm's incentives to espacio. Higher productivity correlations
chase that moving target. among “nearby” producers are predicted by
many theories of spillovers. Este enfoque
4.1 Productivity Spillovers
is still imperfect, however, as the structure
Producer practices can have spillover of common shocks might also be related to
effects on the productivity levels of other distancia.
empresas. These externalities are often discussed Enrico Moretti (2004) explores agglomer-
in the context of classic agglomeration mech- ation-type productivity spillovers by match-
anisms like thick-input-market effects and ing the 1980 and 1990 US Population
knowledge transfers. Knowledge transfers in Censuses with the 1982 and 1992 Census
particular need not be tied to any single geo- of Manufactures by city-industry. He esti-
graphic or input market. Producers are likely mates a plant-level production function that
to attempt to emulate productivity leaders includes the share of workers in other indus-
tries in the metro area who have completed
some college. He interprets the estimated
12 A second, more roundabout mechanism also relates marginal product of this outside educated
greater competition to technology innovation and adop-
labor as a productivity spillover. Moretti
ción. If heightened competition raises the firm's probability
of exit or bankruptcy, the convexity of the firm's payoffs
finds that the marginal product of the local
created by limited liability encourages risk-takinghuman (see, for
capital measure is in fact positive.
example, Susan Rose-Ackerman 1991). In essence, compe-
Furthermore,
tition may drive desperate firms to “throw a Hail Mary” by the measured spillovers are
adopting risky but potentially productive new technologies
stronger across plants that are “close” in both
in the hope of staving off collapse. I will discuss another
the geographic and technological senses.
implication of the convexity of firm payoffs and technology
adoption below. These results are consistent with both the
Página 25
thick-input-market and knowledge-transfer market side. 13 They separately identify these

stories of productivity spillovers. two effects by comparing the impact of firms'
Several studies have focused specifically R&D (instrumented for using federal- and
on the role of knowledge transfers. On one state-level R&D tax incentives) on other
level, it seems
doubtful that they must exist. Esprac-
that productivity-enhancing firms atdistances.
market varying technological
They measure and product
technologi-
tices are completely excludable; negocios cal distance using correlations in firms' pat-
cannot always keep every facet of their pro- enting patterns and product market distance
duction process secret. Por otra parte, using the correlation in firms' sales across
the ubiquity of large and persistent produc- business segments. Because these two dis-
tivity differences within industries suggests tances are not perfectly correlated across
that any such emulation/spillover process is firms, they can measure the separate impacts
far from perfect. Frictions clearly exist that of R&D. They find that both types of spill-
prevent less efficient producers from fully overs matter but technological spillovers
replicating industry leaders' best practices. quantitatively dominate, creating a net posi-
The crucial research questions of these stud- tive externality.
ies, then, are the size of knowledge trans- Bartelsman, Haskel, and Ralf Martin
fers, what features influence this size, and (2008) make an interesting distinction
the channels through which the spillovers between global and economy-specific tech-
funcionar. nology frontiers. They show using micro-
Rachel Griffith, Rupert Harrison, and data from numerous countries that a plant’s
Van Reenen (2006) show that the geo- productivity converges faster toward the
graphic location of a firm's R&D activity productivity level of the domestic leader in
asuntos. Using patent data to pin down its industry than the global industry leader.
the historical locations of UK firms' R&D A second intriguing result is that if a plant
operations (they use presample locations falls sufficiently behind the global frontier,
to minimize endogeneity of the location of any pull toward convergence disappears, but
research activity), they find that UK firms convergence to the national frontier remains
with a greater R&D presence in the United no matter the size of the gap (conditional on
States have faster overall productivity survival, of course).
growth, and that this growth is more highly Gustavo Crespi et al. (2008) and Wolfgang
correlated with the growth of the US R&D Keller and Stephen R. Yeaple (2009) also
stock in the same industry. These patterns look at cross-border productivity conver-
are consistent with a US research presence gence. Crespi et al. focus on measuring the
making it easier for firms to tap into the information flows that could be the source of
knowledge base of the US economy, which this convergence. They combine production
tends to be the technological leader in most microdata with survey data on where firms
industrias. The precise mechanism through gather information used in their innova-
which this technology tapping occurs is tive efforts. They find that, not surprisingly,
unclear, and would be an interesting area “nearby” firms (eg, suppliers and com-
for further exploration. petitors, though less so buyers) are primary
Bloom, Schankerman, and Van Reenen
(2007) point out that spillovers can cut
two ways: technological spillovers can 13 Hans Gersbach and Armin Schmutzler (2003) dem-
benefit everyone, but there can also be onstrate how product market competition can endog-
enously determine the extent of knowledge spillovers via
market-stealing effects on the product labor mobility.
Página 26
fuentes; that much of this information, par- The second mechanism acts through effi-
ticularly from competitors, is free (though ciency increases within plants or firms. Como
surely not given freely ); and that having a discussed above, heightened competition
multinational presence aids these flows. can induce firms to take costly productivity-
Keller and Yeaple (2009) tie productivity raising actions that they may otherwise not.
growth among publicly traded US firms to Besides raising producers' own productiv-
foreign direct investment in those firms' sec- ity levels, this effect of competition leads to
tors by foreign-owned multinationals. FDI- aggregate productivity growth via the “within”
driven spillovers account for a substantial componente. There is a Schumpeterian caveat
portion of productivity growth, especially in to this within-effect of competition, however.
high-tech sectors. As Xavier Vives (2007) points out, under cer-
These papers and others like them suggest tain conditions, heightened competition (at
that spillovers exist and operate through vari- least for a market of fixed size) can actually
ous mechanisms, though again the observed diminish a firm's incentives to make produc-
productivity dispersion also makes clear that tivity-enhancing investments.
substantial frictions to the diffusion and rep- Because of the substantial literature built
lication of best practices remain. Políticas around the productivity impacts of trade
meant to increase such spillovers must be competition, I discuss it in a separate subsec-
careful, however, to not destroy firms' incen- tion below. I first cover general competitive
tives to innovate. If spillover-enhancing poli- efectos.
cies make it too hard for firms to appropriate
4.2.1 Intramarket Competition
the benefits of their innovation, the policies
could do more damage than good in the long A general indicator that product-market
correr. competition is enhancing productivity is a
4.2 Competition positive correlation between productivity and
producer growth and survival. Such correla-
Pressures from threatened or actual com- tions have been a robust finding in the litera-
petitors can affect productivity levels within ture; Foster, Haltiwanger, and Krizan (2001)
an industry. Competition drives productivity offer a broad-based overview, for example. 14
through two key mechanisms; this section Several recent studies have looked at partic-
discusses examples of research into both. ular mechanisms through which competition
The first is Darwinian selection among leads to a Darwinian selection process.
producers with heterogeneous productiv- Syverson (2004a) investigates the connec-
ity levels. Competition moves market share tion between competition and productivity
toward more efficient (ie, lower-cost and in a case study of the ready-mixed con-
generally therefore lower-price) producers, crete industry, which is well suited for this
shrinking relatively high-cost firms/plants,
sometimes forcing their exit, and open-
ing up room for more efficient producers. 14 Foster, Haltiwanger, and Syverson (2008) point out
that these results linking selection to productivity actu-
It also raises the productivity bar that any ally reflect selection on profitability , since intraindustry
potential entrant must meet to successfully price variation caused by idiosyncratic demand differences
enter. In the static model of section 2, these across plants is buried in standard revenue-based produc-
tivity measures. They show that such demand variation is
mechanisms are summarized as an increase extremely important in explaining plant survival patterns,
in A . Such selection underlies the “between” even in their sample of plants in homogeneous-product
component of aggregate productivity growth industrias. This broader interpretation of the evidence to
include demand-side factors will be discussed further in
mentioned earlier. the following section.
Página 27
type of investigation. The industry's physi- example of productivity growth in an indus-

cally homogeneous product and very high try that is driven almost entirely by within-
transport costs make spatial differentiation effect efficiency improvements. He follows
paramount. Differences in competitiveness US iron ore mining during the period the
across markets should therefore be related industry was first facing competition from
to the density of concrete producers in the foreign producers. (Brazilian mines, spe-
mercado. It is harder for inefficient concrete cifically. I will discuss more examples of
producers to be profitable in dense markets trade-induced productivity change in a sepa-
because, if they charge the high prices neces- rate section below.) The case study shows
sary to cover their costs, customers can eas- how competition can drive existing firms to
ily shift to their more efficient competitors. improve their productivity.
This implies the productivity distribution of The US iron ore industry had been pro-
ready-mixed plants will be truncated from tected from foreign competition by the
below as density rises. This is indeed what high costs of transporting ore from its other
holds in the data. Markets with denser con- sources on the globe (eg, Australia and
struction activity have higher lower-bound Brazil). By 1980, however, increased produc-
productivity levels, higher average pro- tion from low-cost Brazilian mines brought
ductivity, and less productivity dispersion. delivered prices for Brazilian ore in the Great
(Construction density is used as an exog- Lakes region in line with delivered prices
enous shifter of concrete producer density from northern Minnesota's Mesabi Range,
because the construction sector buys almost the major ore-producing area of the United
all of the ready-mixed industry's output, yet Estados. Facing competition from abroad for
concrete accounts for only a small share of the first time, the US producers attempted
construction costs.) Syverson (2007) shows to lower costs by making drastic changes in
that these patterns of competition-driven their production operations. Schmitz shows
selection on costs are also reflected in ready- most of these changes centered on loosening
mixed prices. 15 the strict work rules in the US mines. por
Outside of manufacturing, Foster, instance, mine managers originally had very
Haltiwanger, and Krizan (2006) find that little flexibility in their ability to assign differ-
aggregate productivity growth in the US ent workers to different tasks. The initiation
retail sector is almost exclusively through of serious competition allowed the mines to
the exit of less efficient single-store firms gain back flexibility in new contracts, rais-
and by their replacement with more efficient ing their utilization of available labor and
national chain store affiliates. This evokes enabling them to shed unneeded overhead
stories surrounding the growth and competi- trabajadores. The reconfigured contracts were
tive impacts of discount retailers like Wal- extremely successful at raising productivity.
Mart and Target over the past two decades. The industry's average labor productivity had
These studies focus on the selection effect been roughly constant at two tons of ore per
of competition. Schmitz (2005) offers an worker-hour for several decades preceding
1980. By 1985, however, it had doubled to
four tons per hour. As a result, the mines were
15 Such price effects also raise an interesting point given able to remain competitive even in the face of
the common use of revenue-based productivity measures.
Namely, as competition raises the average physical continuously
(ie, falling Brazilian ore prices.
quantity-, not revenue-based) productivity level in the Other recent studies have shown these
market, it also
nue-based reduces prices.
productivity measuresThiswill
means standard
understate detailed
thereve-
true case studies appear emblematic
of much broader competitive effects that
impact of competition on average productivity levels.
Página 28
act across numerous industries and econo- 2001. European firms producing the prod-
Mies. For example, Syverson (2004b) looks ucts that saw the greatest increase in compe-
at the entire US manufacturing sector. tition responded in one of two ways. Some,
Richard Disney, Haskel, and Ylva Heden particularly those using low-tech production
(2003a, 2003b) and the studies described methods, shrank and exited. Others, how-
in UK Office of Fair Trading (2007) show ever, innovated. Their patent rates, R&D,
similar results in the United Kingdom. Y IT adoption, and TFP growth increased con-
Giuseppe Nicoletti and Scarpetta (2005) currently. In aggregate, therefore, Chinese
overview evidence across OECD countries. trade competition increased aggregate TFP
in these markets through both within- and
4.2.2 Trade Competition
between-firm (selection) effects.
As seen in Schmitz's results for the iron Multiple studies using producer microdata
ore industry, the presence—or even just the have found comparable results in other set-
threat—of imports from abroad is another tings. Examples include Marcela Eslava et
form of competitive pressure. This phe- Alabama. (2004), Marc-Andreas Muendler (2004),
nomenon is the focus of a burgeoning line Bernard, Jensen, and Schott (2006), Ana M.
of research, driven in part by the recent Fernandes (2007), and Verhoogen (2008).
theoretical trade literature focusing on The specific mechanisms through which
heterogeneous-productivity producers and trade-oriented competition is postulated
their response to trade, especially Eaton and to increase productivity do vary across the
Kortum (2002) and Melitz (2003). papers, from quality upgrading within plants
Pavcnik (2002) shows how trade liberal- to heightened selection across plants. Mary
ization during the 1970s drove productivity Amiti and Jozef Konings (2007) highlight a
growth among Chilean manufacturers. los separate mechanism through which trade
paper demonstrates that sectors facing new can increase productivity: the expansion of
import competition saw faster productivity the set (or declines in the effective price) of
growth over her 1979–86 sample period than intermediate inputs when imported inputs
sectors producing primarily nontradables. become more available. I will discuss the
Pavcnik goes on to show that these indus- input-market channel further below.
try-level gains came both from existing pro- Interestingly, despite the strong correla-
ducers raising their productivity levels (the tion between the average productivity level
within effect) and from the reallocation of of an industry's plants and that industry’s
activity away from—and sometimes, the exit trade exposure, there is less evidence of large
of—less efficient, formerly protected pro- productivity impacts on the domestic plants
ducers (the selection effect). when they begin exporting. That is, exporters
Bloom, Mirko Draca, and Van Reenen are almost inevitably more productive than
(2011) look at how Chinese import compe- their nonexporting industry counterparts,
tition—the proverbial 800-pound gorilla in but most studies have found that this cor-
trade policy discussions—affected produc- relation largely reflects selection rather than
tivity and innovation in twelve European a causal impact of exporting on productivity.
countries between 1996 and 2007. To iden- Plants that choose to begin exporting were
tify competition's effects, they exploit the already more productive before trade. Esta
differential across-product drops in import is surprising if only because exporting firms
barriers that occurred when China became can leverage the benefits of any productiv-
part (due to its accession into the WTO) of ity gains across larger markets, raising their
the now-expired Multi Fibre Agreement in incentive to engage in innovative activities.
Página 29
That said, Van Biesebroeck (2005) and Jan Refiners were compensated for this tax by
De Loecker (2007a) document cases where quota protection from imports and govern-
exporters' productivity advantage grows after ment-imposed limits on domestic competi-
entry into the export market. (This is some- tion (antitrust law was often thrown to the
times referred to as the “learning-by-export- wind in the construction of New Deal pro-
ing” hypothesis.) Both are in somewhat grams). This transfer scheme led to the stan-
special settings, which might explain in part dard quantity distortions, but it also distorted
why they find postexport productivity growth incentives for efficient production. Agricultores
while many others have not. The postexport received a flat payment per ton of sugar
growth of Van Biesebroeck's (2005) sample contained in their beets, so their optimal
of sub-Saharan African exporters appears to response was to simply grow the largest beets
reduce their credit and contract enforcement posible. The problem is that refining larger
constraints, allowing them to undertake what beets into sugar is less efficient. As beets
were previously prohibitively costly produc- grow larger, their sugar-to-pulp ratio falls,
tivity-raising activities. Such a mechanism requiring more time and energy to extract
raises the question of whether it would apply a given amount of sugar from them. En el
to any firm that chooses to export (if so, why same time, given the restraints on compe-
wouldn't every producer do so?), or whether tition in the refined sugar market, refiners
these effects, while causal, reflect heteroge- had little incentive to improve sugar extrac-
neous treatment effects, with firms most apt tion on the margin. The combined result of
to benefit choosing to export. De Loecker these incentives is readily apparent in the
(2007a) finds that Slovenian firms that begin datos. When the Sugar Act was passed, a ton
exporting during the posttransition period of beets yielded an average of 310 pounds of
saw productivity growth after entering for- refined sugar, a figure that had been steadily
eign markets. Interestingly, firms export- rising from 215 pounds per ton in 1900. But
ing to higher-income regions saw greater this trend suddenly reversed after 1934.
productivity growth. Apparently the export Yields dropped to 280 pounds per ton by
market—not just the exporter itself—mat- 1950 and 240 pounds by 1974, the year the
ters. This raises interesting selection issues Act was repealed. Not surprisingly, yields
about which markets firms choose to export began to climb again immediately after
to, even conditional on the decision to export repeal, to about 295 pounds per ton by 2004.
in the first place. It is a sad testimony to the Act's productivity
distortions that yields seventy years after the
4.3 Deregulation or Proper Regulation
act were still lower than when it was passed.
Poorly regulated markets can create per- Christopher R. Knittel (2002) and Kira R.
verse incentives that reduce productivity. Fabrizio, Nancy L. Rose, and Wolfram (2007)
Deregulating or reformatting to smarter examine how power plant operations react
forms of regulation can reverse this. to changes in the regulatory structure they
Benjamin Bridgman, Shi Qi, and Schmitz operate under. Both studies involve moving
(2009) show how regulations in place for plants away from a traditional cost-plus reg-
decades in the US sugar market destroyed ulated monopoly structure into alternative
incentives to raise productivity. The US formas. Knittel (2002) studies the implemen-
Sugar Act, passed in 1934 as part of the tation of “incentive regulation” programs,
Depression-era restructuring of agricultural where regulators explicitly tie operators'
law, funded a subsidy to sugar beet farm- earnings to the achievement of particular
ers with a tax on downstream sugar refining. operating efficiencies. Fabrizio, Rose, and
Página 30
Wolfram look at the effect of electricity mar- Indeed, there are almost surely comple-
ket reforms that occurred in many regions mentarities between product market and
in the United States during the 1990s. Ambos input market flexibility. If consumers want
studies find that plants experienced effi- to reallocate their purchases across produc-
ciency gains after the shift in the regulatory ers, firms that experience growth in demand
ambiente. Fabrizio, Rose, and Wolfram for their products will need to hire additional
also show that, in line with what one would inputs to meet that demand. The more easily
expect, the productivity gains were largest inputs can move toward these firms, which
among investor-owned utilities and smallest will typically be higher-productivity busi-
in municipally operated utilities. nesses due to the forces described above,
Beyond these case studies, recent work the faster and more smoothly the realloca-
has also taken a broader look at how prod- tion mechanism works. In the context of the
uct market regulations impact productivity model in section 2, flexible input markets
at the micro level. For example, Michael reduce the concavity of the revenue func-
Greenstone, List, and Syverson (2011) show tion, making producer size more responsive
how environmental regulations (the US to productivity differences. This section dis-
Clean Air Act Amendments specifically) cusses recent research tying factor market
reduce manufacturing plants' productivity flexibility to productivity.
niveles. Nicoletti and Scarpetta (2005) and The institutional features of input mar-
Jens Arnold, Nicoletti, and Scarpetta (2008) kets, such as the roles of unions and the
discuss the productivity effects of product- structure of the financial sector, have an
market regulations in OECD economies. ambiguous theoretical impact on flexibil-
A related yet distinct relation between ity If institutions improve match efficiency,
legal structure and productivity is how solve asymmetric information problems, or
privatization affects formerly state-owned otherwise serve efficiency-enhancing roles,
empresas. J. David Brown, John S. Earle, and they make input markets more flexible. Si
Almos Telegdy's (2006) study of formerly they facilitate rent-seeking behavior on the
state-owned
European enterprises
countries in of
is one several Eastern
the more com- otherthe
end, hand, theyofimpede
impact flexibility.
any particular En el
institution
prehensive of such studies. They document is an empirical question—one which several
broad-based productivity growth in plants of the studies in this section investigate.
after privatization but they also find consid- Maksimovic and Phillips (2001) inves-
erable variation in the size of the impacts tigate the market for US manufacturing
across countries, with more than 15 per- plants themselves, as productive assets. Ellos
cent average TFP growth in Romania but a measure how a plant's productivity changes
slightly negative impact in Russia. when it is sold by one firm to another. Ellos
find that, on average, a plant's productivity
4.4 Flexible Input Markets
rises after the sale. That is reassuring: the
I discussed above how competition market tends to allocate inputs in an efficient
increases productivity. If one thinks of com- way, instead of as a response to ambitions of
petition as flexibility in product markets— empire-building managers or other ineffi-
in more competitive markets, it's easier for cient motives. Another of their findings that
consumers to shift their purchases from one is consistent with this efficiency-enhancing
producer to another—it is logical to suppose role is that the plants that are sold tend to
that flexible input markets might also raise come from the selling firm's less productive
niveles de productividad. business lines. In essence, the sellers are
Página 31
moving away from activities at which they revenue TFP levels are too small (large) rela-
are less proficient. tive to an allocatively efficient benchmark. dieciséis
Petrin and Sivadasan (2010) use a novel After measuring these implied plant-level
approach to look at the productivity effects distortions, they compare their distribution
of labor market flexibility. They measure the to the analogous distribution measured in
difference between Chilean plants' marginal US microdata. (This is used as the compari-
products of labor (as derived from industry- son rather than the first-best/zero-distortion
level production functions they estimate) outcomes because it is a more realistic con-
and their average wages. Such gaps can be trol group. The US data contain, and hence
caused by any one of a number of market can be used to control for, gaps that reflect
distortions, like market power, taxes, or the adjustment costs and measurement error
firing costs that are the object of the study. that may be immutable to policy action.)
Allocative efficiency is achieved, at least in Hsieh and Klenow find that Chinese aggre-
the cross section, when this gap is equated gate TFP could increase by 30–50 percent
across plants. (Though of course overall inef- and Indian TFP by 40–60 percent by achiev-
ficiencies still exist unless these gaps are all ing the US level of allocative efficiency with
zero.) Efficiency increases if labor inputs are their existing resources.
moved from low- to high-gap plants because Bartelsman, Haltiwanger, and Scarpetta
the net change in marginal product caused (2009) look at the success of allocation across
by the input shift outstrips the change in several countries. Rather than using a gap-
wage costs. Petrin and Sivadasan find that a type methodology like Hsieh and Klenow,
particular legislative change that raised firing they measure efficiency using the corre-
costs was associated with an increase in the lation between a plant's share of industry
mean gap, suggesting the legislation reduced output and its productivity level. The logic
allocative efficiency. of this metric is straightforward and similar
Several recent papers have taken these to that in the model in section 2 and what
ideas and asked whether, more broadly was discussed at the beginning of the com-
speaking, economies efficiently allocate petition section. Well functioning markets
inputs across heterogeneous production should reallocate output to more produc-
unidades. Hsieh and Klenow (2009) use the tive plants, leading to a positive correlation
measured TFP dispersion across Chinese between output share and productivity. Un
and Indian firms to infer the size of pro- additional advantage of the metric is that it is
ducer-level distortions that jointly depress easy to compute. Its limitation is that it is an
aggregate productivity in those economies. accounting decomposition and, as such, is not
Their methodology is conceptually similar to directly tied to welfare theory the way gap-
Petrin and Sivadasan's gap approach. Su type measures are. However, Bartelsman,
model indicates that in the absence of dis-
tortions, plants' revenue-based TFP levels
(TFP measured using revenues as an output 16 Their model's implication of equal revenue TFP
across plants stems from the standard efficiency condition
measure rather than quantities) should be that inputs' marginal revenue products are equated across
igual. This implies that observed deviations all uses, and the fact that marginal products are propor-
from this equality reflect the presence of dis- tional to average products for a Cobb–Douglas produc-
tion function without fixed costs. Since TFP is an average
tortions. (Note, however, that quantity-based product measure, equal marginal revenue products implies
TFP values are not equated even if there are equal average revenue products and therefore equal rev-
no distortions.) Essentially, their framework enue TFP. Non-Cobb–Douglas technologies and/or fixed
costs can also support persistent revenue TFP differences
implies that plants with relatively high (low) aside from any distortions.
Página 32
Haltiwanger, and Scarpetta show, in a simple high “productivity” businesses may not be
model, how various types of producer-level particularly technologically efficient. Mucho
distortions do in fact lead to reductions in of the literature described above therefore
the output–productivity correlation within documents the joint influence of produc-
an industry. tivity and demand factors that show up in
within-industry price variation.
A new strand of research has begun to
5. Big Questions
extend the productivity literature to explic-
That is a brief summary of what we know itly account for such idiosyncratic demand
about the causes of productivity differences effects as well. These new frameworks—see
at the micro level and why we would want to Sanghamitra Das, Roberts, and James R.
know these causes. I want to emphasize that Tybout (2007), Eslava et al. (2008), Foster,
while the discussion draws out major themes Haltiwanger, and Syverson (2008, 2010), and
of that body of knowledge, it really only just De Loecker (2007b) for example—allow an
scratches the surface of the literature. additional and realistic richness in the mar-
I think a fair reading of the discussion above ket forces that determine producers' fates.
would say that we have learned a lot about The work to this point indicates that demand
productivity since the Bartelsman and Doms factors are indeed important. They exert a
(2000) survey. At the same time, it is hardly considerable influence on businesses' growth
time to declare victory and go home. Muchos and survival. And while many of the basic
pressing issues and open questions remain. En results above that have been checked after
this section, I will briefly lay out what I see adjusting for the supply–demand dichot-
to be the major questions about productivity omy have been robust, the results do sug-
that the research agenda should address. 17 gest some reinterpretations of productivity
effects as inferred from standard measures.
5.1 What Is the Importance of Demand?
The scope of issues that this new line of
Productivity is typically thought of as a research has addressed is still small, how-
supply-side concept. As discussed in sec- nunca. Demand could play an important role
tion 2, it is the component of the produc- in many more settings that have been hid-
tion function unrelated to observable labor, den to this point due to measurement issues.
capital, and intermediate inputs. But pro- This is likely to be especially true when
ductivity as actually measured in producer moving to sectors without well defined out-
microdata generally reflects more than just puts (what exactly does Google produce, for
supply-side forces. Because producer-spe- example, and how should it be measured?).
cific prices are unobserved in most business- Unwinding this knot is a top priority.
level microdata, output is typically measured
5.2 What Is the Role of (or Hope for)
by revenue divided by an industry-level
Government Policies That Encourage
deflator. This means that within-industry
Productivity Growth?
price differences are embodied in output
and productivity measures. If prices reflect Clearly, many of the productivity drivers
in part idiosyncratic demand shifts or mar- discussed above can be influenced by gov-
ket power variation across producers—a ernment policies. Esto es especialmente cierto de
distinct likelihood in many industries—then the “external” drivers in the previous sec-
tion—the elements of the market environ-
ment that can induce business to take actions
17 Conversations with Haltiwanger were very helpful in
writing this section. to raise their productivity or that affect the
Página 33
Darwinian selection process that whittles out Of course, it's quite likely that the quantita-
inefficient producers. tive impact of factors varies across industries
Several policy-related questions are or markets. A concomitant question, then, is
prime targets for research. Allí tienen which factors matter most in what sectors?
been many policy reforms (particularly in Research that ties observable attributes of
trade policy and market regulation design) the industry's technology or demand struc-
that had plausibly productivity-enhancing ture to the quantitative importance of pro-
efectos. Many studies have evaluated spe- ductivity-influencing factors would be an
cific reforms in isolation, taking the policy incredible advance in our ability to explain
change as given. But a policy change, even productivity growth.
one that moves in the right direction, may
5.4 What Factors Determine Whether
not necessarily
reforms, besize
either in optimal. Alternativa
or approach, might Selection or Within-Producer Growth
Is More Important in a Market/Sector/
be more cost effective. Research has typi-
Industry?
cally compared the effects of policy reforms
to a null of no reform, but perhaps an In many settings above, there was a
equally important comparison is among prominent distinction between aggregate
possible reform alternatives. What type of productivity growth coming from “within”
reform is most effective for a given type of (productivity growth at a given plant or firm)
market or friction? What is the optimal size and “between” (reallocation-based selection
and timing of policy changes? Estos son los across existing businesses or entry and exit)
next set of questions the literature should fuentes. Just as the literature still needs to
chase in this area. characterize the relative quantitative contri-
A related issue is why reforms, even if they bution of various influences on producer-level
are welfare enhancing in their productivity efficiency, so too does it need to measure the
effects, don't always happen. There could relative importance of within and between
be economic reasons for this. Established components in explaining aggregate produc-
interests could be earning rents in the unre- crecimiento de tivity
formed environment. Ellos pueden ser capaces de We do know some patterns already. por
stave off reform, especially if its benefits example, aggregate productivity growth in
are diffuse while its losses are concentrated. the retail sector seems to be almost exclu-
Characterizing the nature of these barriers sively from reallocation, at least in the
to aggregate productivity gains—who wins, Estados Unidos. But of course the literature
who loses, and by how much—could be has covered nowhere near the full span of
fruitful. sectors and economies. More importantly,
we do not yet have a good model of what
5.3 Which Productivity Drivers Matter
sectoral features (again on either the sup-
Most?
ply or demand side) might determine
The research described above has framed the relative importance of each. Por que es
which factors might explain variation in pro- within-store productivity growth so small
ductivity levels. The relative quantitative on average in retail, but not manufactur-
importance of each, however, is still unclear. ing, for example? Answering questions like
Summarized succinctly, if we could easily this would go a long way to developing our
measure these factors and add them to the understanding of how micro productivity
production function, which would have the differences drive the aggregate productiv-
largest R 2 ? ity movements.
Página 34
5.5 What Is the Role of Misallocation as lead to more firms taking bets on potential
a Source of Variation in Emerging productivity-increasing activities like IT
Economies? inversión. There is some evidence that this
Productivity differences explain much of is happening, but the literature has yet to
the per capita income variation across coun- show this definitively. Second, if there is an
intentos. As seen above, recent research with upward trend in productivity dispersion, will
producer microdata is building the case that a the forces of selection stem this spread? Si
substantial portion of these productivity gaps so, when? Will a shakeout be strong enough
arise from poor allocation of inputs across to drive dispersion back to its previous level?
production units in developing countries. Third, is this increase in variance something
In some ways, this is a hopeful finding: specific about IT capital, or is it a broader
these countries could become substantially feature of general purpose technologies?
more productive (and raise their incomes) Historical evidence would be very informa-
by simply rearranging the inputs they already tive here. For example, did the diffusion of
tener. Not everything hangs on some unat- the electric motor in the early twentieth cen-
tainable technologies that are out of reach. tury also increase in the variance in produc-
On the other hand, the result also has tivity outcomes across businesses? Or even
discouraging elements. While research has when a particular industry experiences a rev-
identified misallocation as a source of the olution in its standard technology, does this
problem, it hasn't really pinned down exactly lead to temporary increases in productivity
what distortions create gaps between the dispersion followed by a shakeout?
social marginal benefits and costs of inputs 5.7 Can We Predict Innovation Based on
across production units. It is hard to imple- Market Conditions?
ment policies that close these gaps and the
variation between them (ie, reallocate Here I speak of innovation broadly—
inputs more efficiently) without knowing the product and process innovation, measured
nature of the gaps in the first place. or unmeasured by formal R&D numbers.
That said, there has been some early prog- This question is in some ways a corollary to
ress on this front. Witness the efforts to tie the one above about quantifying and pre-
misallocation to various labor market poli- dicting the split between within-producer
and between-producer productivity growth.
cies.isMuch
this remains area
an important to befor
done, however,
further effort.and Within-productivity growth is in many cases
not simply the passive accumulation of effi-
ciencia; it comes in part as a result of the active
5.6 What Is the Importance of Higher
innovative efforts of producers. What market
Variance in Productivity Outcomes?
or technological factors determine how large
Some of the work above, particularly innovative activity will be? Can we predict
that focusing on the role of IT capital, sug- whether product or process innovation will
gests that the variance of productivity out- dominate, based on market features?
comes might be increasing at a very broad
5.8 The Nature of Intangible Capital
nivel. This has several implications. Primero,
the operation of a business is a call option: Many of the primary drivers of productiv-
poor outcomes are truncated because of the ity naturally create persistence in productiv-
possibility of exit. The value of this option ity levels at plants and firms. Éstas incluyen
increases with a mean-preserving spread in learning-by-doing; innovative efforts; y
resultados. As such, higher variance should in many cases investment in higher quality
Página 35
managerial, labor, or capital inputs. An easy causal impact might be limited. En el otro
way to explain such persistence is to think of extreme, if managers don't seem to matter
these productivity enhancements as resulting at all, then it is quite likely that managerial
from producers' investments in intangible practices have a strong causal impact on
capital—know-how about their businesses productividad.
that is embodied in the organization. Esta
5.10 A Plea for Data
conceptual structure also highlights how
productivity gains sourced in intangible Data availability is not a research question,
capital can also be thought of, along with but it is crucial for answering the questions
managerial and unobserved factor quali- posed above. Virtually everything discussed
ties, as arising from mismeasured inputs. Si in this survey we now know because detailed
one really could measure intangible capital data on production practices was available.
(which, alas, is inherently difficult given its But many of these datasets were originally
nature), the productivity differences arising collected by statistical agencies for the pur-
from such sources could be explained. pose of constructing aggregates. Their abil-
Understanding how such intangible capi- ity to offer insights into what happens at
tal stocks are built and sustained would shed the micro level was in many ways a happy
light on many productivity-related issues for externality. Now that we know the value of
this reason. Such insights would also speak the knowledge that such information can
toward active literatures on the subject in generate, economists should push for more
macroeconomics and finance. How much directed efforts to measure business-level
uncertainty is inherent in intangible capital production practices. This could include, for
investment? What is the distribution of rates example, more data on managers and man-
of return across producers, and what predicts agement practices, business-level prices,
¿ellos? Is intangible capital fully excludable input quality measures, proxies for intangi-
or are there spillovers to other firms? Cómo ble capital, non-R&D innovation spending,
well do R&D measures capture investment y así. Obviously, collecting such data is
in intangibles? Are there other proxies that costly, and this sort of push will involve trad-
could augment such measures? eoffs for statistical agencies or a willingness
of researchers to pay private companies for
5.9 Management Versus Managers
the collection efforts. Nevertheless, it seems
We know more about the role of manage- clear that there is much to be gained in
ment than before, but what about managers ? exchange for those costs.
Some good work on CEOs aside, we don't
really know if good managerial practices
6. Conclusion
matter enough to attain productivity gains
or whether they are complementary to the The research into the productivity differ-
skills of those who implement them. Si ellos ences across businesses has come a long way
are complements, what skills matter? Son since Bartelsman and Doms (2000) surveyed
they built by experience, tenure in the indus- the literature a decade ago. We know more
try or on the job, education, or something about what causes the measured differences
¿más? Understanding these issues might also in productivity, and how factors both inter-
help to pin down the causal nature of man- nal and external to the plant or firm shape
agement practices. If good management the distribution. These insights have been
practices reflect in large part the fact that applied to research questions in numerous
they are what good managers do, then the campos.
Página 36
That said, there is still plenty to be learned.Aw, Bee Yan, Mark J. Roberts, and Daniel Yi Xu. 2008.
Fortunately, I see no sign that the rate at “R&D Investments, Exporting, and the Evolution
which researchers accumulate knowledge in of Firm Productivity.” American Economic Review ,
98(2): 451–56.
this area is slowing. I am excited to see what Balasubramanian, Natarajan, and Jagadeesh Siv-
the next several years bring in this research adasan. 2011. “What Happens When Firms Patent?
agenda, as the content of the next decade’s New Evidence from US Economic Census Data.”
Review of Economics and Statistics , 93(1): 126–46.
survey unfolds. Bandiera, Oriana, Iwan Barankay, and Imran Rasul.
2007. “Incentives for Managers and Inequality among
Workers: Evidence from a Firm-Level Experiment.”
Referencias Quarterly Journal of Economics , 122(2): 729–73.
Bandiera, Oriana, Iwan Barankay, and Imran Rasul.
Abowd, John M., John Haltiwanger, Ron Jarmin, Julia2009. “Social Connections and Incentives in the
Lane, Paul Lengermann, Kristin McCue, Kevin Workplace: Evidence from Personnel Data.” Econo-
McKinney, and Kristin Sandusky. 2005. “The Rela- metrica , 77(4): 1047–94.
Bartel, Ann, Casey Ichniowski, and Kathryn Shaw.
tion among Human Capital, Productivity, and Market
Value: Building Up from Micro Evidence.” In Mea- 2007. “How Does Information Technology Affect
suring Capital in the New Economy , ed. Carol Cor-Productivity? Plant-Level Comparisons of Prod-
rado, John Haltiwanger, and Daniel Sichel, 153–98.uct Innovation, Process Improvement, and Worker
Chicago and London: University of Chicago Press. Skills.” Quarterly Journal of Economics , 122(4):
Abowd, John M., Francis Kramarz, and David N. Mar- 1721–58.
golis. 1999. “High Wage Workers and High Wage Bartelsman, Eric J., and Mark Doms. 2000. “Under-
Firms.” Econometrica , 67(2): 251–333. standing Productivity: Lessons from Longitudinal
Ábrahám, Árpád, and Kirk White. 2006. “The Dynam- Microdata.” Journal of Economic Literature , 38(3):
ics of Plant-Level Productivity in US Manufactur-569–94.
ing.” Center for Economic Studies Working Paper Bartelsman, Eric J., Pieter A. Gautier, and Joris de
06-20. Wind. 2010. “Employment Protection, Technology
Acemoglu, Daron, Philippe Aghion, Claire Lelarge,Choice, and Worker Allocation.” Tinbergen Institute
John Van Reenen, and Fabrizio Zilibotti. 2007. Discussion Paper 2010-042/3.
“Technology, Information, and the Decentraliza-Bartelsman, Eric J., John Haltiwanger, and Stefano
tion of the Firm.” Quarterly Journal of Economics Scarpetta.
, 2009. “Cross-Country Differences in
122(4): 1759–99. Productivity: The Role of Allocation and Selection.”
Acemoglu, Daron, and Joshua Linn. 2004. “MarketNational Bureau of Economic Research Working
Size in Innovation: Theory and Evidence from thePaper 15490.
Pharmaceutical Industry.” Quarterly Journal of Bartelsman,
Eco- Eric J., Jonathan E. Haskel, and Ralf Mar-
nomics , 119(3): 1049–90. estaño. 2008. “Distance to Which Frontier? Evidence on
Ackerberg, Daniel, C. Lanier Benkard, Steven Berry,Productivity Convergence from International Firm-
and Ariel Pakes. 2007. “Econometric Tools for Ana-Level Data” Centre for Economic Policy Research
lyzing Market Outcomes.” In Handbook of Econo- Discussion Paper 7032.
metrics, Volume 6A , ed. James J. Heckman andBasu, Susanto, Luigi Pascali, Fabio Schiantarelli,
Edward E. Leamer, 4171–276. Amsterdam and Bos- and Luis Serven. 2009. “Productivity, Welfare and
ton: Elsevier, North-Holland. Reallocation: Theory and Firm-Level Evidence.”
Amiti, Mary, and Jozef Konings. 2007. “Trade Liberal-
National Bureau of Economic Research Working
Paper 15579.
ization, Intermediate Inputs, and Productivity: Evi-
dence from Indonesia.” American Economic Review Benkard,
, C. Lanier. 2000. “Learning and Forgetting:
97(5): 1611–38. The Dynamics of Aircraft Production.” American
Arnold, Jens, Giuseppe Nicoletti, and Stefano Scar-Economic Review , 90(4): 1034–54.
petta. 2008. “Regulation, Allocative Efficiency Bernard,
and Andrew B., J. Bradford Jensen, and Peter K.
Productivity in OECD Countries: Industry and Schott. 2006. “Trade Costs, Firms and Productivity.”
Firm-Level Evidence.” Organisation for Economic Journal of Monetary Economics , 53(5): 917–37.
Co-operation and Development Economics Depart- Bernard, Andrew B., Stephen J. Redding, and Peter
ment Working Paper 616. K. Schott. 2010. “Multiple-Product Firms and Prod-
Arrow, Kenneth J. 1962. “Economic Welfare and the uct Switching.” American Economic Review , 100(1):
70–97.
Allocation of Resources for Invention.” In The Rate
and Direction of Inventive Activity: Economic andBertrand, Marianne, and Antoinette Schoar. 2003.
“Managing with Style: The Effect of Managers on
Social Factors , 609–26. Princeton: Princeton Univer-
sity Press. Firm Policies.” Quarterly Journal of Economics ,
118(4): 1169–1208.
Asplund, Marcus, and Volker Nocke. 2006. “Firm Turn-
over in Imperfectly Competitive Markets.” Review Bloom,
of Nicholas, Mirko Draca, and John Van Reenen.
Economic Studies , 73(2): 295–327. 2011. “Trade Induced Technical Change? The Impact
Página 37
of Chinese Imports on Innovation, IT and Productiv-

Dynamics.” Review of Economic Dynamics , 7(2):
ity.” National Bureau of Economic Research Work-331–53.
ing Paper 16717. Caves, Douglas W., Laurits R. Christensen, and W.
Bloom, Nicholas, Benn Eifert, Aprajit Mahajan, David
Erwin Diewert. 1982. “The Economic Theory of
McKenzie, and John Roberts. 2011. “Does Man- Index Numbers and the Measurement of Input,
agement Matter? Evidence from India.” National Output, and Productivity.” Econometrica , 50(6):
Bureau of Economic Research Working Paper 16658.1393–1414.
Bloom, Nicholas, Christos Genakos, Raffaella Sadun,
Collard-Wexler, Allan. 2010. “Demand Fluctuations in
and John Van Reenen. 2010. “Does Manage- the Ready-Mix Concrete Industry.” Unpublished.
ment Matter: New Empirics and Old Theories.”Cooper, William W., Lawrence M. Seiford, and Kaoru
Unpublished. Tone. 2006. Introduction to Data Envelopment Anal-
Bloom, Nicholas, Raffaella Sadun, and John Van ysis and Its Uses: With DEA-Solver Software and
Reenen. 2009. “The Organization of Firms acrossReferences . Nueva York: Springer.
Countries.” National Bureau of Economic Research
Crespi, Gustavo, Chiara Criscuolo, Jonathan E. Haskel,
Working Paper 15129. and Matthew Slaughter. 2008. “Productivity Growth,
Bloom, Nicholas, Raffaella Sadun, and John Van Knowledge Flows, and Spillovers.” National Bureau
Reenen. Forthcoming. “Americans Do IT Better: of Economic Research Working Paper 13959.
US Multinationals and the Productivity Miracle.”
Cummins, Jason G., and Giovanni L. Violante. 2002.
American Economic Review . “Investment-Specific Technical Change in the
Bloom,
Reenen.Nicholas, Mark Schankerman,
2007. “Identifying andSpillovers
Technology John VanUnited StatesConsequences.”
roeconomic (1947–2000): Measurement and Mac-
Review of Economic
and Product Market Rivalry.” National Bureau ofDynamics , 5(2): 243–84.
Economic Research Working Paper 13060. Das, Sanghamitra, Mark J. Roberts, and James R.
Bloom, Nicholas, and John Van Reenen. 2007. Tybout. 2007. “Market Entry Costs, Producer
“ Measuring and Explaining Management Practices Heterogeneity, and Export Dynamics.” Economet-
across Firms and Countries.” Quarterly Journal of rica , 75(3): 837–73.
Economics , 122(4): 1351–1408. De Loecker, Jan. 2007a. “Do Exports Generate Higher
Bloom, Nicholas, and John Van Reenen. 2010. “Why Productivity? Evidence from Slovenia.” Journal of
Do Management Practices Differ across Firms and International Economics , 73(1): 69–98.
Countries?” Journal of Economic Perspectives De , 24(1):
Loecker, Jan. 2007b. “Product Differentiation,
203–24. Multi-product Firms and Estimating the Impact
Blundell, Richard, and Stephen R. Bond. 2000. “GMM of Trade Liberalization on Productivity.” National
Estimation with Persistent Panel Data: An Applica- Bureau of Economic Research Working Paper 13155.
tion to Production Functions.” Econometric Reviews
Disney,, Richard, Jonathan E. Haskel, and Ylva Heden.
19(3): 321–40. 2003a. “Entry, Exit and Establishment Survival in
Boning, Brent, Casey Ichniowski, and Kathryn Shaw. UK Manufacturing.” Journal of Industrial Econom-
2007. “Opportunity Counts: Teams and the Effec-ics , 51(1): 91–112.
tiveness of Production Incentives.” Journal of Labor
Disney, Richard, Jonathan E. Haskel, and Ylva Heden.
Economics , 25(4): 613–50. 2003b. “Restructuring and Productivity Growth in
Bridgman, Benjamin, Shi Qi, and James A. Schmitz. UK Manufacturing.” Economic Journal , 113(489):
2009. “The Economic Performance of Cartels: Evi- 666–94.
dence from the New Deal US Sugar Manufacturing Doraszelski, Ulrich, and Jordi Jaumandreu. 2009.
Cartel, 1934–74.” Federal Reserve Bank of Minne- “R&D and Productivity: Estimating Endogenous
apolis Staff Report 437. Productivity.” Unpublished.
Brown, J. David, John S. Earle, and Almos Telegdy.Eaton, Jonathan, and Samuel Kortum. 2002. “Technol-
2006. “The Productivity Effects of Privatization: ogy, Geography, and Trade.” Econometrica , 70(5):
Longitudinal Estimates from Hungary, Romania, 1741–79.
Russia, and Ukraine.” Journal of Political Economy , Richard, and Ariel Pakes. 1995. “Markov-Per-
Ericson,
114(1): 61–99. fect Industry Dynamics: A Framework for Empirical
Brynjolfsson, Erik, Andrew McAfee, Michael Sorell, Work.” Review of Economic Studies , 62(1): 53–82.
and Feng Zhu. 2008. “Scale without Mass: BusinessEslava, Marcela, John Haltiwanger, Adriana Kugler,
Process Replication and Industry Dynamics.” Har- and Maurice Kugler. 2004. “The Effects of Structural
vard Business School Working Paper 07-016. Reforms on Productivity and Profitability Enhancing
Bushnell, James B., and Catherine D. Wolfram. 2009. Reallocation: Evidence from Colombia.” Journal of
“The Guy at the Controls: Labor Quality and Power Development Economics , 75(2): 333–71.
Plant Efficiency.” In International Differences in the Marcela, John Haltiwanger, Adriana Kugler, and
Eslava,
Business Practices and Productivity of Firms , ed.Maurice Kugler. 2008. “Plant Survival, Market Fun-
Richard B. Freeman and Kathryn L. Shaw, 79–102. damentals and Trade Liberalization.” Unpublished.
Chicago and London: University of Chicago Press. Fabrizio, Kira R., Nancy L. Rose, and Catherine D.
Campbell, Jeffrey R., and Jonas DM Fisher. 2004. Wolfram. 2007. “Do Markets Reduce Costs? Assess-
“Idiosyncratic Risk and Aggregate Employment ing the Impact of Regulatory Restructuring on US
Página 38
Electric Generation Efficiency.” American EconomicEconometrics and Economic Theory in the Twentieth
Review , 97(4): 1250–77. Century: The Ragnar Frisch Centennial Symposium ,
Faggio, Giulia, Kjell G. Salvanes, and John Van Reenen.
ed. Steinar Strøm, 169–203. Cambridge; Nueva York
2010. “The Evolution of Inequality in Productivityand Melbourne: Cambridge University Press.
and Wages: Panel Data Evidence.” Industrial and Haltiwanger, John, Stefano Scarpetta, and Helena Sch-
Corporate Change , 19(6): 1919–51. weiger. 2008. “Assessing Job Flows across Countries:
Fernandes, Ana M. 2007. “Trade Policy, Trade Volumes
The Role of Industry, Firm Size and Regulations.”
and Plant-Level Productivity in Colombian Manu- National Bureau of Economic Research Working
facturing Industries.” Journal of International Eco-
Paper 13920.
nomics , 71(1): 52–71. Hamilton, Barton H., Jack A. Nickerson, and Hideo
Forbes, Silke J., and Mara Lederman. 2011. “Does Owan.
Ver- 2003. “Team Incentives and Worker Het-
tical Integration Affect Firm Performance? Evidencia
erogeneity: An Empirical Analysis of the Impact of
from the Airline Industry.” Rand Journal of Econom-
Teams on Productivity and Participation.” Journal of
ics , 41(4): 765–90. Political Economy , 111(3): 465–97.
Foster, Lucia, John Haltiwanger, and CJ Krizan. Holmes, Thomas J., David K. Levine, and James A.
2001. “Aggregate Productivity Growth: Lessons fromSchmitz. 2008. “Monopoly and the Incentive to
Microeconomic Evidence.” In New DevelopmentsInnovatein When Adoption Involves Switchover Dis-
Productivity Analysis , ed. Charles R. Hulten, Edwin
ruptions.” National Bureau of Economic Research
R. Dean, and Michael J. Harper, 303–63. ChicagoWorking Paper 13864.
and London: University of Chicago Press. Hopenhayn, Hugo A. 1992. “Entry, Exit, and Firm
Foster, Lucia, John Haltiwanger, and CJ Krizan. 2006.
Dynamics in Long Run Equilibrium.” Econometrica ,
“Market Selection, Reallocation, and Restructuring60(5): 1127–50.
in the US Retail Trade Sector in the 1990s.” Review
Hortaçsu, Ali, and Chad Syverson. 2007. “Cement-
of Economics and Statistics , 88(4): 748–58. ing Relationships: Vertical Integration, Foreclosure,
Foster, Lucia, John Haltiwanger, and Chad Syverson.Productivity, and Prices.” Journal of Political Econ-
2008. “Reallocation, Firm Turnover, and Efficiency:
omy , 115(2): 250–301.
Selection on Productivity or Profitability?” American
Hortaçsu, Ali, and Chad Syverson. 2011. “Why Do
Economic Review , 98(1): 394–425. Firms Own Production Chains?” Unpublished.
Foster, Lucia, John Haltiwanger, and Chad Syverson.
Hsieh, Chang-Tai, and Peter J. Klenow. 2009. “Misallo-
2010. “The Slow Growth of New Plants: Learningcation and Manufacturing TFP in China and India.”
about Demand?” Unpublished. Quarterly Journal of Economics , 124(4): 1403–48.
Fox, Jeremy T., and Valérie Smeets. 2011. “Does Hubbard,
Input Thomas N. 2003. “Information, Decisions,
Quality Drive Measured Differences in Firm Pro-and Productivity: On-Board Computers and Capac-
ductivity?” National Bureau of Economic Research ity Utilization in Trucking.” American Economic
Working Paper 16853. Review , 93(4): 1328–53.
Galindo-Rueda, Fernando, and Jonathan E. Haskel. Hulten, Charles R. 2009. “Growth Accounting.”
2005. “Skills, Workforce Characteristics and Firm-National Bureau of Economic Research Working
Level Productivity in England.” Report prepared Paper 15341.
for the Department of Trade and Industry, Depart-
Ichniowski, Casey, and Kathryn Shaw. 2003. “Beyond
ment for Education and Skills, Office for NationalIncentive Pay: Insiders' Estimates of the Value of
Statistics. Complementary Human Resource Management
Garicano, Luis, and Paul Heaton. 2007. “InformationPractices.” Journal of Economic Perspectives , 17(1):
Technology, Organization, and Productivity in the155–80.
Public Sector: Evidence from Police Departments.”
Ichniowski, Casey, Kathryn Shaw, and Giovanna Pren-
Centre for Economic Performance Discussion Paper nushi. 1997. “The Effects of Human Resource Man-
826.
Gersbach, Hans, and Armin Schmutzler. 2003. “Endog-agement Practices
Finishing on Productivity:
Lines.” American EconomicA Study of ,Steel
Review 87(3):
enous Technological Spillovers: Causes and Conse- 291–313.
quences.” Journal of Economics and Management Ilmakunnas, Pekka, Mika Maliranta, and Jari Vaini-
Strategy , 12(2): 179–205. omäki. 2004. “The Roles of Employer and Employee
Greenstone, Michael, John A. List, and Chad Syver- Characteristics for Plant Productivity.” Journal of
hijo. 2011. “The Effects of Environmental Regula- Productivity Analysis , 21(3): 249–76.
tion on the Competitiveness of US Manufacturing.”
Jorgenson, Dale W., Mun S. Ho, and Kevin J. Stiroh.
Unpublished. 2005. Productivity. Volume 3. Information Technol-
Griffith, Rachel, Rupert Harrison, and John Van ogy and the American Growth Resurgence . Leva-
Reenen. 2006. “How Special is the Special Relation-bridge and London: MIT Press.
ship? Using the Impact of US R&D Spillovers on Jorgenson, Dale W., Mun S. Ho, and Kevin J. Stiroh.
UK Firms as a Test of Technology Sourcing.” Amer- 2008. “A Retrospective Look at the US Productivity
ican Economic Review , 96(5): 1859–75. Growth Resurgence.” Journal of Economic Perspec-
Griliches, Zvi, and Jacques Mairesse. 1998. “Produc-tives , 22(1): 3–24.
tion Functions: The Search for Identification.” In
Jovanovic, Boyan. 1982. “Selection and the Evolution
Página 39
of Industry.” Econometrica , 50(3): 649–70. Nicoletti, Giuseppe, and Stefano Scarpetta. 2005.
Kaplan, Steven N., Mark M. Klebanov, and Morten“Regulation and Economic Performance: Product
Sorensen. 2008. “Which CEO Characteristics andMarket Reforms and Productivity in the OECD.”
Abilities Matter? National Bureau of Economic Organisation for Economic Co-operation and Devel-
Research Working Paper 14195. opment Economics Department Working Paper 460.
Keller, Wolfgang, and Stephen R. Yeaple. 2009. “Mul-
Oliner, Stephen D., Daniel E. Sichel, and Kevin J.
tinational Enterprises, International Trade, and Pro-
Stiroh. 2007. “Explaining a Productive Decade.”
ductivity Growth: Firm Level Evidence from the Brookings Papers on Economic Activity , 1: 81–137.
United States.” Review of Economics and Statistics
Olley,
, G. Steven, and Ariel Pakes. 1996. “The Dynamics
91(4): 821–31. of Productivity in the Telecommunications Equip-
Kellogg, Ryan. 2009. “Learning by Drilling: Inter-firm
ment Industry.” Econometrica , 64(6): 1263–97.
Learning and Relationship Persistence in the Texas
Pavcnik, Nina. 2002. “Trade Liberalization, Exit, and
Oilpatch.” National Bureau of Economic Research Productivity Improvement: Evidence from Chilean
Working Paper 15060. Plants.” Review of Economic Studies , 69(1): 245–76.
Klette, Tor Jakob, and Samuel Kortum. 2004. “Inno-Petrin, Amil, and James Levinsohn. 2005. “Measuring
vating Firms and Aggregate Innovation.” JournalAggregate
of Productivity Growth Using Plant-Level
Political Economy , 112(5): 986–1018. Data.” National Bureau of Economic Research
Knittel, Christopher R. 2002. “Alternative Regulatory
Working Paper 11887.
Methods and Firm Efficiency: Stochastic Frontier Petrin, Amil, and Jagadeesh Sivadasan. 2010. “Estimat-
Evidence from the US Electricity Industry.” Review ing Lost Output from Allocative Inefficiency, with
of Economics and Statistics , 84(3): 530–40. Application to Chile and Firing Costs.” Unpublished.
Lazear, Edward P. 2000. “Performance Pay and Produc-
Rose-Ackerman, Susan. 1991. “Risk Taking and Ruin:
tivity.” American Economic Review , 90(5): 1346–61.Bankruptcy and Investment Choice.” Journal of
Lentz, Rasmus, and Dale T. Mortensen. 2008. “An Legal Studies , 20(2): 277–310.
Empirical Model of Growth through Product Inno- Sakellaris, Plutarchos, and Daniel J. Wilson. 2004.
vation.” Econometrica , 76(6): 1317–73. “Quantifying Embodied Technological Change.”
Levinsohn, James, and Amil Petrin. 2003. “EstimatingReview of Economic Dynamics , 7(1): 1–26.
Production Functions Using Inputs to Control for Schmitz, James A. 2005. “What Determines Productiv-
Unobservables.” Review of Economic Studies , 70(2):ity? Lessons from the Dramatic Recovery of the US
317–41. and Canadian Iron Ore Industries following Their
Levitt, Steven D., John A. List, and Chad Syverson.Early 1980s Crisis.” Journal of Political Economy ,
2011. “How Does Learning By Doing Happen?” 113(3): 582–625.
Unpublished. Schoar, Antoinette. 2002. “Effects of Corporate Diver-
Maksimovic, Vojislav, and Gordon Phillips. 2001. “The
sification on Productivity.” Journal of Finance , 57(6):
Market for Corporate Assets: Who Engages in Merg- 2379–2403.
ers and Asset Sales and Are There Efficiency Gains?”
Schreyer, Paul. 2001. Measuring Productivity: Mea-
Journal of Finance , 56(6): 2019–65. surement of Aggregate and Industry-Level Produc-
Maksimovic, Vojislav, and Gordon Phillips. 2002. “Dotivity Growth: OECD Manuel . Paris and Washington,
Conglomerate Firms Allocate Resources Ineffi- DC: Organisation for Economic Co-operation and
ciently across Industries? Theory and Evidence.” Desarrollo.
Journal of Finance , 57(2): 721–67. Shearer, Bruce. 2004. “Piece Rates, Fixed Wages and
Malmendier, Ulrike, and Geoffrey Tate. 2009. “Super-Incentives: Evidence from a Field Experiment.”
star CEOs.” Quarterly Journal of Economics , 124(4):
Review of Economic Studies , 71(2): 513–34.
1593–1638. Syverson, Chad. 2004a. “Market Structure and Pro-
Marschak, Jacob, and William H. Andrews. 1944. “Ran-ductivity: A Concrete Example.” Journal of Political
dom Simultaneous Equations and the Theory of Pro- Economy , 112(6): 1181–1222.
duction.” Econometrica , 12(3–4): 143–205. Syverson, Chad. 2004b. “Product Substitutability and
Mas, Alexandre. 2008. “Labour Unrest and the Qual- Productivity Dispersion.” Review of Economics and
ity of Production: Evidence from the ConstructionStatistics , 86(2): 534–50.
Equipment Resale Market.” Review of Economic Syverson, Chad. 2007. “Prices, Spatial Competition and
Studies , 75(1): 229–58. Heterogeneous Producers: An Empirical Test.” Jour-
Melitz, Marc J. 2003. “The Impact of Trade on Intra-nal of Industrial Economics , 55(2): 197–222.
industry Reallocations and Aggregate Industry Pro-
Thornton, Rebecca Achee, and Peter Thompson. 2001.
ductivity.” Econometrica , 71(6): 1695–1725. “Learning from Experience and Learning from Oth-
Moretti, Enrico. 2004. “Workers' Education, Spill- ers: An Exploration of Learning and Spillovers in
overs, and Productivity: Evidence from Plant-Level Wartime Shipbuilding.” American Economic Review ,
Production Functions.” American Economic Review 91(5):
, 1350–68.
94(3): 656–90. UK Office of Fair Trading. 2007. “Production and
Muendler, Marc-Andreas. 2004. “Trade, Technology, Competition: An OFT Perspective on the Productiv-
and Production: A Study of Brazilian Manufacturers,ity Debate.” OFT 887.
1986–1998.” CESifo Working Paper 1148. van Ark, Bart, Mary O'Mahony, and Marcel P. Timmer.
Página 40
2008. “The Productivity Gap between Europe andof Rents: Wages and Innovation in a Panel of UK
the United States: Trends and Causes.” Journal ofCompanies.” Quarterly Journal of Economics ,
Economic Perspectives , 22(1): 25–44. 111(1): 195–226.
Van Biesebroeck, Johannes. 2003. “Productivity Veracierto, Marcelo. 2008. “Firing Costs and Business
Dynamics with Technology Choice: An Application Cycle Fluctuations.” International Economic Review ,
to Automobile Assembly.” Review of Economic Stud-49(1): 1–39.
ies , 70(1): 167–98. Verhoogen, Eric A. 2008. “Trade, Quality Upgrading,
Van Biesebroeck, Johannes. 2005. “Exporting Raisesand Wage Inequality in the Mexican Manufacturing
Productivity in Sub-Saharan African ManufacturingSector.” Quarterly Journal of Economics , 123(2):
Firms.” Journal of International Economics , 67(2):
489–530.
373–91. Vives, Xavier. 2007. “Innovation and Competitive
Van Biesebroeck, Johannes. 2008. “The SensitivityPressure.”
of IESE Business School Working Paper
Productivity Estimates: Revisiting Three Important
634.
Debates.” Journal of Business and Economic Statis-
Walker, Francis A. 1887. “The Source of Business
tics , 26(3): 311–28. Profits.” Quarterly Journal of Economics , 1(3):
Van Reenen, John. 1996. “The Creation and Capture 265–88.
Página 41
Copyright of Journal of Economic Literature is the property of American Economic Association and its content
no se pueden copiar o enviar por correo electrónico a múltiples sitios o publicar en una lista de correo sin el expreso del titular de los derechos de autor
permiso escrito. Sin embargo, los usuarios pueden imprimir, descargar o enviar artículos por correo electrónico para uso individual.

¿Que Determina La Productividad?

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

¿Que Determina La Productividad?

Transféré par

Droits d'auteur :

Formats disponibles

18/2/2018 ¿Qué determina la productividad?

Journal of Economic Literature 2011, 49: 2, 326-365

¿Qué determina la productividad?

1. Introducción persistentes diferencias de productividad medidas

es pequeño en relación con la variación de productividad

niveles (por ejemplo, Vojislav Maksimovic y Gordon

G ( A , K , L , M ). Totalmente diferenciando Sin

para el flujo de servicios de capital? Si el Incluso después de determinar cómo con-

las elasticidades α j pueden construirse directamente.Este

2.3 Un modelo de dentro de la industria tomar como dado. En imperfectamente competitivo

el nivel de costo de corte A. Una condición de entrada There

aggregate industry productivity—the ratio of may be favorable enough to induce some of

7 I mean long proposed: Francis A. Walker (1887) pos-

338 Journal of Economic Literature, Vol. XLIX (June 2011)

Newer work using matched employer- productivity dispersion as a metric of a newly

thick-input-market and knowledge-transfer market side. 13 They separately identify these

type of investigation. The industry's physi- example of productivity growth in an indus-

Syverson: What Determines Productivity? 361

of Chinese Imports on Innovation, IT and Productiv-

Vous aimerez peut-être aussi