Vous êtes sur la page 1sur 26

INCOME INEQUALITY AND THE MIDDLE CLASS IN TENNESSEE:

CENSUS 1990 & 2000 DATA ANALYSIS

MPA Paper Prepared By:

Mark Kleiner

Master of Public Administration


POLS 590
University of Tennessee at Chattanooga
24 April 2003
TABLE OF CONTENTS

Research Problem………..…………………………. 1

Literature Review………..…………………………… 3

Study Synopsis……………………………………….. 8

Methodology…………………………………………... 9

Findings………………………..…………………….. .11

Commentary…………………..…………………….. 13

Appendix A, figures #1 - #8……..………………...…16

Bibliography…………………………..……………... 22

1
RESEARCH PROBLEM:

The current field of study concerned with the dynamics surrounding the

middle class is one of the most relevant and fascinating research pursuits today.

According to an unofficial definition of the middle class suggested by the US

Department of Commerce, households that fall above and below the top and

bottom 20% of the population are considered part of the middle class (Census,

2003). Since the large majority of Americans find themselves in the “middle

60%,” on the scale, these studies naturally grow in significance. The ultimate

policy developed as a result of this research will most definitely impact the middle

class of America the hardest, since they comprise the greatest presence on the

quintile scale.

Current work on the middle class, much like other significant studies

surrounding important issues, is fraught with controversy and debate not

necessarily limited to the political arena. The controversy has affected this field

of research by creating misunderstandings centered on the relationship between

methodology and results. To begin, a standard definition has eluded

researchers. It is well known that definitions of variables are typically the

foundation of most studies. The lack of proper identification of a variable yields

findings that cannot be compared across studies, along with varying

2
methodologies that are not reliable. Secondly, any policy development stemming

from research will be compromised as to its total effectiveness on a target

demographic if it is based on methodology that has not been generally accepted

by its own academic community.

Unfortunately for the study of the middle class, its own sheer size

mandates a measuring tool that is uniformly accepted by the public and private

sectors. The government’s primary social responsibilities are contingent upon its

ability to tax the population of the US. The private sector must accept this policy.

However, private businesses exert a powerful influence on Washington, so the

expected policy development compromises are present because of the ability to

manipulate data through customized, self-serving methodologies propounded by

researchers.

This work examines the current state of research on the middle class and

the questions surrounding the topic. The major concerns are the size of the

middle class, and whether or not it is shrinking. A primary tool used by

researchers to study the size of the middle class, regardless of its definition, is

the Gini coefficient, which will be explained in detail. The Gini coefficient

measures income inequality in a population group. So by default, the study of

the middle class is overshadowed by the study of income inequality.

Additionally, this paper shows how measuring income inequality at the

state level (Tennessee) allows for confirmation of the current body of research

through application of local results to the conclusions presented by researchers.

3
This paper confirms that income inequality in Tennessee has increased between

1990 and 2000, while the income shares of the middle class have decreased.

LITERATURE REVIEW:

There is not enough research being conducted on the dynamics of the

middle class. The researchers that are paying attention to it find themselves in

the early, developmental stages of research. Evidence of this is found in the

extreme variation between studies (Dugger, 1998, 289).

However, one agreement shared among researchers is the demographic

profile of the middle class (Hacker, 1996, 41). The middle class family is

comprised of working parents, earning exactly the median estimated income for

their population. They both have a vehicle. The age of married couples is rising,

and the number of singles and broken families are on the rise, contributing to the

older couple phenomenon as well. They do not have children yet, waiting for

more financial security first. They are driven consumers, trying to maximize

every dollar. At the same time, retailers are trying to take every dollar. Times

are tight by any measure.

This being said, the motives behind studying the middle class are noble

enough: to brighten the prospects of a diminishing class (Howland, 1997, 300).

Three perspectives are circulating among researchers concerning what has

happened to the middle class. First, and most accepted, is that the middle class

is shrinking. There are myriad indicators identified by researchers of this

4
decrease in the middle class. They can be split into two different general

categories: Economic and Social.

Economic Rationale

Beginning with the economic indicators, wage stagnation and inflation are

cited as the number one reason for a decrease in the middle class (McMahon,

1997, 35). The effects of wage stagnation, according to McMahon are reflected

in the decay of our urban centers over the past three or four decades. He argues

that cities are synonymous with the middle class, and that they essentially

created one another. He argues that when wages flatten out, those in the middle

60% either drop or rise on the scale, and unemployment skyrockets. Judging

from the suburban exodus over the years, America’s cities are left with the low

end of the scale within their limits. On an even larger scale, the decline in the

middle class can be attributed to the decrease in available manufacturing jobs

(Kacapyr, 1996, 31). According to Kacapyr, the long, downward trend of the

proportion of jobs in manufacturing has coincided with the increase in Gini

coefficients.

Social Rationale

Changes in household composition over the last 40 years are the primary

social reason for the decline in the middle class. This includes a gradual decline

in the respect for traditional values that our parents held dear (McMahon, 1997,

33). These values included a respect for savings, as well as a commitment for

5
living within means and paying bills regularly. Along these same lines, some

researchers ascribe the decline to a loss of sincerity in our social policy planning

(Little, 2001, 341). Little argues that originally, there was a sense of genuine

care built into social programs that is lacking today. She cites Civil War benefit

programs, the GI Bill, and 1935 style social security. However, none of this deals

with the problem of the disappearing middle.

Solutions

Researchers are quick to point out the reasons for the decline, and even

quicker to explain how to fix the dilemma. The solutions of curbing the decline in

the middle class are centered on trying to alleviate income inequality in the US,

which alone is an interesting notion. A logical place to begin the fix is with the

government (Dugger, 1998, 286). Fixing the government’s perceived

inefficiency, as a solution, almost seems like an escapist approach to the topic. It

does lead to other interesting suggestions though, such as instituting tax codes

that will allow for half of the lower income bracket to move up into the middle

60% (Ehrle, 1996, 19).

Other proactive solutions are to deal with poverty directly (Feldstein, 1999,

137). Feldstein recommends that we should tackle poverty instead of the afore-

mentioned redistributive policy fixes. The difficulty in defining the middle class

equally applies to the attempt to define “poverty.” To approach a problem that

has no official definition to work on another issue that shares the same dilemma

compounds the difficulty.

6
The Kacapyr group offers other solutions to increase job wages. They

suggest unionizing, job training, and tax reform as the most reasonable ways to

curb income inequality (Kacapyr, 1996, 33). They also identify several shifts in

demographic trends that may affect the Gini coefficient in the near future. First,

labor force growth is slowing down. As less people are available for permanent

positions, as indicated by the temporary service industry becoming America’s

number one growth industry last year, the wages offered at these positions will

increase, fighting wage stagnation. This ultimately would cause the low end to

slide into the middle 60% because they are earning more. The labor force is also

aging, translating into higher pay and benefits to the older workers. Lastly, the

trend in securing education is much like the job training to capitalize on worker

skill sets. More education always translates into higher pay.

Alternative Explanations

Common sense dictates that when trying to implement a solution strategy,

properly defining the problem is the logical place to start. So the most

appropriate solution is to develop a standard measurement for the middle class.

As previously mentioned, few researchers even agree on what is happening with

the middle class. A more extreme perspective states that it has already vanished

(Ehrle, 1996, 20). (See figure #1) This illustrates how Ehrle measures the

middle class. She added capital gains and insurance supplements to the

income, subtracting government transfer payments (EIC) and Social Security

taxes. This was her attempt to remove what she describes as “government

7
tampering,” to expose what is really happening to the middle class strictly at the

household income level. The findings are startling, especially seeing that 37.9%

of the population makes less than $20,000 dollars per year. One can see that

based on the arbitrary definition of the middle class, or the lack thereof,

measurement is not conclusive. Based on the $45,000 cut off, Ehrle was able to

make the middle class “vanish.” This illustrates the need for standard measures

for the middle class.

An alternate perspective states the direct opposite of the Kacapyr group’s

notion of a vanished middle class. In the retail community, based on several

tenets of consumerism, the researchers are of the opinion that the middle class

has not disappeared at all, but remains alive and well (Levy, 1989, 139). Levy

argues that in terms of sales, the middle class has always been based on

traditional values. Levy goes on to identify a return of middle class values in

buying habits, buying traditional items more often than not. This is happening at

the same time that researchers are reporting that the middle class is vanishing.

How can this be? Also, consumer spending is on an increasing long-term trend,

but if the middle class is disappearing, who is doing all the buying? Levy

suggests that Americans are living well above their means. This explains why

the middle class is experiencing the largest personal debt load they have ever

experienced before, brought on by credit cards and loans.

Upon review of the problems and solutions for income inequality, and the

varying views on what is actually happening to the middle class, the following

study shows exactly what the Gini coefficient measures, and outlines a very

8
direct, traditional methodology that uses state level Census tables that determine

where income is distributed in Tennessee. In the Findings & Commentary

section, this study will identify the major fallacy in the field by clarifying the

problem.

STUDY SYNOPSIS:

Before discussing the methodology employed to arrive at my

conclusions, a little on the Gini measure is appropriate. Since the 40’s, the Gini

coefficient has been widely used as a variable in the majority of studies

concerned with income distribution (Cowell, 1995, 17), providing some degree of

uniformity to the study of income distribution. The Gini coefficient ranges from 0-

1. It is a measure of overall income inequality in a population. According to the

scale, perfect income equality is represented by a 0 score. This would mean, for

instance, that 10% of the population has received 10% of the income distribution,

20% received 20% income distribution, and so on. Graphically, this would be

illustrated by a 45° positive slope line with a range of 0-1. A score of 1 indicates

perfect income inequality, meaning that one household has all the share of

income (Lynch, 1998, 78). (See figure #2)

However, the actual graph of income distribution is traditionally a concave

curve (Lorenz curve) that hangs below “perfect equality.” The coefficient

measure itself is equal to the algebraic area under the line of equality and said

curve: the greater the area, the larger the income inequality, and in turn, the

larger the Gini coefficient (Atkinson, 1970, 244).

9
The attractive aspect of the Gini is that it shows in concise notation the

percentage of income share in a population group. As stated above, it shows

how much aggregate income is held by what percentage of the population. With

the Gini Interpolator program provided by the US Department of Commerce, all

one needs are the very basic variable inputs to complete the prompts on the

program. Without the processor, any researcher would be required to search old

tables for the coefficients, which would be very time consuming. Besides, the

tables are not accurate anymore. To derive a Gini through the base algorithm

step-by-step would require completing a 30 plus step equation, for each income

interval! Instead, all the necessary data was found at the Census website, and

defaults are worked into the system, to be precise as possible. These defaults

control for poverty, population, and race.

METHODOLOGY:

Hypothesis

This study hypothesizes that the Gini coefficients for 1990 and 2000 have

increased along with a decrease in income share of the middle 60% of the

population.

Data Sets

For the purpose of this study, the data inputs required were obtained from

Census Bureau table P080, which is “Household Income in 1989,” and QT-P32,

10
which is “Income Distribution of Households in 1999.” The tables were entered

into the interpolator, and processed. (See figures #3 and #4) The only

modifications to these tables were the removal of the family composition data

columns that were of no consequence to this study.

Results

The only omited information is the interval prompt for each interval, which

would be very lengthy. The only variables required for our purposes are the

calculated Gini, and income share data. According to analyst and program

developer Kirby Posey of the Department of Commerce Statistics Branch of the

Census Bureau, the standard error indicates that not all prompts were input

during the procedure. Prompts such as “input the aggregate income of each

income interval, if available,” automatically input a default list of values that are

derived from national trends. Similarly, the race and median income for each

income interval are set at defaults if the information is not available. For the

purpose of this study, the default more than compensates, first because they are

virtually the same as each other, and secondly, the aim of this study is to

produce a comparative analysis between two years, 1990 and 2000, and not

necessarily plot a trend line from year to year.

FINDINGS:

11
This study shows quite clearly that the Gini Coefficient for the state of

Tennessee increased from 1990 to 2000. This means more income inequality.

Before the Census 2000 information was available, this comparison could not be

made, not just because of the lack of availability of data sets. In 1993 data

collection methods were changed at the Census Bureau (Census Bureau, 2003)

This is very significant in that Ginis for the following year were skewed. The

availability of the 2000 data sets last summer were the first comprehensive sets

that could be used in a decade comparison since 1990, due to the aggregate

nature of the Gini coefficient which defines distribution in terms of aggregate data

such as “household income,” and due to similarities to the previous data

collection schemes before the switch.

The Gini in 1990 was 4.517, and in 2000 it has increased to 4.614. This

is an increase of .097, which on a scale of 0-1 is very significant, especially when

this change occurred in the span of a mere decade. To plot these Ginis on the

Lorenz Curve in figure #2 and #7, divide the Ginis by 10. As you can see if you

plot the Gini value for 1990 as a curve, .4517 concisely shows the income spread

of the percentage group of a population and its percentage share of the

aggregate income. (See figure #7)

Points of significance from figure #5 show that the lower quintile holds

only 3.3% of the aggregate income, and a salary cap of under $10,000 for 1990.

This situation was just as bleak in 2000, the lowest quintile held only 3.34% of

the aggregate, virtually no change over 10 years. The top salary in the lowest

fifth in 2000 was $15,550, not much of an improvement over ten years. That is

12
an average increase for the top wage of the low 20% of approximately $550 a

year. At the same time, the top 20% gained a whole percentage point, which

translates into an average gain of $10,317 per year through 2000. The top 20%

held just under half the wealth in the state, and in 2000 they were less than half a

percentage point away from clearing 50%. (See figure #8)

While the disparities between the rich and poor are interesting, what is

most important to this study is to identify the income share of the middle class.

As mentioned before, this middle is defined as the middle 60% of the quintile

scale. One sees from figure 5 and 8 that the middle class owns 48.27% of the

state aggregate income in 1990. Simply add the middle three quantiles together

to obtain this figure. In 2000, the middle class experienced a 1.35% drop in

income share to arrive at 46.92% of the aggregate. See figure #8.

With the given methodology which isolates the population as the

independent variable, decreases in Ginis were experienced by the middle and

upper-middle income groups, and since the lower 20% didn’t experience any

significant increase in income share, then it is clear where the money shifted.

The dollars shifted into the upper income bracket. Note that the upper limits

between the two years are $100,000 apart. The cap, as calculated by default in

the Gini coefficient processor multiplies the upper limit of the last closed interval

in the scale by approximately 2 to cap the open-ended interval (Welniak, 6).

13
COMMENTARY:

Based on a re-examination of the literature review, it becomes clear that

the study of the middle class rests on measurement methodology. There are two

avenues available to researchers who are faced with this type of subject: they

can all agree on a comprehensive definition of a variable of question, or refuse

to, as is the case of the body of work regarding the middle class, and wallow in

ambiguous results based on personalized definitions. Adding to the general

misunderstanding surrounding the field is the fact that most researchers do not

directly identify their independent variable, or their approach, while displaying

their results without this distinction.

The two most popular approaches at measuring the middle class are

based on the two variables used in this study, people and money. The two

methodologies define, from opposite directions, the theory behind the question of

size that permeates all the studies, this one included. The first approach asks

the question “has the middle class shrunk in terms of numbers of households?”

It becomes mandatory, then, to define the middle class by their income level

(independent variable), which would allow researchers to count households

(dependent variable). The obvious problem is that a standard definition of the

middle does not exist, allowing for every study to define it as they see fit. This

allows for manipulated data.

The second avenue focuses on the other variable, population. This is

much more tangible. The US government realizes that to study class dynamics,

starting with a tangible independent variable makes more sense than taking a

14
pluralistic approach by creating self-serving definitions for income brackets. The

illustration below shows the general concept behind the two major income

inequality strategies being used by academics today.

MONEY $$$ number of people change


Avenue 1

Avenue 2 PEOPLE ## amount of money changes

The second avenue concentrates on the population, rather than income

brackets. The government believes that it is more important, economically

speaking, to know who has the money, and where it is, rather than know how

many people are falling within a certain boundary on a scale. More importantly, it

is important to have a standardized, reliable methodology. The US government

has used the Gini coefficient since 1940. So the question for this approach

becomes “how much money does “X” percent of the population hold?” as

opposed to “how many people make “X” amount of dollars and ‘become’ middle

class?”

Fortunately for research conducted thus far on the topic of the middle

class, regardless of which methodology one employs, the results do ultimately

say the same thing, even though they tackle the issue from directly opposite

positions, by focusing on different independent variables. All results, some more

extreme than others, point to the fact that wealth is being shifted upwards and

downwards, and that the traditional middle, or what we may call “majority

stakeholders” in our economy are not “moving” but losing their monetary power.

15
This conclusion is based on the approach that isolates the population as the

independent variable, so that a percentage share of the aggregate income can

be identified. If the operative definition is “middle 60%” of a population, then that

number of households never changes, only their share of the wealth. On the

other hand, if the operational definition of the middle is an income bracket, then

one could track the annual change in number of households falling into the

interval.

This is the biggest misunderstanding that runs throughout the literature.

According to the government’s approach, the people in the middle are not going

anywhere, but their dollars are shifting. The confusion presents itself when

researchers define the middle with confusing and convenient terms, and don’t

specify what their independent variable is. In this literature review, researchers

have chosen to take the first approach mentioned above, by highlighting the

population as the transient variable, when according to the government’s

approach the only mobile variable is the dollar.

APPENDIX “A”

16
Figure #1

A Two-Tiered Stratification Model

(includes capital gains and employee health


insurance less
Social Security payroll taxes and government
transfers)

UPPER CLASS
Upper Upper $100,000+ 6.3%
Middle Upper $60,000-$99,999 14.7%
Lower Upper $45,000-$59,999 11.8%

Percent of households over $45,000: 32%

LOWER CLASS
Upper Lower $35,000-$44,999 10.1%
Middle Lower $20,000-$34,999 19.2%
Lower Lower $0-$19,999 37.9%

Percent of households under $45,000: 67.2%

Source: U.S. Bureau of the Census Population Reports

17
Figure #2

Gini Coefficient Plot

18
Figure #3 Figure #4

P080. HOUSEHOLD INCOME IN 1989 - QT-P32. Income Distribution in 1999 of Households and
Universe: Households Families: 2000
Data Set: 1990 Summary Tape File 3 (STF Data Set: Census 2000 Summary File 3 (SF 3) - Sample Data
3) - Sample data Geographic Area: Tennessee

Tennessee
Less than $5,000 163648
$5,000 to $9,999 207221
$10,000 to $12,499 104740 Total 2,234,229
$12,500 to $14,999 89526 Less than $10,000 267,405
$15,000 to $17,499 101753 $10,000 to $14,999 161,773
$17,500 to $19,999 88474 $15,000 to $19,999 160,371
$20,000 to $22,499 97894 $20,000 to $24,999 165,763
$22,500 to $24,999 79647 $25,000 to $29,999 162,795
$25,000 to $27,499 88525 $30,000 to $34,999 157,126
$27,500 to $29,999 70372 $35,000 to $39,999 140,047
$30,000 to $32,499 84031 $40,000 to $44,999 133,864
$32,500 to $34,999 61093 $45,000 to $49,999 114,184
$35,000 to $37,499 67147 $50,000 to $59,999 196,781
$37,500 to $39,999 52969 $60,000 to $74,999 208,583
$40,000 to $42,499 57308 $75,000 to $99,999 179,559
$42,500 to $44,999 42799 $100,000 to $124,999 80,699
$45,000 to $47,499 46215 $125,000 to $149,999 36,080
Figure #3
$150,000 to $199,999 31,095
$47,500 to $49,999 34541
$50,000 to $54,999 67313 $200,000 or more 38,104
$55,000 to $59,999 49414
$60,000 to $74,999 94201
$75,000 to $99,999 56341 Mean income (dollars) 48,688
$100,000 to $124,999 20626
$125,000 to $149,999 8143
$150,000 or more 19,574

Mean income 29338

19
Figure #5 Figure #6

Gini Coefficient, 1990 Gini Coefficient, 2000

GINI 1990 GINI 2000


GINI= 4.517996E-01 STAND ERROR= GINI= 4.614518E-01 STANDARD
2.862447E-02 ERROR= 2.692356E-02

QUINTILE UPPER LIMIT SHARE QUINTILE UPPER LIMIT SHARE


1 9995.00 3.31 1 15549.85 3.34
2 19604.65 9.19 2 29249.12 9.21
3 30559.32 15.53 3 44636.30 15.05
4 46332.13 23.55 4 67911.36 22.66
5 299998.00 48.42 5 399998.00 49.74
QUARTILE UPPER LIMIT SHARE QUARTILE UPPER LIMIT SHARE
1 12207.09 5.05 1 19032.75 5.11
2 24806.11 14.37 2 36494.27 14.17
3 41386.50 25.32 3 60819.53 24.38
4 299998.00 55.27 4 399998.00 56.33
DECIILE UPPER LIMIT SHARE DECIILE UPPER LIMIT SHARE
1 5522.68 .88 1 8354.39 .86
2 9995.00 2.43 2 15549.85 2.48
3 14745.44 3.84 3 22433.79 3.91
4 19604.65 5.35 4 29249.12 5.30
5 24806.11 6.91 5 36494.27 6.74
6 30559.32 8.62 6 44636.30 8.31
7 37231.60 10.55 7 54518.41 10.16
8 46332.13 13.00 8 67911.36 12.50
9 61487.34 16.59 9 92486.65 16.14
10 299998.00 31.83 10 399998.00 33.61

20
Figure #7

Tennessee Gini Comparison Plots for 1990 & 2000

Lorenz Curve: Income Inequality by quintile share


Tennessee Household Income, 1990 & 2000

1990
2000

21
Figure #8

Comparison Chart for Tennessee: 1990 & 2000

Item: 1990 2000


Gini Coefficient 4.517 4.614

Low 20% quintile share 3.31% 3.34%

High 20% quintile share 48.42% 49.74%

Middle 60% share 48.27% 46.92%

Upper limit, Low 20% $9995 $15,549

Upper limit, High 20% $299998 $399998

Highest 10% decile share 31.83% 33.61%

Lowest 10% decile share .88% .86%

22
BIBLIOGRAPHY

Atkinson, AB. “On the measurement of inequality.” Journal of Economic Theory


1970; 2:244-263.

Cowell, FA. Measuring Inequality. London, England: Prentice Hall/ Harvester


Wheatsheaf; 1995 p.17.

Dugger, W.M. “Against Inequality.” Journal of Economic Issues. 1998; 32:


286(18).

Ehrle, Lynn H. “The myth of the Middle Class.” The Humanist, 1996; 56:17(4)

Feldstein, Martin. “Reducing Poverty, Not Inequality.” Public Interest. 1999;


137: 13(8).

Hacker, Andrew. “Meet the Median family.” Time. 1996; 147: 41(3).

Howland, Marie. “Stemming Middle-Class Decline: The Challenges to Economic


Development Planning.” Journal of the American Planning Association.
1997; 63:300(11).

Jones, A.F.; Weinberg, Daniel. “The Changing Shape of the Nation’s Income
Distribution.” US Department of Commerce, Current Population Reports.
P60-204, 2003.

Kacapyr, E; Francese, P; Crispell, D. “Are you middle class?” American


Demographics. 1996; 18:30(6).

Kuttner, Robert. “Top-Down Class Warfare.” The American Prospect. 2000;


11:4(3).

Levy, Frank. “Income trends, the return of traditional values, and the vanishing
middle class.” Journal of Retailing. 1989; 65:137(7).

Little, Margaret. “The Middle Class and Social Policy.” Journal of the American.
Planning Association. 2001; 67: 341(7).

Lynch, John W; George Kaplan; Elsie R. Pamuk; Richard Cohen. “Income


inequality and mortality in metropolitan areas of the United States.”
American Journal of Public Health, 1998; 88:1074(6).

23
McMahon, Thomas L; Angelo, Larian. Hollow in the Middle: The Rise and
Fall of New York City’s Middle Class. CUNY Center for Urban Research/
CUNY Data Servic, 1997.

Patterson, George. Technical Paper 17: Trends in the income of families and
Persons in the US 1947-1964. Bureau of the Census; 1965.

Ryscavage, Paul. “A Surge in Growing Inequality?” Monthly Labor Review.


1995; August (51-61).

Skocpol, Theda. The Missing Middle: Working Families and the Future of
American Social Policy. Norton and Company, New York, 2000.

US Census Bureau. Current Population Survey.


www.census.gov/hhes/www/income00.html (see report “Money Income
in the United States).

US Census Bureau. Current Population Survey.


www.census.gov/hhes/www/income/midclass/midclsan.html (see report
“Income Inequality- Middle Class Narrative”) 2003

US Census Bureau. American Community Survey.


www.factfinder.census.gov/servlet/BasicFactsServlet

Welniak, Ed. Memorandum: “Calculating Indexes of Income Concentration from


Grouped Data: An Empirical Study.” 1988, US Dept. of Commerce,
Income Statistics Branch.

24
25

Vous aimerez peut-être aussi