The eﬀect of polytechnic reform on migration

This paper examines the effect of polytechnic reform on geographical mobility. A polytechnic, higher education reform took place in Finland in the 1990s. It gradually transformed former vocational colleges into polytechnics and also brought higher education to regions that did not have a university before. This expansion of higher education provides exogenous variation in the regional supply of higher education. We find that the reform increased the migration of high school graduates. The migration propensities increased particularly close to graduation from high school, but some results also suggest a smaller positive effect over a longer period.


INTRODUCTION
A polytechnic education reform took place in Finland in the 1990s. It gradually transformed former vocational colleges into polytechnics offering a Bachelor's degree and expanded the supply of higher education to all regions. That is, the reform brought higher education also to regions that did not have a university in the pre-reform system.

2
The polytechnic reform was the largest single education reform in Finland since the reform of the comprehensive school system in the early 1970s. The polytechnics constituted a new non-university sector in higher education. The number of graduates from polytechnics has risen very rapidly. By 2000 the number of new polytechnic graduates exceeded the number of new university graduates. This paper exploits the polytechnic reform to examine internal migration patterns. We argue that the lessons from the reform have a broader interest, because they help to understand how the expansion of education, experienced by most advanced countries, affects internal mobility patterns.
The fear among the policy-makers has been that the polytechnic reform may have resulted in increased out-migration of the highly educated graduates from the peripheral regions ('brain drain'), for example, because job opportunities for the highly educated are less local. This outcome would be undesirable from the perspective of regional policy objectives, since the highly educated migrants possess above average skills and also earn above average incomes. Therefore, the prospects of economic growth in the peripheral regions weaken and the tax burden of those who remain rises. Consequently, regional disparities may increase substantially. Obviously, for individuals it is desirable to improve human capital and leave a declining area. 1 Although prior analyses of the relationship between education and migration behaviour are extensive, only the recent studies by Hickman (2009), Machin, Pelkonen and Salvanes (2011), and Malamud and Wozniak (2010) have utilized policy reforms to study the relationship between education and migration. Hickman (2009) considers the extent to which a merit-based aid programme in Florida has affected the location decisions of the college-educated. His results show that those eligible for the programme are significantly more likely to locate in Florida after completing their education than those who were not eligible.
In Machin et al. (2011) and Malamud and Wozniak (2010) the emphasis is on the effect of education on internal migration. Using a Norwegian primary school reform, Machin et al. (2011) find that the length of compulsory education has a positive causal impact on migration. One additional year of education increases the annual migration rates by 15 percent from a low base rate of one percent per year. Malamud and Wozniak (2010) 3 use variation in college attainment in the U.S. induced by draft-avoidance behaviour during the Vietnam War. Their results imply that the additional years of higher education significantly increased the likelihood that the affected men, later in life, resided outside the states where they had been born.
In this paper, the aim is to explore the effects of the polytechnic reform on interregional migration. The analyses are based on a particularly rich longitudinal data on graduated high school students from 1988 to 1998. The polytechnic reform provides us with the exogenous variation in the supply of higher education across regions and over time. The results point out that the expansion of higher education increased the migration of high school graduates. The migration propensities increased particularly close to graduation from high school. However, some estimates also suggest a smaller positive effect over a longer period.
The remainder of the paper is organized as follows. The next section describes the higher education system in Finland and the polytechnic reform. In Section 3 we discuss reasons why the polytechnic reform should affect interregional migration. Section 4 introduces the data. Section 5 describes the empirical approach, and the results are reported in Section 6. The last section concludes.

HIGHER EDUCATION IN FINLAND AND THE POLYTECHNIC REFORM
Compulsory comprehensive schooling for Finnish children begins at the age of seven and it lasts for nine years. 2 Roughly 50 percent of the pupils continue to a high school (general upper secondary school, "lukio" in Finnish), which lasts for three years and ends with a matriculation examination. At the beginning of the 1990s, vocational schools and colleges were a diverse group of schools. Some took most of their students directly from comprehensive schools and provided them with two or three years of vocational education. In some vocational colleges most students had completed high school before entering vocational college. For example, a business degree from a vocational college typically required three years of schooling after a comprehensive school or two years of schooling after a high school.
Since the polytechnic education reform the higher education system in Finland has comprised two parallel sectors: universities 3 and polytechnics. In essence, the reform brought higher education to areas that did not have a university before the reform. The polytechnic degrees are bachelor-level degrees with a vocational emphasis. These degrees take from three and a half to four years to complete. A major difference between the sectors is that polytechnic schools are not engaged in academic research like universities. Education is free at both levels.
The first 22 polytechnics were established under a temporary licence in 1991 (e.g., Lampinen, 2004). The polytechnics were created by gradually merging 215 vocational colleges and vocational schools. The gradual implementation of the reform is clearly reflected in the fact that students who had started their studies before a particular vocational college transformed itself into a polytechnic continued their studies along the old college lines and they eventually graduated with vocational college degrees. 4 Hence, the timing of the reform varied considerably across schools and regions, as described in Böckerman et al. (2009, p. 674-675); see also Figure 2 below. Seven new temporary licences were granted during the 1990s. The first graduates from the new polytechnics entered the labour market in 1994. The experimental phase was judged to be successful and since 1996 the temporary polytechnics have gradually become permanent.
Currently there are 27 multidisciplinary polytechnics in Finland. Unlike the university sector, the network of polytechnics covers the whole country.
The supply of education is controlled by the Ministry of Education through its decisions on the number of study places and the funding of schools. Until the end of the 1990s the number of polytechnic study places increased very rapidly (Figure 1). By 1996 the number of new polytechnic students exceeded the number of new university students.
The number of applications to universities and to the most popular polytechnics exceeds the number of available places by a factor of four.
--- Figure 1 around here ---3 The Finnish university sector consists of 20 universities and art academies, all of which carry out research and provide education-awarding degrees up to doctorates. For further details on the university sector, see e.g. Ministry of Education (2005). 4 The reform changed the curriculum to a different extent in different fields (Böckerman et al., 2009, p. 675). The changes were relatively minor in engineering and nursing education but substantial in business education. The average length of the studies in business education increased from two years to three and a half years. Figure 2 reveals the significant regional differences in the availability of polytechnic (higher) education and the changes in it during the 1990s. Since the availability has been relatively constant thereafter, the following analysis of the reform focuses on the 1990s and early 2000s.
--- Figure 2 around here --- The main aim of the reform was to respond to new demands for vocational skills that were seen to arise in the local labour markets. Furthermore, the geographically broad network of higher education was regarded as a means to equalize regional development, for example, by reducing the brain drain from the less developed regions to the metropolitan areas and therefore to lessen the concentration of the workforce to the central regions. However, the regional disparities in economic growth and unemployment rates have increased considerably in Finland since the depression of the early 1990s; see e.g. Kangasharju and Pekkala (2004) and Tervo (2005). Today, there are pressures to concentrate higher education and research into fewer units, which probably implies that there will be a decline in the number of universities and polytechnics in the future, particularly in the peripheral regions.
The polytechnic reform has previously been evaluated by comparing the employment and earnings of graduates from the polytechnics with those who had obtained vocational college degrees in the pre-reform system. Hämäläinen and Uusitalo (2008) find that the relative earnings of vocational college graduates decrease in the field of business and administration after polytechnic graduates start to enter the labour market, which is inconsistent with the pure human capital model and can be interpreted as evidence that supports the signalling model of education. Böckerman et al. (2009) conclude that the reform had considerable positive effects on the earnings and employment levels for graduates in business and administration but no significant effects in other fields. To the best of our knowledge, no study has, however, examined the regional aspects of the reform. Hence, it is not known how the polytechnic education reform affected interregional migration streams even though these concerns have frequently been raised in policy debates.

MIGRATION?
We are interested in what ways the polytechnic reform affected the migration of recent high school graduates. In the following theoretical discussion, we will consider the possibility that this school reform affected migration not only directly but also indirectly through the changes in the level of education.
To begin with, the reform may have increased the propensity to move directly because fewer high school graduates were able to access education in their home municipality after vocational schools were gradually converted into bigger polytechnic units. That is, it is possible that people who would have otherwise attended local vocational school will now have to move to another town in order to attend a polytechnic. Also, the incentives for school-to-school moving may have increased because (free) higher education became more available.
Yet it is known that the reform expanded higher education, especially to regions that did not have a university, which may have reduced the need of some high school graduates to move in order to obtain higher education. However, regardless of the reform, the poorer educational opportunities in the peripheral regions may have induced migration to the central areas, where most institutions of higher education are located. Hence, it is important to control for the regional differences in the educational and economic opportunities in the empirical analysis.
If the reform affected the school-to-school migration, it is likely that it also had an impact on the school-to-work migration, because those who have moved in the past are more likely to move again (see, e.g., DaVanzo, 1983;Haapanen and Tervo, 2012). This pattern would be consistent with the "hobo syndrome" reported by Munasinghe and Sigman (2004).
In the long run, the polytechnic reform may also operate indirectly through the changes in the level of education. If the reform generally increased the level of education of young adults, this increase may in turn make them more likely to move. Note that if all the vocational colleges were simply relabelled as polytechnics, then this indirect, educational effect should be zero but the overall effect of the reform on migration could still be non-zero. Extensive prior analyses suggest that the propensity to move increases with the level of education (e.g., Jaeger et al., 2010;Faggian, McCann and Sheppard, 2007;Tunali, 2000). 5 However, only recently have policy reforms provided evidence in support of the positive causal relationship; see Hickman (2009), Machin et al. (2011), and Malamud and Wozniak (2010. Although our aim is not to investigate the effect of education on migration in general, 6 it is worth discussing the explanations provided for the observed positive relationship briefly. The first one is the existence of a greater earnings differential between regionsthus greater potential benefits from moving -for the highly educated (Armstrong and Taylor, 2000, p. 155). Education is a form of general human capital, which is easily transferable to different geographical locations. Second, education increases a person's capability of obtaining and analysing employment information, and of using more sophisticated modes of information and search methods (Greenwood, 1975, p. 406).
Hence, highly educated workers may have a better access to information about the potential job prospects and living conditions in other regions.
Third, a higher level of educational attainment may open up new opportunities in the labour market (e.g., Greenwood, 1975, p. 406). As education improves, the market for individual occupations at each level of education tends to become geographically wider but quantitatively smaller in a given location (Schwartz, 1973(Schwartz, , p. 1160. Lastly, psychic costs resulting from the agony of departure from family and friends are likely to be nonincreasing with education (Schwartz, 1973). Education may also reduce the importance of tradition and family ties and increase the individual's awareness of other localities and cultures.

DATA
The empirical analyses are based on the Longitudinal Census File and the Longitudinal Employment Statistics File constructed by Statistics Finland. These two basic register files were updated annually from 1987 to 2004. By matching individuals' unique 5 See also the reviews by Greenwood (1975;1997). Finnish migration has been studied recently by Ritsilä and Ovaskainen (2001), Pekkala and Tervo (2002), Hämäläinen and Böckerman (2004), Haapanen and Ritsilä (2007), Jauhiainen (2008) and Haapanen and Tervo (2012). However, none of these studies has used education reforms to examine migration patterns. 6 Instead of estimating the effect of education on migration using the reform as an instrument, we estimate the reduced-form specifications of the polytechnic reform on migration. personal identifiers across the censuses, these panel data sets provide a variety of reliable register-based information on the residents of Finland. That is, contrary to surveys, for example, the comprehensive register-based data contain very little measurement error; cf. also Malamud and Wozniak (2010, p. 14). Furthermore, register data on spouses and the region of residence are merged with the individual records.
The working sample comprises a seven percent random sample of the individuals who resided permanently in Finland in 2001. 7 The sample was further restricted to the individuals who had completed high school (general upper secondary school). 8 With few exceptions high school education is required for tertiary-level studies. In the following analysis we focus on 18 to 21-year-old graduates 9 from 1988 to 1998. During this period, the availability of higher education made its dramatic rise; see Figures 1-2 above.
Throughout the analyses, the migration event is defined as long-distance migration between the 18 Finnish NUTS3 regions, following, for example, Nivalainen (2004). See Appendix, Figure A1, for a map illustration of migration patterns using the NUTS3 regional classification. 10 These migration flows allow us to examine the changes in the geographical distribution of human capital. Focusing on migration between the NUTS3 regions is also practical, because the location of the educational institution where an individual graduates is known at this regional level in the data. Furthermore, migration of shorter distances between municipalities or sub-regions most likely reflects housing market conditions rather than labour market prospects. However, we will also check robustness of the results below to the use of NUTS3 classification instead of NUTS4.
The average propensities to move during the matriculation year or the following two years are illustrated in Figure 3. It shows a marked increase in the interregional migration rates over time as the polytechnic reform was executed. The regional differences in the migration rates are also substantial. Comparison of time trends in Figures 2 and 3 suggests that there is a positive relationship between the regional expansion of higher education and the interregional migration. However, to investigate whether this relationship still holds after the potential confounding factors have been fully controlled for, a specification of an econometric model is required.

EMPIRICAL APPROACH
A significant proportion of high school graduates migrate in order to receive further education. To understand the implications of the polytechnic reform of the 1990s for interregional migration, we model the migration propensities during the matriculation year and the following years using probit models. Therefore, we assume that migration propensity is determined according to a latent variable * ijyt M : is a dummy variable indicating whether an individual i matriculated in region j in year y migrates ( * M > 0) or not ( ≤ * M 0) t years after the matriculation.
First we estimate the school-to-school migration propensities for the matriculation year and the following two years (t = 0, 1, 2). Later we extend the follow-up period, which also allows us to investigate the changes in the school-to-work migration.
The variable of interest, jy Z , measures the supply of polytechnic education for an individual i when graduating from high school. The supply is measured as the number of new polytechnic study places in the region of residence in the year of the matriculation. Note that the reform may have also attracted persons other than the recently matriculated individuals to enter higher education (i.e. polytechnics). To control for this, we have later defined the supply of polytechnic education not only during the matriculation year but also over a longer three-year period; see the robustness checks below.
All the control variables, ijy X , are measured in a year before an individual matriculates from high school, so that the consequences of migration are not confused with the causes of migration. This decision also ensures that the supply of polytechnic education does not affect the (future) values of the control variables and hence bias the results.
Following, for example, Nivalainen (2004) and Haapanen and Tervo (2012), we use the standard set of covariates. Concerning personal characteristics, we control for gender, age and annual earnings subject to state taxation. Household characteristics comprise marital status, having children, and a spouse's labour income, employment status and the level of education. It is important to control for the household income level.
Otherwise, the differences in the ability to finance the migration costs can partly create the observed positive association between the reform and migration. Another potential determinant of migration is prior scholastic achievement. Matriculation exam scores 11 from high school are therefore used as the measure of achievement. It is expected that the ability of an individual is positively correlated with migration because of his or her attendance to university. 12 In addition, we control for the effects that are specific to the matriculation year ( y λ ), the matriculation region ( j γ ) and the duration of time after the matriculation ( t τ ). Since interregional mobility tends to closely follow cyclical fluctuations in the economy (Milne, 1993;Venhorst, van Dijk and van Wissen, 2011;Saks and Wozniak, 2011), matriculation year fixed effects are used. The regional fixed effects pick up all the regional differences in the migration intensity of matriculated students that are stable over time. Time trend dummies t τ are added to the model to capture the number of years passed after graduation. 13 These dummies allow for the general changes in the migration rates over time after matriculation (cf. Haapanen and Tervo, 2012).
Finally, we also use several sub-regional (NUTS4) characteristics, such as the local unemployment rate and the share of service sector workers, as well as whether the individual matriculates from his or her sub-region of birth, which captures otherwise 11 The matriculation examination is a national compulsory final exam taken by all students who graduate from high school. The answers in each test are first graded by teachers and then reviewed by associate members of the Matriculation Examination Board outside the schools. The exam scores are standardized so that their distribution is the same every year. The range of the matriculation exam scores is 1-6. 12 On average, universities tend to be located further away from high school graduates than lower level educational institutions. 13 Because of low number of observations, a single duration dummy is used for t > 10. unobserved differences in migration behaviour; see Appendix (Table A1) for the detailed definitions of the control variables and their mean values.
The identification strategy is based on the assumption about the exogeneity of the polytechnic reform. Consequently, we assume that the supply of polytechnic starting places is exogenously determined after controlling for other factors potentially influencing migration decisions. For the correct identification of the effect of the reform it is, however, not necessary that the supply is independent of the fixed regional characteristics or those related to the matriculation year since we control for such factors with two sets of fixed dummies. Still, some time-varying regional characteristics could be related to the changes in the regional expansion of polytechnics over time. If such characteristics also influence migration decisions, their exclusion could violate the exogeneity assumption. Thus, following, for example, Black et al. (2005) and Machin et al. (2011), we assess what factors, if any, predict the expansion of polytechnics across regions and over time. Table 1 reports the results from linear panel regressions of the supply of polytechnic starting places on other regional characteristics. In the first column we regress the polytechnic starting places on regional value added, its growth rate, unemployment rate, total population, population aged 19-24-years, and a dummy variable indicating whether the region has a university. In the second column we add the year dummies to the model and in the third column we also add the full set of fixed effects for the NUTS3 regions (but exclude the university dummy because it does not change over time within regions). 14 --- Table 1 around here --- The results show that after controlling for the fixed time and regional effects, only population size is a statistically significant determinant (Column 3 of Table 1). Other factors play little role. Thus, the number of polytechnic starting places positively correlates only with the size of the region after controlling for other regional factors. To rule out the possibility that the omission of the population size would bias the estimates, we have added it to the set of regional controls together with the number of 19 to 24-

12
year-olds in the NUTS3 region that captures the differences in the regional demand for polytechnic education.

Baseline estimates
To begin with the short-run effects, Table 2  shows that the migration probability is, on average, influenced by the regional supply of polytechnic education during matriculation. The average marginal effect is positive and significant: 0.7 percentage points per 1,000 study places in the region.
--- Table 2 around here ---To explore the long-run effects of the polytechnic reform on the migration probability of the matriculated students, we then proceed to study the effect over a longer observation period. Since the last year in the data is 2004 we are able to follow those individuals who matriculated, for example, in 1998 and 1988 for 7 and 17 years, respectively (on average 11.7 years). Again, the supply of polytechnic education is measured during the matriculation year. The results from the six specifications (1-6) reported in Table 3 correspond to those in Table 2. The preferred specification (6) points out that the effect of the polytechnic reform on migration is smaller, and insignificant, in the long run after controlling for the relevant covariates.
--- Table 3 around here --- 15 The average marginal effects were computed as averages over all observations as discussed in Cameron and Trivedi (2005, p. 467).

13
To better illustrate the quantitative magnitude of increase in the supply of polytechnic education, we have also computed the short-run and long-run elasticities (Table 4).
First, it is useful to note that the regional average of the polytechnic study places has grown from zero to roughly 1,800 between 1990 and the early 2000s. Hence, the shortrun marginal effect (0.0068) implies that an increase of 1,800 on polytechnic places has enhanced the annual migration rate in the Finnish regions by 1.2 percentage points. This represents a substantial increase, because the baseline migration rate is 3.7 percent for those who matriculated in 1988. Second, the estimated semi-elasticity implies that a 1,000 increase in the new polytechnic places has resulted in a 12 percent increase in the migration rate. Finally, the short-run elasticity of migration with respect to increase in new polytechnic places is estimated at 0.147. These short-run effects are significantly different from zero, whereas the long-run elasticities are all insignificant.

Sensitivity of the baseline results
To study the robustness of the baseline results reported in Column 6 of Tables 2-3, we have estimated several additional model specifications (Table 5). In Panels A-B we have experimented with changes to the calculation of standard errors. In Panel A robust standard errors are used instead of robust and clustered both by NUTS3 regions and years as in the baseline. In Panel B non-robust standard errors are applied instead of using the baseline specification. The short-run estimates remain all significant at the 5 percent level and the long-run estimates remain insignificant.
--- Table 5 around here ---To allow for spatial correlation in Z, we have constructed an additional control that measures the number of new polytechnic places in the neighbouring regions (Panel C); cf. Oakes (2004Oakes ( , p. 1935. Again, the findings remain intact. Both in the short-run and long-run cases the neighbouring effect is estimated to be insignificant.
In Panels D-E the regional supply of polytechnic education is measured as a three-year average rather than during the matriculation year. The motivation for this specification is that not all individuals make their schooling and migration decisions immediately after matriculation, because of the limited number of study places, a voluntary decision 14 to take a year off from school or enrolment in military service. In Panel D, the supply is measured in the current region of residence and in Panel E in the region of matriculation. In both of these cases the long-run effect is insignificant, pointing out that the effect does not depend on whether the supply is measured only during the matriculation year or also two years after. However, in Panel D the short-run effect roughly doubles and it is significant; indicating that the effect of the reform could be higher than the one reported in Table 2. This having been said, this measure can also reflect reverse causality: the matriculated move to the regions with a high number of starting places. The small, insignificant short-run effect reported in Panel E supports this conclusion.
In addition, we have made several changes to the estimation sample. In Panel F the sample also covers the matriculated from 1999-2001. This extension increases the number of observations. As a result, the estimated standard errors are slightly smaller, but the quantitative magnitude of the effects does not change. In Panel G the sample is limited to the matriculated from 1991-1998 only. This is the time period during which the transformation of the system and the rapid increase in the number of polytechnic graduates took place (see Figure 2). Thus, the time period is crucial for the identification of the effects of the reform. The short-run effect remains unchanged in this specification, but now the long-run effect (0.003) is estimated with a much greater precision and it is also significant at the five percent level. 16 In Panel H the long-run analysis is restricted to observations with number of observations for each individual (that is, t < 7). The results correspond to those in Panel G: the long-run marginal effect is significant at 10 percent level. Thus, we conclude that the polytechnic reform is likely to have a positive, but smaller effect on migration in the long run than in the short run.
Finally, we have altered the definition of migration in Panel I. When we consider shortdistance migration between NUTS4 sub-regions, instead of NUTS3 regions, the effect of the polytechnic reform is estimated to be at almost the same size as previously, but now the estimate is less precise.

Heterogeneity of the effects
Next we examine whether the effect of new polytechnic places is heterogenous with respect to the duration of time and the region of matriculation. For brevity, the results of the specification tests are reported in Table 6, but the detailed estimation results are only available upon request. In Panel A we check to see whether the effect of the reform is constant over duration of time after matriculation. Consistent with short-run and longrun results above (Tables 3-5), the estimates do point to significant heterogeneity in this respect. The effect of the reform on migration is clearly largest close to matriculation and becomes smaller as the time passes.
As regards to the spatial differences, Panel B implies that there are no significant differences in the size of the effect between the university and non-university regions. In Panel C the effect of the reform is allowed to vary across regions more freely. We find that the estimates do not, overall, reveal significant regional differences in the effect.
However, a closer look at the estimated regional parameters reveals that in the short run (school-to-school) migration has increased particularly for the matriculated individuals from the regions of Oulu and Kainuu. The long-run estimates reveal a different picture. Now the migration rates increase for the matriculated from the Helsinki metropolitan area and decrease particularly for the matriculated from the regions of Etelä-Pohjanmaa and Joensuu (i.e. Pohjois-Karjala); see Appendix, Figure A2 for a map illustration.

Effect on residing in the region of matriculation
So far we have considered the effect of the polytechnic reform on the migration propensity in the short run and long run. Thus, the emphasis has been particularly on the intensity of migration. To understand the effects of the reform on the regional distribution of labour better, it is important to acknowledge that a significant proportion of the school-to-school migrants may return to their region of origin after graduation from specialized education. Therefore, we will consider next the extent to which the reform affected the propensity of residing in the region of matriculation. The analysis is parallel to the baseline probit models used earlier (see e.g. Columns 6 in Tables 2 and   3), but now the dependent variable is redefined as a dummy indicating whether an 16 individual resides in the region of matriculation. That is, the estimation samples and also the control variables remain unchanged. Table 7 reports the short-run and long-run average marginal effect of the reform on the propensity to reside in the region of matriculation (Model 1). Both marginal effects are now significant at the 5 percent level and they are larger than the estimated effects on migration intensity. Namely, an increase in polytechnic starting places by 1,000 in the region has, on average, decreased the propensity to reside in the region of matriculation by 1.0 and 1.4 percentage points in the short run and in the long run, respectively. The significant negative marginal effects suggest that the reform has resulted in a spatial redistribution of labour.
--- Table 7 around here ---One goal of the reform was to improve the supply of highly educated labour in the nonuniversity regions. To investigate whether this goal was reached, in Model 2 we have interacted the supply of polytechnic education with information on whether or not the matriculation has occurred in a university or non-university region. Although the differences are not significant, the results suggest that, the reform may have increased the out-migration of individuals from the non-university regions. Figure A3 in the Appendix shows that the out-migration has been particularly intense in the regions of Kainuu and Keski-Pohjanmaa.

CONCLUSIONS
In this paper, we explored the effect of the polytechnic education reform on the migration of graduated high school students in the Finnish context. The reform gradually transformed former vocational colleges into polytechnics offering a Bachelor's degree and expanded the supply of higher education to all regions. The estimates confirm that the reform provides relevant exogenous variation in the regional supply of education. The results showed that the expansion of polytechnic education increased the migration rate of high school graduates. The migration propensities were increased particularly close to graduation from high school. Furthermore, the estimates based on the most intensive period of transformation also point to a smaller, positive effect over a longer period. The study made no attempt to distinguish whether the effects of the reform on migration were due to extension in the length or improvement in the quality of education.
One important reason for the creation of the polytechnic schools was to decrease the brain drain from the less developed regions to the metropolitan areas. The results suggest that this policy aim has not been fulfilled. However, the increased migration

Variables
(1) (2) (3) (4) (5) -0.003 -0.015*** 0.010*** -0.002*** 0.007*** 0.007*** Supply of polytechnic education (0  -likelihood -14,857.6 -14,225.8 -14,230.5 -13,930.3 -13,673.7 -13,552.1 LR-test over restricted specification p < 0.001 (df=2) Notes: Average marginal effect is computed as average over all observations using predictions from probit model. Sample: Individuals are observed during the matriculation year and the following two years. The number of observations is 61,509 in all estimations. Dependent variable: NUTS3 migration during the current year. The explanatory variable of interest reported in the table: The number of 1 st year polytechnic students in the NUTS3 region (in 1,000). The set of controls is defined in Appendix, Table  A1. * (**, ***) = statistically significant at the 0.10 (0.05, 0.01) level. Robust standard errors reported in parentheses allow for clustering at the matriculation year and regional level. p = p-value; df = degrees of freedom. (i) LR-test of (3) vs. (1). (ii) LR-test of (4) vs. (2).  -likelihood -60,175.7 -59,123.3 -58,246.4 -57,908.5 -57,593.9 -57,108.1 LR-test over restricted specification   Tables 2 and 3. M = NUTS3 migration during the current year (0, 1). Z = The number of 1st year polytechnic students in the NUTS3 region (in 1,000). (i) 3-year follow-up period; cf. Table 2. (ii) Extensive follow-up period; cf. Table 3; * (**, ***) = statistically significant at the 0.10 (0.05, 0.01) level. Robust standard errors reported in parentheses allow for clustering at the matriculation year and regional level. Using supply of polytechnic education in the neighbouring regions as an additional control (iii) (0.0026) (0.0013) 0.0159*** 0.0024 Panel D: Three-year supply of polytechnic education in the region of residence (iv) (0.0059) (0.0036) 0.0024 0.0004 Panel E: Three-year supply of polytechnic education in the region of matriculation (iv) ( Using NUTS4 migration (short-distance migration) as the dependent variable (0.0034) (0.0016) Notes: Average marginal effect of the polytechnic reform on migration from probit model is reported. Same controls are used as in the last specification of Tables 2 and 3. In Panels A-C and F-I, the absolute number of polytechnic study places in the region where individual matriculates (in 1,000, measured during the matriculation year) are used. (i) 3-year follow-up period; cf. Table 2. (ii) Extensive follow-up period; cf. Table 3. (iii) LR-test indicates insignificance of the neighbouring effect in both cases. (iv) The regional supply of polytechnic education is measured as a three-year average rather than during the matriculation year. * (**, ***) = statistically significant at the 0.10 (0.05, 0.01) level. Robust standard errors reported in parentheses allow for clustering at the matriculation year and regional level.  -0.0092** -0.0126** Supply of polytechnic education × matriculated from a university region (ii) (0.0038) (0.0063) -0.0130* -0.0196** Supply of polytechnic education × matriculated from a nonuniversity region (ii) (0.0079) (0.0086) Average propensity to reside in the region of matriculation 0.8969 0.7326 Notes: Average marginal effects from two probit models are reported. Dependent variable: Resides in the region of matriculation. Same controls are used as in the last specification of Tables 2 and 3. (i) The explanatory variable of interest: the number of 1st year polytechnic students in the NUTS3 region (in 1,000). In Model 2, this variable has been interacted with dummies indicating whether or not an individual matriculated from a university region. (ii) The difference is neither statistically significant in the short run nor in the long run. See also notes to Table 5.  1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 Polytechnic students University students   Figure A1.
Source: own sample data. (1) Observations from the matriculation year and the following two years; (2) All possible observations after matriculation. The explanatory variables also include region and year of matriculation dummies, and duration time dummies. .

Satakunta
Keski-Pohjanmaa Kymenlaakso Figure A2. The estimated short-run and long-run average marginal effect of the polytechnic reform on the probability to move by NUTS3 matriculation region. a) Short-run follow-up period b) Long-run follow-up period Notes: Average marginal effects are based on the interaction of supply of polytechnic education with the matriculation region dummies. Same controls are used as in the last specification of Tables 2 and 3. See also Table 6 and notes to Table 5. Underlining of the name of the region indicates significance at 10 % level.

Keski-Pohjanmaa
Kymenlaakso Figure A3. The estimated short-run and long-run average marginal effect of the polytechnic reform on the probability to reside in the region of matriculation by NUTS3 matriculation region. a) Short-run follow-up period b) Long-run follow-up period Notes: Average marginal effects are based on the interaction of supply of polytechnic education with the matriculation region dummies. Same controls are used as in the last specification of Tables 2 and 3. See also Table 7 and notes to Table 5. Underlining of the name of the region indicates significance at 10 % level.