International Journal of Manpower

Emerald Article: Resources in vocational education and post-schooling outcomes Mika Maliranta, Satu Nurmi, Hanna Virtanen

Article information:
To cite this document: Mika Maliranta, Satu Nurmi, Hanna Virtanen, (2010),"Resources in vocational education and post-schooling outcomes", International Journal of Manpower, Vol. 31 Iss: 5 pp. 520 - 544 Permanent link to this document: http://dx.doi.org/10.1108/01437721011066344 Downloaded on: 28-08-2012 References: This document contains references to 27 other documents To copy this document: permissions@emeraldinsight.com

Access to this document was granted through an Emerald subscription provided by UNIVERSITI PENDIDIKAN SULTAN IDRIS For Authors: If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service. Information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information. About Emerald www.emeraldinsight.com With over forty years' experience, Emerald Group Publishing is a leading independent publisher of global research with impact in business, society, public policy and education. In total, Emerald publishes over 275 journals and more than 130 book series, as well as an extensive range of online products and services. Emerald is both COUNTER 3 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0143-7720.htm

IJM 31,5

Resources in vocational education and post-schooling outcomes

Mika Maliranta
The Research Institute of the Finnish Economy, Helsinki, Finland

520

Satu Nurmi
Statistics Finland, Helsinki, Finland, and

Hanna Virtanen
The Research Institute of the Finnish Economy, Helsinki, Finland
Abstract
Purpose The purpose of this paper is to examine the determinants of labour market outcomes after the initial vocational basic education (ISCED 3). Design/methodology/approach A multinomial logit model is used for examining the effect of school resources and other factors on students post-schooling outcomes dened as employment, further studies, non-employment and dropping out. Analysis has been done by using unique linked register data on students, their parents, teachers, educational organisations and business companies in Finland. Findings The results indicate that teaching expenditures do not matter but teachers characteristics have a role to play. Teachers with a university degree increase the employment probability of the students, whereas the formal competence of the teachers does not have such positive effects. The students characteristics and performance in comprehensive schools play an important role in determining the outcomes. Local business conditions affect the outcomes of boys but less those of girls. The ofcial quality evaluations adopted in recent years seem to pay attention especially to such aspects of education production that are important for providing capabilities for further studies but less so for employability. Originality/value Employability seems to be a great challenge to the initial vocational basic education. The ndings for local business conditions give support to the view that measures of education policy do not sufce but need to be complemented with those of regional or employment policy, for example, policies aiming to increase regional mobility of the labour force. Such complementary tools are particularly important for boys. Keywords Education, Vocational training, Further education, Regional development Paper type Research paper

International Journal of Manpower Vol. 31 No. 5, 2010 pp. 520-544 q Emerald Group Publishing Limited 0143-7720 DOI 10.1108/01437721011066344

The authors are indebted to Rita Aspund, Virve Ollikainen and two anonymous referees for their insightful comments. They also thank conference participants at the 23rd Meeting of Finnish Economists in Jyvaskyla (2006) and The 29th Annual Meeting of the Finnish Society for Economic Research (2007) for useful suggestions and discussions. The usual disclaimer applies. This paper is a part of a project that studies the performance indicator system in Finnish vocational education and training, funded by the Ministry of Education (project 2/290/2004). All estimations have been carried out in Statistics Finland following the rules that are set to ensure protection of individual integrity and condentiality. The SAS and Stata codes used in this study are available from the authors on request.

1. Introduction Over the past few decades, a large number of empirical studies have been conducted to examine how the increasing supply of educated workers affects the economic growth of the nation or the returns on educational investments reaped by the individual and the whole society (e.g. Krueger and Lindahl, 2001). Also the literature studying rms demand for educated workers has expanded rapidly over the past few years (Hornstein et al., 2006). A third strand of the economics of education literature focuses on educational production with special emphasis placed on school-related determinants of student achievement. This line of research provides outstanding possibilities to explore more closely also the quality aspect of education, which arguably is a foil for the vantage points of the supply of and demand for education. The present paper aims to contribute to the literature on education production by using an extensive set of linked register data on the students in initial vocational education, the educational organisations (and their institutions) providing this type of secondary-level education, and the companies demanding these kinds of qualications. The dataset offers a unique opportunity to study empirically how various quantitative and qualitative aspects of school resources affect student performance. We use a multinomial logit model to examine the effect of school resources on students post-schooling outcomes dened as employment, further studies, non-employment and dropping out. We also evaluate the effects of a specic performance indicator recently adopted by the Finnish Ministry of Education as an additional decision-making tool for distributing statutory core funding to initial vocational education organisations. The Finnish system for initial vocational education is briey described in the Appendix. Hence, our study departs from the mainstream literature on educational production in that the students educational outcome is gauged on the basis of their employability (or further education ability etc.) rather than their test scores or earnings. As vocational education is primarily aimed at producing skilled labour (and pushing students into further studies), measures describing students post-school performance can be argued to be more relevant in this particular context as compared to studies focusing on students achievements in comprehensive schools or general upper secondary schools (Gymnasiums). Employability stands out as a more relevant measure also compared to earnings, not least in view of the wages of most European employees with a vocational education being determined through collective bargaining (Piekkola and Snellman, 2005). Accordingly our sample is not truncated by putting aside students who have not obtained a degree (i.e. a test score) due to failure in the exam or dropping out, or students with no earnings because of further studies or non-employment. Moreover, with our rich dataset we are able to carefully examine and control for a number of crucial aspects such as the effects of a broad set of student characteristics or local business conditions on their propensity of employment or further studies. Our main ndings are the following. Teaching expenditures do not seem to matter but teachers characteristics do. The students characteristics and performance in comprehensive school play an important role in determining the outcomes. Parental background has strong effects even after careful control for the other factors. Local business conditions affect the outcomes of boys but have a negligible impact on girls. The ofcial quality evaluations implemented by the Ministry of Education seem to pay attention especially to those aspects of vocational education production that are important for providing the students with capabilities for further education but less so

Resources in vocational education 521

IJM 31,5

522

for their employability. Finally, the performance indicator (tulosrahoitusmittari ) currently used in Finland as one of the decision-making tools for distributing funds to vocational education providers does not predict well the students propensity of the positive outcomes of employment or further studies. The rest of the paper is structured as follows. Section 2 provides a review of the theoretical underpinnings for our study. Section 3 describes the empirical framework and the data used. In Section 4 we present the results of our empirical analysis. Section 5 concludes. 2. Relevant previous literature The Coleman et al. (1966) report is commonly seen as the starting point for empirical research using educational production functions. This line of research attempts to identify major determinants of student performance, traditionally measured by test scores. A large number of factors have been examined over the years with the most important ones relating to student characteristics, including family background, as well as school resources in the form of per-pupil expenditure, class size and teacher characteristics. The empirical research on educational production has expanded rapidly since the Coleman report. The specication and estimation of educational production functions involve several econometric challenges, the solutions to which are not always straightforward. Indeed, this also seems to provide one important explanation for the differing results obtained in relation to inter alia school resources and especially class size (see Ammermuller and Lauer, 2008). Among the shortcomings that the reported estimates typically suffer from are omitted variables, student selection and, hence, endogeneity of school resources (Webbink, 2005). One way to overcome this endogeneity problem is to make use of experimental data or natural experiments (e.g. Angrist and Lavy, 1999; Krueger, 1999; Hoxby, 2000). Such data are seldom available, though. Recent studies have also exploited panel data to remove xed student, teacher and school effects (e.g. Hakkinen et al., 2003; Rivkin et al., 2005). Some studies have also beneted from improved longitudinal data which have enabled the inclusion of additional variables to control for student characteristics (e.g. Dearden et al., 2002; Meer, 2007). As already noted, the dependent variable in the estimated educational production functions usually measures the educational output, i.e. student outcomes, by means of test scores. While achievement tests are important not least for the selection of young people into further education, test scores may also be argued to represent too narrow a view on student performance (Hanushek, 1979; Rumberger and Palardy, 2005). This holds true especially when the analysis concerns post-comprehensive schooling[1]. Some studies have examined student achievements also by use of alternative measures (drop-out rate, continuing studies etc.). Lee and Barro (2001) used the drop-out rate as one measure for the schooling outcome in their specication of the education production function. This study utilised a panel dataset for a broad number of countries. They concluded that higher average school resources (such as a lower pupil-teacher ratio, higher average teacher salaries) and family inputs (such as richer and better educated parents) are associated with better performance of the pupils. An education outcome closely related to the drop-out rate is the students decision to remain in school. Krueger and Whitmore (2001) conducted a follow-up analysis of the

students who participated in the Tennessee STAR experiment[2]. In their analysis they concluded that attending a small class is related to an increased likelihood of taking the ACT or SAT college entrance exam. The effect was larger among minority students. Dustmann et al. (2003) examined the effect of class size on the students decision to remain in school beyond the age of 16. They found a sizeable and signicant effect. Furthermore, the effect of staying on had a signicant positive effect on their future wages. The earlier work using alternative measures of the schooling outcome has suggested that a higher level of school resources is associated with a higher level of positive education outcomes (such as further studies) and a lower level of negative outcomes (such as the drop-out rate). Furthermore, it has been concluded in several studies that the students characteristics (such as their parents education) seem to have an important role in explaining the students post-schooling outcomes. 3. Empirical setting and data With our longitudinal dataset we are able to trace a students way through upper secondary vocational education, and examine his or her later labour market status (see Figure 1). Our data allow us to explore the effect of educational expenditures and teachers characteristics, the effects of the students social background and individual characteristics, and the effect of the local business conditions on the students post-education destinations. 3.1 Linking data We use a rich dataset that is constructed by linking information on individuals, educational organisations (i.e. education providers) and their institutions (i.e. establishments) as well as local business conditions of industry and commerce. The data concerning students, teachers and local business conditions comes from Statistics Finland. Using data on applications for secondary education in the joint application system ( yhteishaku), we have chosen our initial sample based on persons born in 1982 (primary age group) or 1981 and who applied for secondary education in autumn 1997 or spring 1998. Using annual data on students, we trace those students registered in basic vocational education in 1999 and 2000. Information on the examined school resources are gathered from two sources. We have acquired information on the expenditures of the educational organisations in 1999 to 2001 from the Educational Expenditure Registers of the National Board of Education. The expenditure register covers information on all VET (vocational education and training) providers in Finland. The data on educational organisations have been linked to students based on their educational organisation and main eld of education (seven main elds) in each year. As explanatory variables, we use the person-level averages over the period 1999-2001. The use of the three-year averages can be justied on the basis of the validity (the vocational education typically takes about three years) as well as of the reliability (information for a single year may sometimes be inaccurate). Using the code of the educational institution, we are able to link institution-specic information on the average characteristics of the teachers calculated from the person-level data on teachers in educational institutions for the years 1999-2001. In

Resources in vocational education 523

IJM 31,5

524

Figure 1. Empirical setting and linking of data

order to take into account possible switches of the students between the places of study, the institution-specic information has been linked to the students on the basis of the students educational institution in each year and only then have we calculated for each student the averages over the period 1999-2001. The measures of schooling outcomes are generated by combining information from two sources. The individuals labour market status (employment, student etc.) is taken from the Finnish Longitudinal Employer-Employee Data FLEED in the year 2003. These data allow us to observe also the background characteristics of the individuals. Data on qualications provides us information on all the completed qualications until the year 2003. The person-level FLEED data can be linked to the Business Register by using the establishment code of the workplace. Using this feature of data we have calculated measures of the regional labour demand conditions over the period 2000-2002, which is the time period we expect the students to be entering the labour market.

3.2 Dependent variable The schooling outcomes are measured: . on the basis of the main activity during the year 2003; and . using the situation at the end of the year. In our base models we have used the former and in the robustness checks the latter way of dening the outcome. We have divided the education outcome into four exclusive categories: employed, pursuing further studies, non-employed and drop-outs. The students in the rst group employed may have completed their initial vocational education or not. The students in further education had completed their initial vocational education and were now continuing their studies. The students in the category non-employed had completed their initial vocational education but were now passive, i.e. these young people were neither furthering their education nor employed. The remaining students can be referred to as drop-outs. They have no job and no secondary educational degree[3]. The aim of vocational education and training is to improve the skills of the work force, but it also gives general eligibility for polytechnic and university studies. Hence, the rst two categories may be interpreted as positive outcomes, whereas the latter two categories are seen as negative educational outcomes. Table I shows that, on the basis of the main activity during 2003, some 62.2 per cent of the students were employed[4], 12.8 per cent pursued further studies, 15.5 per cent were non-employed ve years after starting their studies and the remaining 9.5 per cent were drop-outs. Substantial variation can be found between the different main elds of education. 3.3 School resources In the empirical literature one of the main indicators of resources invested into education is the educational expenditure. We have tested several measures for educational expenditures. In our base models we have used the teaching expenditure per student and in the robustness checks also other measures described below. Teaching expenditures contain teachers salaries, teaching materials, and other teaching-related costs. Work-life expenditures include all the costs of the on-the-job-learning period (the compensation paid to employers and the costs of supervising). Teacher training expenditures comprise expenses aimed at developing the skills of teachers and other staff members. Career teaching expenditures include student services which help students with career planning and job applications as well as recruiting services for rms. Teaching hours denotes all the hours spent on teaching or instructing students. In addition, we have included information on teacher characteristics measured by age, university degree and formal competence. Formal competence of teachers is dened by law and it can be acquired by having a sufcient amount of pedagogical education, for example. Teachers with a university degree do not necessarily have formal competence and vice versa. We also have information on whether the educational organisation has received an award for good conduct on the basis of performance indicators (rewarded for conduct) or has applied and received an award for high quality operations (rewarded for quality) (see the Appendix).

Resources in vocational education 525

IJM 31,5

526

Occupation 5,226 63.40 1,112 13.49 1,271 15.42 634 7.69 8,243 100.00 3,223 100.00 440 13.65 234 8.30 2,819 100.00 468 14.52 543 19.26 410 12.72 264 9.37 248 13.11 230 12.16 190 10.04 1,892 100.00 1,905 59.11 1,778 63.07 1,224 64.69

Employment

Studies

Non-employment

Drop-out

Total

Note: The full names of the main elds and their subelds are reported in Table III

Table I. Destinations of students in 2003 by the eld of education Technology Adm. and Commerce Hotel Social and Health Culture 296 50.34 122 20.75 97 16.50 73 12.41 588 100.00 Humanist 67 67.68 6 6.06 7 7.07 19 19.19 99 100.00 Total 10,924 62.23 2,247 12.80 2,717 15.48 1,665 9.49 17,553 100.00

Natur.

428 62.12

85 12.34

101 14.66

75 10.89

689 100.00

Furthermore, two variables describing the characteristics of the education organisations are included: the size of the education organization and the share of the special-case students (including disabled and immigrant students) in their educational organisation and the main eld of education in each year. Table II presents descriptive statistics of the key variables used in the estimation models below. The average teaching expenditure for the total sample is 3.80 thousand euros per student. The data nevertheless display considerable variation in this respect. The interquartile range is 1.07 ( 4:27 minus 3.20) thousand euros. Teaching expenditures account for most of the total expenditures. The average number of teaching hours is 75 (not reported in Table II). The average share of teachers with a university degree or formal competence is less than 30 per cent in secondary-level vocational education institutions. The average share of teachers aged below 35 is 11.2 per cent and the share of those aged 50 or more is 38.4 per cent. 3.4 Student characteristics Previous research has shown that a wide variety of individual student characteristics are related to student outcomes. We have included personal information on the date of
Percentile Mean Std p1 p25 p50 p75 p99 7.62 79.5 71.9 26.4 60.1 45.1 1.11 0.9 8.2 18.0 8.2 Proportion (%)

Resources in vocational education 527

Continuous variables Teaching exp./student (1,000e) 3.80 1.07 1.99 3.20 3.81 4.27 University degree teachers (%) 29.2 17.9 5.4 17.7 22.9 36.8 Formal qualied teachers (%) 27.4 13.2 3.8 19.8 25.5 32.4 234 year teachers (%) 11.2 5.5 1.9 7.9 10.7 14.2 50 2 year teachers (%) 38.4 8.7 13.4 32.4 38.9 44.3 Size of organization (00s) 7.5 7.9 0.5 2.9 5.5 8.5 Grade difference in elementary school 20.54 0.68 22.03 21.01 20.56 2 0.10 Peer effect of elementary school 0.8 0.1 0.5 0.8 0.8 0.9 Size of elementary school (00s) 3.5 1.6 0.0 2.4 3.4 4.5 Parents schooling years 12.2 2.1 9.0 12.0 12.0 14.0 Employment growth in region (%) 20.6 3.1 28.5 21.9 20.5 1.3 Separations of the above 50-years old per all workers in the business sector of the region (retirements in region) (%) 7.6 1.2 5.2 6.7 7.5 8.1 Binary variables Rewarded for quality (organization) Rewarded for conduct (organization) Switching education eld Male student Born in 1982 (ref. 1981) Selected to rst request Senior high school studies Parents earnings . e30,000 Finnish-speaking student Number of observations

10.6 9.4 21.2 3.2 59.6 90.5 72.7 6.0 14.3 94.6 17,553

Table II. Descriptive statistics on estimation sample, selected variables

IJM 31,5

528

birth, sex and native language[5]. In addition, we use information on the fathers or mothers education and annual income in the year 1995, depending on which one of the parents has higher education or income. As another important control, we have included a variable that measures how much the students grades differ from the average grades in his or her comprehensive school (grades vary from 4 to 10). The Table II shows the well-known fact that students who attain a vocational education have generally had below-average grades in comprehensive school. According to the table, the difference is 0.54 but the variation is considerable, the interquartile range being 0.91. As a measure of the environmental background factor we have calculated the proportion of the school-mates in primary school (i.e. in the comprehensive school) who have graduated from at least upper-secondary education before the end of 2003. We call this variable a peer effect set off in quotation marks since the identication of the causal link would require a more detailed analysis. The variable is included primarily for control purposes. The average proportion is 82 per cent and the interquartile range 9 per cent. We also have information on the main eld (koulutusala) and subeld (opintoala) of education (for detailed description of the elds of studies, see Table III). To obtain reliable estimates we control in detail for the subeld of education. There are 30 subelds of education initially, but for computational reasons we have dropped subelds with less than 100 student observations. As a consequence, eight subelds and 261 students are omitted from the analysis. We are ultimately left with an estimation sample of 17,553 students. 3.5 Local business conditions Finally, we have included variables describing employment growth and retirements[6] by region. These local business conditions may make a difference on employment possibilities etc. of the individuals. We dene the geographical area as the students municipality of residence, including all adjoining municipalities, in 2000. A typical area consists of ve to ten municipalities that lie within the travel-to-work distance from the place of residence[7]. As a robustness check, estimations were also done by measuring the local factors alternatively on the basis of the place of vocational education in 2000. The average employment growth rate in the neighbourhood of the students residence between 2000 and 2002 is 2 0.56 per cent. Again, however, the interquartile range is not negligible, being 3.21 per cent. 4. Econometric analysis 4.1 Multinomial logit method Multinomial models can be used when the dependent variable consists of several outcomes. We use the multinomial logit model (MNL), proposed by Luce (1959), which is the simplest alternative of multinomial models. It is a suitable method when the regressors do not vary over alternatives, as is the case in our analysis. This means that the value of an explanatory variable (e.g. the grades in the comprehensive school) of a certain student is the same for each of the four outcomes[8]. The multinomial probit model (MNP) is an alternative to MNL. It is, however, computationally much more demanding. When regressors do not vary over alternatives, MNP does not have any

Code 05 06 08 09 10 17 18 19 20 21 22 24 25 26 27 28 29 37 38 40 31 32 44 45 12 13 64 65 63 66

Field of studies and training Natural resources sector Agriculture Horticulture Fishery Other primary industries Forestry Technology and transport sector Graphics technology Heating and ventilation Machinery and metal technology Vehicles and transportation Textiles and clothing Food industry Electrical engineering Land survey technology Construction technology Wood industry Surface treatment Paper and chemical industry Seafaring Other technology and transportation Administration and commerce sector Business and administration Hotel, catering and home economics sector Hotel, restaurant and catering Home economics and cleaning services Social and health care services sector Social and health care services Beauty care Culture sector Crafts and design Communications and visual arts Music Theatre and dance Humanist and teaching sector Leisure activities Physical education

Koulutus- ja opintoala Suomeksi Luonnonvara-ala Maatilatalous Puutarhatalous Kalatalousa Muu luonnonvara-alaa Metsatalous Tekniikan ja liikenteen ala Graanen ala LVI-ala Kone- ja metalliala Auto- ja kuljetusala Tekstiili- ja vaatetusala Elintarvikeala Sahkoala Maanmittausalaa Rakennusala Puuala Pintakasittelyala Paperi- ja kemianteollisuudenala Merenkulkualaa Muu tekniikka ja liikennea Kaupan ja hallinnon ala Kaupan ja hallinnon ala Matkailu-, ravitsemis- ja talousala Hotelli-, ravintola- ja suurtalousala Koti-, laitostalous- ja puhdistuspalveluala Sosiaali- ja terveysala Sosiaali- ja terveysala Kauneudenhoitoala Kulttuuriala Kasi- ja taideteollisuusala Viestinta- ja kuvataideala Musiikkialaa Teatteri- ja tanssialaa Humanistinen ja opetusala Vapaa-ajan toiminta Liikunta-alaa

Resources in vocational education 529

Note: aExcluded from the analysis due to a small number of students (,100) in the group

Table III. Fields of study in vocational studies and training in Finland (the classication by the Finnish National Board of Education, the version of year 1995)

appreciable advantage over MNL. Both methods should produce nearly identical results (Long and Freese, 2006). Our experimentations indicated that this is true also here. We therefore report only the MNL results. We focus on marginal effects on the probabilities of a change in the regressor. The marginal effects are calculated at the means of the independent variables. The sum of these marginal effects over the four outcomes is zero by design.

IJM 31,5

530

4.2 Results Table IV presents the marginal effects from the MNL estimations and their standard errors. The results are divided into three parts. The rst part focuses on the effects of our primary interest, i.e. that of the school resources (the expenditures, the characteristics of teachers and education organisations). The results in the rst part show that teaching expenditures per student are not signicantly related to any of the different choices. We have also estimated models that, in addition, included work-life expenditures and teacher training expenditures per student but all three expenditure variables had insignicant relationships with the alternative outcomes (not reported here). Instead of expenditures we have also used teaching hours per student as an indicator of education input. Again, insignicant results were found (not reported here). Table IV shows that the share of teachers having a university degree has a signicant positive effect on the students employment probability. A bit surprisingly, formal competence for teaching has an independent negative effect on the employment probability. When the variable for the teachers formal competence is dropped, the results for the effects of the teachers university degree remain similar (not reported here). Broadly speaking similar ndings can be made when the university degree variable is dropped; the formal competence of the teachers has a negative effect on employment ( p , 0.01) (not reported here). A weak indication is obtained here that young teachers are able to lower the students non-employment probability. The size of the educational organisation has statistically insignicant negative effects on the outcomes considered here. In order to better capture the more qualitative aspects of education production at the initial vocational education level we have also added variables related to awards for quality and conduct. The reward for quality indicates whether or not the educational institution has received an award for high quality operations (laatupalkinto) at least once in the period 2001-2005. The estimated coefcient indicates a positive relationship with further studies but a negative relationship with employment. Hence, the results suggest that although the award criteria have changed over time these evaluations seem to pay continuous attention to such aspects of education production that are important for providing capabilities for further education but less so for employability. Our other main ndings remain largely intact after the inclusion of this quality variable (results are not reported here). Policy makers have become increasingly aware of the importance of providing incentives (in addition to resources) to the organisations for improving their education production. To this end one part of the funding is nowadays distributed on the basis of a performance indicator that is constructed in co-operation with Statistics Finland (for details, see the Appendix). It is a composite index that gives the greatest weights to employability and further studies (see Virtanen, 2006). The indicator is relatively simple and neglects a number of potentially important factors that are found important in this study, such as the eld of education at a detailed sublevel and the initial quality of the students skills, for example. Our results show that the award winning organisations (rewarded for conduct) do not fare any better in increasing the employment or the further studies probability of their students as compared to the non-winners. One of the most interesting ndings of our study is that teaching expenditures spent per student do not seem to have any effect on the students destination after their initial

Emp. b/se n 2 0.003 0.001 * * 2 0.001 * * 0.002 0.001 0.001 2 0.054 * * 0.002 2 0.331 * * 20.003 20.000 * 0.001 * * * * 0.001 0.000 0.000 0.007 0.001 0.057 20.073 * * * * 20.293 * * * * 20.001 0.092 * * * * 0.002 20.019 * * * 20.017 * * * * 0.025 * 0.001 20.015 * * * 0.013 0.004 0.038 0.002 0.020 0.009 0.009 0.006 0.009 0.001 0.006 0.013 0.001 0.003 2 0.004 * * * 2 0.006 * 0.005 0.044 0.002 0.013 0.009 0.011 0.006 0.017 0.001 0.009 0.012 0.001 0.003 20.001 * 20.000 0.040 * * * * 2 0.140 * * 0.011 * * * * 2 0.097 * * * * 0.047 * * * * 0.047 * * * * 0.035 * * * * 2 0.019 2 0.008 * * * * 0.043 * * * * 2 0.039 * * 0.007 0.056 0.003 0.025 0.013 0.012 0.008 0.024 0.002 0.011 0.018 0.002 0.005 20.000 0.000 0.005 * * * 0.007 17,553.0 0.065 217,574.2 0.063 * * * * 0.297 * * * * 20.004 * * * 0.062 * * * 20.030 * * * * 0.006 20.016 * * * 20.010 0.008 * * * * 20.009 20.019 2 0.031 * * * * 0.136 * * * 2 0.006 * * * 2 0.057 * * * * 2 0.019 * * 2 0.034 * * * 2 0.002 0.004 2 0.000 2 0.019 * * 0.045 * * * * 0.008 0.000 0.001 0.001 0.001 0.001 0.024 0.017 0.157 0.003 20.000 * 0.000 20.001 20.000 20.001 0.031 * * 0.002 0.029 0.004 0.000 0.000 0.001 0.000 0.000 0.012 0.007 0.078 0.003 2 0.000 0.001 * 2 0.002 * 2 0.000 2 0.000 0.016 2 0.005 0.244 * * * 0.006 0.000 0.000 0.001 0.001 0.001 0.014 0.011 0.089 0.003 0.000 0.000 0.001 0.000 0.000 0.007 0.007 0.063 0.004 0.024 0.001 0.019 0.005 0.007 0.005 0.012 0.001 0.005 0.008 0.001 0.002 Robust standard errors

Studies b/se Robust standard n errors

Non-emp. b/se Robust standard n errors

Drop-out b/se Robust standard n errors

School resources Teaching exp./student, e1,000 Univer. degr. teachers (%) Formal qualif. teachers (%) 234 year teachers (%) 50 2 year teachers (%) Size of organization (00s) Rewarded for quality Rewarded for conduct Special-case student share

Characteristics of student Grade dif. in element. sch. Peer effect of element. sch. Size of element. sch. (00s) Switching edu. eld Male student Born in 1982 (ref. 1981) Selected to rst request Senior high school studies Parents schooling years Parents earnings . e30,000 Finnish-speaking

Local business conditions Emp. growth in region Retirements in region n Pseudo R 2 Log likelihood

Notes: Coefcients refer to the average marginal effects; Robust standard errors, clustered on the basis of the educational organisation and the main eld of education (446 clusters); *p , 0.1; * *p , 0.05; * * *p , 0.01; * * * *p , 0.001

Resources in vocational education 531

Table IV. MNL estimations, base model

IJM 31,5

532

vocational education. The measurement problem of this variable is one possible explanation for this nding (see Ollikainen, 2007). Part of this problem arises from the fact that we use the average expenditures per student in the given organisation and main eld of education (452 groups) as our explanatory variable. This ratio may be inaccurate because the costs can vary between different students within the same group. In particular, there are some special-case student groups that may be more (or less) costly to teach than others and therefore our average expenditure variable may be inaccurate. To control for this possibility we included a proxy variable that measures the share of the special-case students (including disabled and immigrant students) in the students group (see below). The second two parts present the results for the student characteristics and local business conditions, respectively. Table IV shows that the students characteristics predict well their different outcomes. Unsurprisingly, high grades in comprehensive school (compared to the average grades in the students comprehensive school) increase the probability of nding a job or starting further studies, and, conversely, decrease the probability of becoming non-employed or dropping out (i.e. exiting the labour force). The peer effect of the comprehensive school level on the propensity for further education is positive and highly signicant while its relationship with employment is negative. The size of the comprehensive school is positively related to employment and negatively to further studies and non-employment. Students having switched their main eld of education in initial vocational education have a higher probability to pursue further studies and a lower probability to become employed or non-employed. Male students are 4.7 percentage points more likely to get a job than are female students, whereas female students more frequently continue their studies (the probability is 3 percentage points higher for female than for male students) upon graduation from initial vocational education. The drop-out propensity is similar for both genders. As expected, the younger students of the two cohorts studied (for instance, those who did not repeat any class during comprehensive schooling) have a clearly higher probability to enter employment and a lower probability to enter non-employment or to drop out compared to their older counterparts. A dummy variable that indicates when the student was selected to his or her rst request in the joint application procedure shows that these students have a higher employment probability and lower probability to pursue further studies or to drop out than the other students. To account for the possibility that some students may have complemented their vocational education with general education and thus may have paved their way for further studies, we have included a dummy variable indicating inclusion of senior high school studies. The coefcient is statistically insignicant for other outcomes except for the positive probability to drop out. Parental background has a major role to play in determining the students choice. The parents (the one who has the higher education) education years have a strong effect on the students choice between employment and further studies, decreasing the former and increasing the latter at the same rate. However, the independent effect of the parents income (measured by a dummy indicating whether the annual income of either of the parents exceeds 30,000 euros in 1995) on the students probability to pursue further studies is insignicant[9]. This nding suggests that it is the norms (or social networks) rather than the pecuniary resources that determine the students choice between employment and further studies. Furthermore, we nd that a student

having a high-income parent has a high employment probability and a low drop-out and non-employment probability[10]. Our results also indicate that the Finnish-speaking students have a low employment and a high non-employment probability. This nding is likely to reect primarily the fact that the Swedish-speaking people, which is the other major group in Finland, fare well in the labour market. High net job creation in the regions business sector increases employment and decreases non-employment probabilities. Signicant effects are found though one would expect migration to reduce the role of local effects. Local labour demand conditions are unrelated to the probability of further studies. So, we do not obtain any evidence that further studies have elements of the disguised unemployment (that is that further studies are a way to escape unemployment when labour markets are tight). We have also studied the potential impact of separations of older workers in the region as the ageing of the workforce may have profound effects on the local labour markets. We nd that a high separation rate of workers aged 50 or more, which can be thought as a proxy for retirements[11], in the neighbourhood lowers the students non-employment probability. We have also controlled for the subeld of the education (these 22 subelds are described in Table III). We do not show these results in the tables, but we note in this context that the effects of the different subelds vary quite considerably implying that detailed subelds are worth controlling for in this kind of analysis. For example, the Vehicle and Transportation eld (auto- ja kuljetusala) and the Machinery and Metal Technology eld (kone- ja metalliala) are characterised by relatively high employment probabilities whereas the Textile and Clothing eld (tekstiili- ja vaatetusala) and, in particular, the Communications and Visual Arts eld (viestinta- ja kuvataideala) point to weak employment probabilities. The difference in employment probability between the two extremes is about 30 percentage points even after careful control for other determinants (such as the students characteristics) of employability. In view of the current discussion about the differential success of boys and girls in the educational system, we have also estimated our models separately for girls (Table V) and boys (Table VI). The amount of resources used in the educational organisation seems to be ineffective for both sexes. We nd evidence that girls are more responsive to the teachers characteristics: the high share of the university educated teachers increases the employment probability and decreases the non-employment probability of girls but not of boys. Interestingly, the organisations receiving awards for providing high quality education increase the probability of further studies by girls but not by boys. However, those educational organisations that have been rewarded for high conduct do not seem to be able to produce different outcomes for either gender. Gender differences arise also with respect to local business conditions as these are an important determinant for boys but not for girls. High net job creation in the proximity of the students residence has a signicantly positive effect on the employment probability of boys whereas the employability of girls does not seem to be dependent on local business conditions. This might reect girls higher propensity to migrate, for example. On the basis of our estimates and descriptive statistics (given in Table II) we can evaluate the economic signicance of the different determinants. For boys, the difference between well-performing (measured at the third quartile of net job

Resources in vocational education 533

IJM 31,5

534

School resources Teaching exp./student, e1,000 Univer. degr. teachers (%) Formal qualif. teachers (%) 234 year teachers (%) 50 2 year teachers (%) Size of organization (00s) Rewarded for quality Rewarded for conduct Special-case student share 2 0.001 0.002 * * * * 2 0.001 * * 0.000 2 0.001 0.003 * * 2 0.048 * * 2 0.007 2 0.634 * * * * 0.058 * * * * 2 0.223 * * * 0.016 * * * * 2 0.045 * 0.043 * * 2 0.004 0.041 * 2 0.010 * * * * 0.071 * * * * 2 0.079 * * * 0.010 0.073 0.004 0.027 0.020 0.014 0.024 0.003 0.016 0.026 0.002 0.002 0.001 0.002 0.007 0.064 0.003 0.023 0.015 0.011 0.014 0.002 0.012 0.018 0.000 0.008 7,086.0 0.066 27,312.0 0.050 * * * * 0.289 * * * * 2 0.006 * * 0.045 * * 0.010 0.006 2 0.014 0.009 * * * * 2 0.019 0.014 2 0.034 * * * * 0.225 * * * * 2 0.007 * * 2 0.070 * * * * 2 0.039 * * 0.013 2 0.045 * * * 0.001 2 0.036 * * * 0.066 * * * * 2 0.003 * 2 0.011 * * 0.007 0.067 0.003 0.015 0.016 0.009 0.016 0.002 0.012 0.019 0.002 0.010 0.000 0.001 0.002 0.001 0.001 0.022 0.020 0.187 0.005 2 0.000 2 0.000 0.000 2 0.000 2 0.001 0.038 * * * 0.014 0.129 0.006 0.000 0.000 0.001 0.001 0.001 0.012 0.011 0.109 0.004 2 0.001 * 0.001 * * 2 0.001 0.000 2 0.001 2 0.009 2 0.017 0.365 * * * * 0.008 0.000 0.000 0.001 0.001 0.001 0.016 0.013 0.111

Characteristics of student Grade dif. in element. sch. Peer effect of element. sch. Size of element. sch. (00s) Switching edu. eld Born in 1982 (ref. 1981) Selected to rst request Senior high school studies Parents schooling years Parents earnings . e30,000 Finnish-speaking

Local business conditions Emp. growth in region Retirements in region n Pseudo R 2 Log likelihood

Notes: Coefcients refer to the average marginal effects; Robust standard errors, clustered on the basis of the educational organisation and the main eld of education (405 clusters); *p , 0.1; * *p , 0.05; * * *p , 0.01; * * * * p , 0.001

Table V. MNL estimations, for girls Emp. b/se n Robust standard errors Studies b/se Robust standard n errors Non-emp. b/se Robust standard n errors Drop-out b/se Robust standard n errors 2 0.008 * 2 0.001 * * * * 0.001 0.000 0.001 2 0.001 0.019 0.011 0.140 * 2 0.074 * * * * 2 0.291 * * * * 2 0.002 0.070 * * * * 2 0.014 2 0.015 * * 0.019 0.000 2 0.016 * * 2 0.001 0.000 0.002 0.005 0.000 0.000 0.001 0.000 0.000 0.014 0.008 0.082 0.005 0.035 0.002 0.020 0.010 0.007 0.015 0.002 0.008 0.014 0.001

Emp. b/se n 2 0.002 0.000 2 0.003 * * * * 0.002 0.001 2 0.001 2 0.045 0.004 2 0.262 0.037 * * * * 2 0.068 0.009 * * * 2 0.165 * * * * 0.054 * * * * 0.053 * * * * 2 0.076 * * 2 0.006 * * * 0.029 * * 2 0.009 0.009 0.072 0.003 0.042 0.014 0.010 0.033 0.002 0.013 0.023 0.002 0.006 20.001 20.001 0.001 0.003 2 0.004 * * * 2 0.003 0.005 0.034 0.002 0.035 0.010 0.006 0.010 0.001 0.007 0.016 0.007 * * * * 0.006 0.076 210,103.1 0.063 * * * * 0.276 * * * * 20.003 * 0.089 * * 0.002 20.026 * * * * 20.007 0.006 * * * * 20.005 20.037 * * 2 2 0.027 * * * * 0.081 2 0.005 * * 2 0.038 2 0.033 * * * 2 0.008 0.054 * * 2 0.001 2 0.009 0.025 0.007 0.056 0.002 0.025 0.012 0.008 0.024 0.002 0.011 0.016 0.001 0.004 20.072 * * * * 20.289 * * * * 20.001 0.114 * * * * 20.024 * * * 20.018 * * * 0.029 * 0.001 20.015 * * 0.021 * * 20.002 * * 20.002 0.010 0.001 0.001 0.002 0.001 0.001 0.032 0.019 0.214 0.001 20.000 0.001 * 20.001 * 20.001 20.000 0.018 20.003 0.009 0.004 0.000 0.000 0.001 0.000 0.000 0.015 0.008 0.099 0.000 2 0.000 0.001 * * 2 0.002 2 0.000 0.000 0.034 0.008 0.186 0.007 0.000 0.001 0.001 0.001 0.001 0.021 0.015 0.127 0.000 0.000 0.001 * * * 0.001 0.000 0.001 * * * 20.006 20.008 0.067 0.004 0.000 0.000 0.001 0.000 0.000 0.008 0.008 0.079 0.005 0.032 0.002 0.030 0.009 0.006 0.016 0.001 0.006 0.009 0.001 0.003 Robust standard errors

Studies b/se Robust standard n errors

Non-emp. b/se Robust standard n errors

Drop-out b/se Robust standard n errors

School resources Teaching exp./student, e1,000 Univer. degr. teachers (%) Formal qualif. teachers (%) 234 year teachers (%) 50 2 year teachers (%) Size of organization (00s) Rewarded for quality Rewarded for conduct Special-case student share

Characteristics of student Grade dif. in element. sch. Peer effect of element. sch. Size of element. sch. (00s) Switching edu. eld Born in 1982 (ref. 1981) Selected to rst request Senior high school studies Parents schooling years Parents earnings . e30,000 Finnish-speaking

Local business conditions Emp. growth in region Retirements in region n Pseudo R 2 Log likelihood

Notes: Coefcients refer to the average marginal effects; Robust standard errors, clustered on the basis of the educational organisation and the main eld of education (360 clusters); *p , 0.1; * *p , 0.05; * * *p , 0.01; * * * *p , 0.001

Resources in vocational education 535

Table VI. MNL estimations, for boys

IJM 31,5

creation variable) and badly performing regions (1st quartile) is 2.2 percentage points in the employment rate and -1.3 percentage points in the non-employment rate (calculated from the coefcients in Table VI). 4.3 Consideration of the omitted variable problem and robustness A cause for concern is whether the interpretation of statistical relationships found here may be misled by the omitted variable problem. This might be the case if there is selection of the students on the basis of such characteristics that are unobservable to us. However, our view is that this is not a major problem here. First, it is worth reminding that the set of conditioning variables is quite large and detailed including various important determinants such as parental background, for instance. Second, in the Finnish joint application system the selection for vocational studies is made primarily on the basis of the grades in the school-leaving certicate. In addition, a student gets some extra points for his or her rst request[12]. In some rare cases, education providers may use application tests, in which case they might be able to screen out students on the basis of such relevant characteristics that we cannot observe in our data. However, these tests are common and important in only four elds: Crafts and design, Communication and visual arts, Social and health care services and Beauty care. It should be noted that in our baseline model we have a control for the grade in the school-leaving certicate, a dummy variable that indicates whether the student has started those studies that were his or her rst wish, and, in addition, a set of dummy variables for the elds of studies. Yet, we have performed an additional robustness check with a subsample of students that excludes those sub-elds where application tests are used. These results do not challenge our main ndings (not reported here). On the other hand, there may be selection from the students side. Students with certain characteristics that are unobservable to us (like motivation) may want to apply to certain kinds of educational organisations or institutions and this selection mechanism might potentially generate spurious statistical relationships between the characteristics of organisations (or institutions) and the destinations of their students. We have made an effort to tackle this potential problem by excluding those students that have been selected to their rst request. The idea is that in this sub-group there should be a considerable amount of randomness in the sorting of the students into educational organisations and institutions. The sample size decreases from 17,553 to 4,785, which may weaken the accuracy of our estimates. The results of this experiment are reported in Table VII. Again, we nd no relationship between the amount of resources used and the outcome. The share of university-educated teachers has a positive effect on the employment probability but now the coefcient is not statistically signicant at the conventional levels of signicance ( p-value is 0.137). Now we nd somewhat stronger evidence that young teachers lower the non-employment probability of the students. It is well-known that MNL assumes that the ratio of two alternatives does not depend on other alternatives (i.e. that the assumption of independence of irrelevant alternatives, IIA, holds). For the IIA assumption it is important that the alternatives are reasonably distinct and dissimilar (McFadden, 1973; Amemiya, 1981). Our view is that this is the case here. It also turns out that statistical tests give support to our conjecture. Hausman IIA tests do not reject the IIA assumption. We have also

536

Emp. b/se n 20.006 0.001 20.000 0.000 0.001 20.000 20.053 20.020 20.296 0.053 * * * * 20.234 * * 0.009 * 20.108 * * * 20.015 0.087 * * * * 20.013 20.016 * * * * 0.046 * * 20.033 0.014 0.095 0.005 0.040 0.020 0.025 0.034 0.003 0.021 0.043 0.003 0.008 20.001 0.003 0.001 0.004 20.002 0.003 0.007 0.062 0.003 0.027 0.012 0.013 0.014 0.002 0.012 0.022 0.003 0.400 * * * * 20.002 20.040 * 0.025 * 20.052 * * * 20.006 0.005 * 20.032 * * 0.036 0.008 * * * 0.002 7,086.0 0.065 2 7,322.9 0.051 * * * * 0.265 * * * * 20.005 * 0.038 20.017 0.036 * * * 0.004 0.009 * * * * 20.006 20.015 0.010 0.083 0.004 0.023 0.015 0.019 0.028 0.002 0.014 0.029 0.002 0.005 2 0.107 * * * * 2 0.431 * * * * 2 0.002 0.109 * * * 0.007 2 0.071 * * * * 0.015 0.003 2 0.009 0.012 2 0.005 * * * 2 0.007 0.014 0.001 0.001 0.002 0.001 0.001 0.035 0.026 0.259 0.001 0.000 20.000 0.001 20.001 0.001 0.019 0.005 0.174 0.005 0.000 0.000 0.001 0.001 0.000 0.019 0.011 0.116 0.006 20.000 0.000 20.003 * * 20.001 20.001 0.013 0.016 0.146 0.010 0.000 0.001 0.001 0.001 0.001 0.025 0.016 0.170 2 0.000 2 0.000 0.001 0.002 0.000 0.000 0.021 2 0.001 2 0.023 0.008 0.000 0.000 0.001 0.001 0.001 0.024 0.018 0.166 0.010 0.055 0.003 0.035 0.013 0.019 0.022 0.002 0.012 0.023 0.002 0.005 Robust standard errors

Studies b/se Robust standard n errors

Non-emp. b/se Robust standard n errors

Drop-out b/se Robust standard n errors

School resources Teaching exp./student, e1,000 Univer. degr. teachers (%) Formal qualif. teachers (%) 234 year teachers (%) 50 2 year teachers (%) Size of organization (00s) Rewarded for quality Rewarded for conduct Special-case student share

Characteristics of student Grade dif. in element. sch. Peer effect of element. sch. Size of element. sch. (00s) Switching edu. eld Male student Born in 1982 (ref. 1981) Senior high school studies Parents schooling years Parents earnings . e30,000 Finnish-speaking

Local business conditions Emp. growth in region Retirements in region n Pseudo R 2 Log likelihood

Notes: Coefcients refer to the average marginal effects; Robust standard errors, clustered on the basis of the educational organisation and the main eld of education (405 clusters); *p , 0.1; * *p , 0.05; * * *p , 0.01; * * * *p , 0.001

537

Table VII. MNL estimations, for those who were not selected to their rst request

Resources in vocational education

IJM 31,5

538

performed Wald and LR tests to check whether there are any pairs of outcomes that can be combined in our analysis. Both of these tests unmistakably reject this for all six different cases. Likelihood-ratio tests for our explanatory variables give support to our decision to control for the subeld of education. In the models that are reported in the tables we have allowed for correlation between students who study in the same organisation and in the same main eld of education, i.e. we use standard errors that are clustered by these groups (446 clusters). As a robustness check, we have also calculated standard errors that are robust with respect to heteroskedasticity, but all observations are assumed to be independent (i.e. no clustering). In this case, the standard errors are generally smaller. 5. Conclusions This paper contributes to the literature on the effectiveness of initial vocational education production by use of an extensive set of linked register data. We examine the determinants of alternative labour market outcomes after initial vocational education (ISCED 3). Our categorical dependent variable distinguishes between four separate outcomes: (1) employment; (2) further studies; (3) non-employment; and (4) dropping out. The explanatory variables are classied into three main groups: (1) the school resources, the educational organisations (and their institutions); (2) the students characteristics; and (3) the local business conditions. Our main ndings are the following. Teaching expenditures do not seem to matter but teachers characteristics have a role to play. Teachers with a university degree increase the employment probability of the students whereas the formal competence of the teachers does not have such positive effects. The students characteristics and performance in comprehensive school play an important role in determining the outcomes. Local business conditions affect the outcomes of boys but less those of girls. The ofcial quality evaluations adopted in recent years seem to pay attention especially to such aspects of education production that are important for providing capabilities for further studies but less so for employability. An import policy issue in Finland, as in many other countries, is how to improve the employability of the young. When it comes to initial vocational education in Finland, some steps have already been taken in this direction. The idea is to identify and reward those education providers that are able to increase the students employment probability or encourage them to pursue further studies. Against this background our ndings are somewhat worrying. The traditional policy tool of increasing expenditures does not seem to help. The qualitative approaches do not seem encouraging either. Those organisations that, according to ofcial evaluations (laatupalkinto), have high quality educational production are able to increase the further studies probability but

at the cost of a lower employment probability while the non-employment probability is unaffected. In particular, we have found evidence that the qualitative aspects, as the current quality reward systems identify them, seem to have some signicance for girls. For boys, however, these institutional characteristics seem to be ineffective. Our ndings seem to indicate that the determinants of high employability may be deeper and more case-specic than those affecting the capability of pursuing further studies. Our results also highlight the important role of student background characteristics, including performance in comprehensive schooling, as well as parental background, in creating the prerequisite skills for further success. Employability thus poses a great challenge to education policy. The point is how to provide education providers with incentives for nding and implementing those tools that are most effective in varying situations. The eld of education strongly affects outcomes. Our results show that a difference of 20 percentage points in employment probability rates between two subelds of education is not unusual at all. These ndings indicate that the balance of labour demand and supply, and thus the level of employment, can be substantially improved by reallocating resources (and students) between elds of study according to the contemporary needs. Our ndings for local business conditions give support to the view that measures of education policy do not sufce but need to be complemented with those of regional or employment policy, for example, policies aiming to increase regional mobility of the labour force. Our results suggest that such complementary tools are particularly important for boys. In this study we have only scratched the surface of the complicated mechanisms of educational demand and supply. Further analysis using large-scale register data would give important new knowledge for policymakers in developing the educational system and its incentives. For example, the links between business life and educational organisations and the effects of these connections on employability could be studied more extensively.

Resources in vocational education 539

Notes 1. Vocational qualication has been studied in a few studies, e.g. Dearden et al. (2002) and Meer (2007). 2. STAR-experiment is probably the best known, large-scale education experiment. In this experiment, students and their teachers were randomly assigned to small, regular-size, or regular-size classes with a teacher aide in the rst four years of school. 3. These students may be enrolled in education but they do not have a secondary education degree. 4. Most of these students (84%) had completed their initial vocational education while the rest (i.e. 16%) had a job without graduating. 5. Finland has two ofcial languages: Finnish and Swedish. Finnish is spoken by most of the population. Only about 5.5 percent of the population have Swedish as their mother tongue. 6. As a proxy measure for the retirement intensity in the region we have used the share of the separated older workers (above 50 years old) of total employment in the regions business sector rms. For the denition of region (see footnote 7).

IJM 31,5

540

7. More precisely, the numbers for each area are calculated by taking a weighted average of all municipalities in the area around the central municipality. Each municipality is the central municipality of its labour force area. So, calculations are performed separately for each municipality so that the number of areas is equal to the number of municipalities in Finland. The weight of an adjoining municipality is the share of employees of the total number employees in the central municipality who have their residence in that adjoining municipality. The weight of the central municipality is the share of those employees in the central municipality who do not commute between municipalities. About this approach of dening regions and calculating indicators for these regions, see Maliranta and Nurmi (2004) 8. See Cameron and Trivedi (2005) for a comprehensive description of multinomial models. 9. We have used a dummy variable in the models due to our concerns about the quality of information in the left tail (and extreme right tail) of distribution. However, we have also done a robustness check by using the log of parental earnings in place of the dummy variable. The results were similar to what was found in our baseline model. 10. As expected, parental earnings and education are correlated (r 0:2684). Still, the results were quite similar in the two cases when one or the other of these two parental variables were dropped from the model. 11. Ilmakunnas and Maliranta (2007) uses the same variable in their analyses and they nd that in two cases out of three the destination of these separations was retirement or so-called pension unemployment. 12. In some cases applicants may receive some additional points on the basis of work experience. The importance of this factor, however, is quite minor in general, and in our sample in particular because we focus on the students that are quite young.

References Amemiya, T. (1981), Qualitative response models: a survey, Journal of Economic Literature, Vol. 19, pp. 1483-536. Ammermuller, A. and Lauer, C. (2008), School quality and educational outcomes in Europe, in Dolton, P., Asplund, R. and Barth, E. (Eds), Education and Inequality across Europe, Edward Elgar, Cheltenham, and Northhampton, MA, pp. 47-66. Angrist, J.D. and Lavy, V. (1999), Using Maimonides rule to estimate the effects of class size on scholastic achievement, Quarterly Journal of Economics, Vol. 114 No. 2, pp. 533-75. Cameron, A.C. and Trivedi, P.K. (2005), Microeconometrics. Methods and Applications, Cambridge University Press. Coleman, J.S., Campbell, E.Q., Hobson, C.J., McPartlant, J., Mood, A.M. and Weinfeld, F.D. (1966), Equality of Educational Opportunity, US Government Printing Ofce, Washington, DC. Dearden, L., McIntosh, S., Myck, M. and Vignoles, A. (2002), The returns to academic and vocational qualications in Britain, Bulletin of Economic Research, Vol. 54 No. 3, pp. 249-74. Dustmann, C., Rajah, N. and van Soest, A. (2003), Class size, education and wages, The Economic Journal, Vol. 113 No. 485, pp. F99-F120. Hakkinen, I., Kirjavainen, T. and Uusitalo, R. (2003), School resources and student achievement revisited: new evidence from panel data, Economics of Education Review, Vol. 22 No. 3, pp. 329-35. Hanushek, E.A. (1979), Conceptual and empirical issues in the estimation of educational production functions, Journal of Human Resources, Vol. 14 No. 3, pp. 351-88.

Hornstein, A., Krusell, P. and Violante, G.L. (2006), The effects of technical change on labour market inequalities, in Aghion, P. and Durlauf, S. (Eds), Handbook of Economic Growth. Hoxby, C.M. (2000), The effects of class size on student achievement: new evidence from population variation, Quarterly Journal of Economics, Vol. 115 No. 4, pp. 1239-85. Ilmakunnas, P. and Maliranta, M. (2007), Aging, labor turnover and rm performance, Discussion Papers No. 1092, ETLA, Helsinki. Krueger, A.B. (1999), Experimental estimates of education production functions, Quarterly Journal of Economics, Vol. 114 No. 2, pp. 497-532. Krueger, A.B. and Lindahl, M. (2001), Education for growth: why and for whom?, Journal of Economic Literature, Vol. 39 No. 4, pp. 1101-36. Krueger, A.B. and Whitmore, D.M. (2001), The effect of attending a small class in the early grades on college-test taking and middle school test results: evidence from Project Star, Economic Journal, Vol. 111 No. 468, pp. 1-28. Lee, J.-W. and Barro, R.J. (2001), Schooling quality in a cross-section of countries, Economica, Vol. 68 No. 272, pp. 465-88. Long, J.S. and Freese, J. (2006), Regression Models for Categorial Dependent Variables Using Stata, 2nd ed., Stata Press, College Station, TX. Luce, R.D. (1959), Individual Choice Behavior: A Theoretical Analysis, John Wiley, New York, NY. McFadden, D. (1973), Conditional logit analysis of qualitative choice behavior, in Zarembka, P. (Ed.), Frontier of Econometrics, Academic Press, New York, NY, pp. 105-42. Maliranta, M. and Nurmi, S. (2004), Do foreign players change the nature of the game among local entrepreneurs?, Discussion Papers No. 942, The Research Institute of the Finnish Economy. Meer, J. (2007), Evidence on the returns to secondary vocational education, Economics of Education Review, Vol. 26 No. 5, pp. 559-73. Ollikainen, V. (2007), Ammatillisen Peruskoulutuksen Kustannustehokkuus 2001-2003, Tutkimuksia, No. 132, VATT. Piekkola, H. and Snellman, K. (Eds) (2005), Collective Bargaining and Wage Formation. Performance and Challenges, Springer. Rivkin, S.G., Hanushek, E.A. and Kain, J.F. (2005), Teachers, schools, and academic achievement, Econometrica, Vol. 73 No. 2, pp. 417-58. Rumberger, R.W. and Palardy, G.J. (2005), Test scores, dropout rates, and transfer rates as alternative indicators of high school performance, American Educational Research Journal, Vol. 42 No. 1, pp. 3-42. Virtanen, H. (2006), Tulosrahoitusmittariston Arviointi - Palkkiomuotoinen Tulosrahoitus Ammatillisessa Peruskoulutuksessa (Evaluation of Performance Indicators Rewarding Results in Initial Vocational Training, Abstract in English). Etla B 222, Taloustieto Oy, Helsinki. Webbink, D. (2005), Causal effects in education, Journal of Economic Surveys, Vol. 19 No. 4, pp. 535-60. Appendix. Initial vocational education and training in Finland Post-compulsory level in Finland is divided into general and vocational education (Figure A1)). In 2004 there were 64,100 students in the top form (year 9) in the comprehensive school. Of these students 54 per cent went on to general upper secondary education, 38 per cent to vocational education and training (VET), 3 per cent to the additional 10th form, and 5 per cent did not

Resources in vocational education 541

IJM 31,5

542

Figure A1. Education system in Finland

immediately continue studying. Students can enter the VET not only from comprehensive school but also in later stages in their education. All in all about 147,000 students are enrolled in vocational education every year. A vocational education takes on average three years to complete. It aims to provide necessary vocational competence, knowledge, and skills for working life and to encourage life-long learning. It also gives general eligibility for polytechnic and university studies. Vocational education and training includes theoretical instruction given by vocational institutions and a supervised on-the-job-learning period (at least six months). It can also take the form of

Resources in vocational education 543

Figure A2. Financing of vocational education and training in Finland

apprenticeships in which case 70 to 80 per cent of the training takes place at the workplace. Initial vocational education and training is available in seven elds: Technology and Transport, Business and Administration, Health and Social Services, Culture, Natural Sciences, Leisure and Physical Education, and Tourism, Catering and Home Economics (the current classication). Applications for vocational education and training are made through a national joint application system. Students can apply simultaneously to ve different degree programmes in all vocational institutions involved in the system and to indicate their preferences of the ranking of these degrees. The education provider makes the decisions on admissions. Student selection is typically based on grades in the school-leaving certicate. Other selection criteria are various entrance or aptitude tests, work experience and success in previous studies. Vocational education and training providers are responsible for organising training in their areas, matching provision with local market needs, and devising curricula based on the core curricula and requirements set by the Ministry of Education. There are 210 vocational education providers in Finland. They may be a local authority, a municipal training consortium, a foundation or other registered association, or a state company. Responsibility for funding the vocational education and training is divided between the state and municipal governments. The present system was adopted in 1997 and reformed in 2002 (Figure A2). The statutory nancing is based on unit costs (average cost), transactions (student numbers) and costs (eld-specic; special tasks). It does not depend on the actual expenditure. In addition, two per cent of the funding is based on performance indicators which evaluate the performance of the education providers (award for conduct). First, the indicators estimate students placement in employment and further education after their graduation (impact). They also include drop-out and graduation rate measures (process). Furthermore, formal competence of teachers and staff development is evaluated (staff). Vocational education providers may also receive quality awards based on separately selected themes. It should be noted that only a proportion of education providers have decided to apply for the reward.

About the authors Mika Maliranta is Head of Unit at ETLA. In 2003, he obtained a PhD in economics at the Helsinki School of Economics and Business Administration. Since obtaining his MSc (Econ.) in 1993 he has been researcher at ETLA and during certain periods also at Statistics Finland. He has written a number of articles and monographs on several productivity-related issues including, notably, the micro-level dynamics of productivity growth and the use of ICT. Recent research interests include the role of skills in technological development, and the production of education at the primary and secondary levels, including early school leavers and drop-outs. Mika Maliranta is the corresponding author and can be contacted at: mika.maliranta@etla.

IJM 31,5

544

Satu Nurmi is Head of Research at Statistics Finland. In 2004, she obtained a PhD in economics at the Helsinki School of Economics and Business Administration. Since obtaining her MSc (Econ.) in 1998 she has also been working as a research assistant at Helsinki School of Economics and a researcher at Statistics Finland. She has studied the growth and survival of Finnish manufacturing plants. Recent research interests include the dynamics of exporting, the effects of foreign ownership, entrepreneurship and the production of education. Hanna Virtanen is Teaching Assistant and a third year PhD student at Helsinki School of Economics. In 2005, she obtained her MSc (Econ.). She has been a researcher at ETLA for more than three years during the period of 2002-2008. She has conducted studies about the Finnish music industry and the performance-based funding in vocational education and training in Finland. Recent research interests include the career and education choices of young people and their success in the labor market and society, as well as the production of education at the primary and secondary levels.

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints