A NON-EXPERIMENTAL EVALUATION OF UNEMPLOYMENT RISK IN CRETE AND THE IONIAN ISLANDS: REGIONAL EVIDENCE FOR GREECE

The basic aim of this paper is to investigate the impact that individuals’ level of education and participation in training programmes (apprenticeship, intra-fi rm training, continuing vocational training, popular training) have on their job prospects in Crete and the Ionian Islands, both among Greece’s top tourist destinations, during the implementation of the fi rst Community Support Framework CSF (1989-1993). We try to see whether the level of education itself and participation in training programmes increased the chances of fi nding a job. More specifi cally, we examine the social and demographic characteristics that increase the chances of someone in the examined population fi nding a job, how those chances change (if they do) after the introduction of training courses and, also, whether university graduates, in contrast to most of the rest of the EU member states, face greater diffi culties in fi nding a job than non-university graduates, as a series of studies for Greece conclude. In our analysis we use individual anonymised records (micro-data) of the Labour Force Survey (LFS) for both employed and unemployed at NUTS-2 level.


Introduction
The aim of the paper is to study the impact that education and training programmes (apprenticeship, intra-fi rm training, continuing vocational training-CVT, popular training) had on the labour market in the Greek regions of Crete and the Ionian Islands during the implementation of the Community Support Framework (CSF) CSF-1 (1989-93).Namely, we try to see whether the level of education itself and participation in training programmes increased the chances of fi nding a job.The vocational training programmes of the CSF-1 in the regions under examination started in March 1990 and ended in March 1994.During the examined time period, both regions belonged to Objective 1 of the EU Structural Funds.The total population of the above two regions constitutes 7.5% of Greece's total.
We work fi rst at the regional level, and then at the urban, semi-urban and rural levels.
The main questions to be answered are: (i) What are the social and demographic characteristics that increase the chances of someone in the examined population fi nding a job?
(ii) How does the participation in training courses affect the chances of getting an employment?
(iii) Whether university graduates, in contrast to most of the rest of the EU member states, face greater diffi culties in fi nding a job than non-university graduates, as a series of studies (see Meghir et al., 1989;OECD, 1990;Iliades, 1995; IN.E./GSEE-ADEDY, 1999; Katsikas, 2005) for Greece conclude.
We test the human capital theory, which underpins many of the important developments in modern economics and provides one of the main explanations for wage and salary differentials by age and occupation, and the uneven incidence of unemployment by skill (education and training).We try to research whether the more educated and the more trained a person is, the higher the probability of him or her fi nding a job.
A O P 1 8 ( 4 ) , 2 0 1 0 , I S S N 0 5 7 2 -3 0 4 3 Previous labour market studies for Greece were based on qualitative research and LFS aggregated data.The analysis of investigating the impact of training on the Greek labour market -at NUTS 2 level -is based on the micro-data of the Greek LFS.The access to the individual anonymised records of the Greek LFS was not allowed to researchers until the summer of 2005, due to the Data Protection Act.Also, all training programmes in Greece are co-fi nanced by the EU funds and so we try to see what happened with the EU money during the period of the CSF-1 in the domain of training.Thus, we do not simply research into the early 1990s but, for a specifi c reason, namely the management of EU funds in the fi rst programming period.There is the opportunity for other researchers to compare the 2nd and 3rd CSFs with CSF-1 based on empirical analysis.
The article starts with the issue of over-education and why it is important to this research.Then, we examine the impact of training programmes on the employment prospects of individuals in the EU and the rest of the OECD according to a series of studies; the results are based on both cross-sectional and longitudinal data.We discuss the vocational training policies for the unemployed in Greece.Finally, we refer to the socio-economic characteristics of the examined regions and follow a logistic regression for the years 1988 and 1992 -based on micro-data of the Greek LFS -for the two regions under study.The article concludes with the impact of training on employment probability in Europe and the examined areas, and ends with some general comments on the merit and value of this study.

The definition of educational mismatches and over-education
A great deal of American and European empirical evidence, in fact, is now available on the subject of over-education.On the other hand, as Joop Hartog showed, "a solid relation (of the over-education / under-education literature) with a formal theory of the labour market is lacking" (Hartog, 1997).A further diffi culty is the confusion between the term 'over-education' specifi cally, and very different notions like 'qualifi cation infl ation' or 'credentialism' (Green et al., 1999).According to human capital theory, over-education is not a permanent occurrence and is the result of a temporary poor match between the employer and the employee.This appears to go against the empirical evidence, which suggests there is always a large percentage of labour force that is over-educated (Green et al., 1999).Hartog (1997) extensively surveyed the literature and found signs that over-education has grown in frequency over the years (i.e., especially since the 1970s, whilst under-education is less frequent, in Europe, anyway).In the UK, there is less evidence on under-education, but Sloane et al. (1995) showed a 17% of under-education for the labour market generally.A reduced evaluation of 8-10% of under-education was shown by Groot and van den Brink (1997).In the vast majority of studies, over-education is a great deal more frequent than under-education and this outcome fi ts in with that of Green et al., 1999.According to the OECD (1990, p. 67, Table 2.3), in Greece -contrary to what happens in many other European countries -the unemployment rate of university graduates was higher than that of the less educated, whilst, mainly after the late 1980s, a lot of graduates of tertiary education, especially of certain old traditional specializations, faced problems of absorption into the labour market (Iliades, 1995).Also, according to Katsikas (2005), university graduates in Greece face greater diffi culties fi nding a job in comparison to the less educated.Meghir et al. (1989) analysed the main determinants of female participation in the labour force and male unemployment duration in Greece using data from the 1981 Greek LFS.An interesting fi nding is that male unemployment duration increased with education.Also, according to the study of IN.E./GSEE-ADEDY (1999), based on the processing of ESYE (National Statistical Service of Greece) aggregated data, the probability of an unemployed person becoming long-term unemployed depends on his/her age, gender and family status.Contrary to the common perception, this probability did not depend on the level of education.
Greek farming and especially Greek industry consisted of many pre-eminently small businesses of traditional activities, which did not require administrative and technical staff with higher education and specialization (Kanellopoulos, 1984).Besides, the family character of many Greek businesses made their owners avoid hiring staff (including those with high qualifi cations) or implementing innovative ideas of highly skilled people, with the result that industry was unable to create enough new positions for people with relatively high specialization and unable to absorb the increased numbers of graduates (Kanellopoulos, 1984).Exactly the opposite happened in the public sector, where many new positions were created to absorb unemployed graduates.Although this waiting (queuing for a public sector job) raised the apparent unemployment of graduates (see, for instance, Krueger and Summers, 1987, p. 44), some of them held tempor¬ary jobs, often in the concealed economy (Glytsos, 1990).Moreover, the public sector limits its action to essentially bureaucratic competence and activities or the provision of non-exchangeable services internationally.Greece seems to manifest over-edu¬cation by any of the criteria mentioned above.The relative remuneration of university graduates was decreasing through time, mainly because of their over-supply (Glytsos, 1990).
For more information on the causes of graduate unemployment, see Johnes et al., 1987;Sanyal, 1987;Dolton and Vignoles, 2000.4. The human capital approach and the human capital theory During the late 1950s and early 1960s, the current neoclassical theory of the labour market emerged with the development of the human capital theory.Gary Becker (1964-2nd ed., 1975) published a book with the title "Human Capital", which developed a theory of human capital formation and analysed the rate of return on investment in education and training.However, investment in human capital remains a controversial issue (Woodhall, 1987).
A O P 1 8 ( 4 ) , 2 0 1 0 , I S S N 0 5 7 2 -3 0 4 3 Whilst the human capital literature has highlighted a number of productivityrelated characteristics, human capital theorists place most emphasis on the importance of education and training as the main component of productivity (Blaug, 1975).Education, it is suggested, provides the basic skills of reading and writing, cognitive skills, and the "ability to learn" which will increase an individual's productivity in all jobs (general human capital), whilst vocational education, on the other hand, will increase an individual's productivity in a narrower range of jobs by providing more specifi c skills (specifi c human capital).Becker (1962) distinguishes between general and specifi c human capital of workers, and within specifi c human capital, between employer-and employee-fi nanced on-the-job training.Most broadly, the theory of specifi c human capital predicts that where the fi xed costs of employment, due to on-the-job training, are the greatest, unemployment is the lowest (Rees, 1973, pp. 118-20).Following Becker's (1964) analysis of the economic role of human capital, particularly education, there is now a considerable amount of empirical research on the closely related topics of education and skills (see Prais, 1995;Murray and Steedman, 1998) and, more specifi cally, the increasing role of skilled labour in the economy (Berman et al., 1994;Machin, 1996;Green et al., 1998;Machin and van Reenen, 1998).

Impact of training at micro-economic level
The early European evaluation studies are mostly characterized by the fact that research was not based on longitudinal and non-experimental data, as is the norm in the secondgeneration studies (see section 5.2), but on cross-sectional and (quasi) experimental data.Experimental evaluations are common in the U.S. but scarce in Europe (Bjorklund and Regner, 1996).The micro-economic studies on active labour market policies (ALMPs) were effectively summed up in OECD (1993a) and Fay (1996).Regarding training, the basic conclusion was a frequently weak return to the training of the unemployed.In the majority of cases, the most signifi cant force decreasing the return was deadweight (i.e., a trained job-hunter is taken on but would have been employed in any case without training) (Jackman et al., 1996).

The findings of European training evaluations (first-generation studies)
Among the ALMPs, the greatest advance has been in the evaluation of training programmes, whilst the majority of the training studies focused on the impact of training on future remuneration or on the likelihood of re-employment.The impact on the duration of the following employment period, too, has just been examined in studies done lately (e.g.Kaitz, 1979;Ridder, 1986;Card and Sullivan, 1988;Ham and Lalonde, 1991;Gritz, 1993;Bonnal et al., 1994;Torp, 1994;Zweimuller and Winter-Ebmer, 1996) -it is important to separate the length of employment from the duration of job tenure (Cockx et al., 1998).
Examination of accessible micro studies on training forces us to realize that it has been remarkably diffi cult to be clear about the foreseeable positive impact on those taking part (Jackman, 1995).It could be thought more extraordinary, according to Calmfors and Skedinger (1995), in view of the powerful theoretical points suggesting a positive impact when programmes were concentrated on a set of outsiders like these, that there is no more defi nite evidence on the impacts of centring on the young.
A large number of different sorts of training programmes and their impacts were studied by OECD (1993a).In general, it was found that programmes aimed at a few people only whose requirements are easily recognizable and at quite a high cost per person, frequently seemed to succeed relatively well in improving the remuneration and job possibilities of a number of the participants (this might account for the fact that training programmes in Norway, which were not that large, seemed to have succeeded much more effectively than in Sweden; Calmfors, 1995).In contrast, wider programmes involving more participants at quite a low cost per person normally appeared to make almost no difference (if any) to the prospects of those involved (Jackman, 1995).According to Rosholm and Skipper (2003), training raises the employment rate of participants but this effect disappears over time and this would indicate a locking-in effect.
These fi ndings can be explained in different ways.One is that the characteristics of the unemployed differ to a great extent and taking into account their age, education and occupational backgrounds, just a few were able to gain from more training.Therefore, the only training programmes that had economic returns were those aimed at particular groups.Another explanation is that due to greater returns to training, only programmes with large inputs, i.e., targeted programmes, succeed.For instance, this could apply where the trainees are not used to the kind of skills they are learning, or for those not used to gaining skills by formal means (Jackman, 1995).
It follows that a labour market policy desirous of putting all unemployed people on a programme, or giving them temporary work, cannot be largely made up of effective training programmes (OECD, 1993b;Calmfors, 1994).

Findings of recent European Programme evaluation on training (second-generation studies)
In contrast to the early European evaluation studies -cited in section 5.1 -cross-sectional data is hardly to be found and training research in Europe has replaced it with the more useful longitudinal data, allowing for the possibility that impact assessments will be more robust (Kluve and Schmidt, 2002).Namely, the studies in section 5.2 examined the same population groups over time, except those of Winter-Ebmer (2006) and Cueto and Mato (2009), which used only one reference year in their research; also, only one study (that of Malmberg-Heimonen and Vuori, 2005) used experimental data.
These results show that the more expensive programmes having a signifi cant amount of training appear to be most effective at increasing employment prospects (see Kluve et al., 1999;Brodaty et al., 2001;van Ours, 2001;Kluve and Schmidt, 2002;Raaum and Torp, 2002).Lately, national studies do not all fi nd positive impacts (Gerfi n and Lechner, 2000;Regner, 2002); but bearing in mind that job creation and subsidies for employment in the public sector usually do not succeed (Kluve et al., 1999;Brodaty et al., 2001), especially if their one aim is to remove unemployed people from the register (Lechner, 2000), training seems to have a signifi cant impact.
Concerning the most recent research (Weber and Hofer, 2003;Graversen, 2004;Graversen and Jensen, 2004;Hujer et al., 2004;Rosholm and Svarer, 2004;Centeno et al., 2005 -on earnings as well;Hogelund and Holm, 2005;Aakvik and Dahl, 2006), there is no impact of training on employment probability on the European labour markets.Also, according to a series of studies (Lechner et al., 2005 - [Cockx, 2003;Hamalainen and Ollikainen, 2004;Leetmaa and Vork, 2004;Albrecht et al., 2005 -on employment effects for young men (research on earnings as well, but no impact on income effects); Arellano, 2005 -higher positive effects for women than for men; Cavaco et al., 2005;Fitzenberger and Speckesser, 2005 -more  In conclusion, up-to-date evaluation studies point to minor impacts of European training policies and they are most likely less signifi cant and not always as positive as those responsible for designing them had wished.Although the cross-national fi gures show a few positive results of programmes, it is impossible to disregard the more negative results.The fi ndings allow us to conclude that training programmes seem to have some positive effects on employment and no effects on earnings.Moreover, effects diminish over time.The negative effects reported by several evaluations can be explained, on the one hand, by a locking-in effect, and on the other, by the fact that some participants seem to enrol in training merely in order to collect unemployment insurance benefi ts (Cueto and Mato, 2009).The conclusions based on the recent studies are somewhat similar to those of Heckman et al. (1999) and Stanley et al. (1999) for the U.S.

Vocational training policies for the unemployed in Greece
The structure of expenditures on "active" interventions in 1997 shows that the level of expenditures in Greece (0.35%), as a percentage of the GDP, is behind that of the EU-15 average (1.13%) concerning all specifi c interventions, with the exception of "measures for the young" (youth vocational education and training, etc.; 0.10%) which are comparable to the European average (0.13%).Furthermore, there is an extremely low level of expenditures on the training of adults (0.06% for Greece in comparison to 0.29% for the EU-15) (OECD, Employment Outlook, 1999).
The system of CVT in Greece was developed mainly due to its incorporation in Community funding programmes (Iliades, 1995;Chletsos, 1998;Papakonstantinou, 1998).Policies concerned with training and retraining of the unemployed have been confi ned to continuing training programmes.Vocational training programmes for the unemployed were wholly unconnected with employment policies, and were thus wasteful of training resources (Gravaris, 1991, p. 37;Christodoulakis and Kalyvitis, 1995;Balourdos and Chryssakis, 1998).This is refl ected in the fact that the unemployment rate for those (20-29 years old) with complementary vocational training in Greece was 20%, compared to 14% for those with only compulsory schooling; the corresponding fi gures for the EU were 11.5% and 23.5% ((Economic and Social Committee of Greece, 1998, p. 31).
Particularly with regard to training programmes for the unemployed in Greece, the method of identifying skill requirements, on the basis of which the programmes were offered, was wholly inadequate.It was based on changes in labour force categories derived from the LFS, on estimates of the impact of investment programmes on employment (where these existed or where such estimates were possible) and on Job Market Surveys.These last record shortages of skills on the basis of company estimates of their own shortages, which were often inaccurate or did not correspond to the capacity of the fi rms to utilise the skills demanded (Linardos-Rylmon, 1998).

The Region of Crete
The Region of Crete contains the NUTS-3 areas of Iraklio, Lasithi, Rethymno and Chania.In 2001, the per capita GDP was equal to 67% of the EU-15 average (69% for Greece as a whole), whereas in 2003 the regional GDP per head was 97% of the country's mean (98% in 1995) and 78% of the EU-25 mean (80.9% for the country as a whole).With Iraklio as its centre, 5.5% of the country's population is concentrated in the region with a distinct upward trend, noting that between the censuses of 1991 and 2001, the region had the second highest rate of population increase (11.3%) nationally after the Southern Aegean.It produces 5.3% of the national GDP, 7.9% of the agricultural produce, 1.3% of manufacturing and 5.9% of services.Seventy-fi ve percent of its produce comes from services, with its signifi cant role in tourism, noting that 15% of its gross product stems from hotels and restaurants (2003).It accounts for 7.5% of cultivated land in the country and 35% of total olive oil production (fi rst in the country in 2001).Unemployment in Crete fell for the third consecutive year by 0.2 units in 2001, to 6.7% of the workforce (10.5% for the whole of Greece), the second lowest proportion nationally (source: www.economics.gr).

The Region of the Ionian Islands
The Region of Ionian Islands contains the NUTS-3 areas of Zakynthos, Corfu, Kefallinia and Lefkada.The region consists of 32 small and big islands, but only 13 of them are inhabited (the four aforementioned islands plus Antipaxoi, Ereikousa, Ithaka, Kalamos, Kastos, Mathraki, Meganisi, Othonoi and Paxoi).Between the censuses of 1991 and 2001, the population rose by 9.9%, the third largest rise nationally after the Southern Aegean and Crete (source: www.economics.gr).
With Corfu as its centre, the region is inhabited by 2% of the country's population and produces 1.7% of the GDP (2003).In 2001, the per capita GDP was equal to 61% of the EU-15 average (69% for Greece as a whole), whereas in 2003, the regional GDP per head was 87% of the country's mean (85% in 1995) and 71% of the EU-25 mean (80.9% for the country as a whole).The region produces 2% of the country's agricultural produce, 0.15% of manufacturing production and 2% of services.Eighty percent of the region's produce comes from services, with an important contribution of tourism, considering that 20% of the regional GDP stems from hotels and restaurants (the second highest proportion in the country after that of Southern Aegean) (2003 data).In 2001, it accounted for 2% of cultivated land in the country.Unemployment in the region rose by 1.2 units in 2001 to 10.2% of the workforce, the largest rise in the country (11.3% in 2003; 10.5% for the whole of Greece) (source: www.economics.gr).

Numbers of records in the LFS samples
The questionnaire of the European (and Greek) LFS was greatly modifi ed in 1992 (Felstead et al., 1998).The originality of this research is that we use individual anonymised records (micro-data) of the LFS for both the employed and the unemployed (about 1.5% of the total population of each region).
Table 1 shows the numbers of records eligible for analysis in the LFS samples for the two regions under examination in 1988 and 1992.Apart from the system missing records, following the age limitation (15-64 years old) and removing the non-active population, we ended with the following numbers of records eligible for analysis in each region (in the spring, namely from the 14th to 26th week of the year): The basic aim of the econometric analysis is to test the impact that training programmes (apprenticeship, intra-fi rm training, CVT, popular training) and level of education had on people's job prospects in the Regions of Crete and the Ionian Islands during the implementation of the CSF-1 (1989-93), accounting for demographic characteristics such as age, gender, marital status and area of residence.We try to see whether participation in training programmes and level of education increased the chances of fi nding a job.In the paper, we use a logistic regression model for studying differences between those that did participate in training programmes and those that did not.Regression models allow group comparisons adjusting for demographic and socio-economic variables.It should be noted that regression-adjusted comparisons may still provide misleading results when other important variables that might have an effect are omitted.
The dependent variable takes two possible values (employed versus unemployed).The explanatory variables (six for 1992 and fi ve for 1988) are the participation in training courses (with fi ve categories, including the four types of training completed, as mentioned above, and non-participation in training courses as the reference category), six levels of education, gender, age level (four categories), marital status and place of residence (urban areas, semi-urban areas, and rural areas).
The effect of demographic variables such as age, gender, marital status, place of residence, as well as level of education and participation in training programmes on the employment status, is investigated with a logistic regression model due to the categorical nature of the dependent variable.The logistic regression model is written as: where P(y = 1|x 1 ,...,x k ) and 1-P(y = 1|x 1 ,...,x k ) denote the conditional probability for a randomly selected individual to be 'unemployed' and 'employed' respectively.The coeffi cient β k denotes the effect that a unit increase in the explanatory variable x k has on the log odds of being 'unemployed' than 'employed' controlling for all other variables in the model, and β 0 is the intercept of the model and the value of the logit when all the explanatory variables take the zero value.More specifi cally, a unit increase in the explanatory variable x k multiplies the odds by e β k controlling for all other variables in the model.
Solving the above formula with respect to the conditional probability, we have: Due to data limitations, we cannot explore the impact that the duration of courses, thematic fi elds, number of participants, and duration of unemployment period of the trainees have on unemployment.Another limitation of the research is that the data available are cross-sectional rather than longitudinal and therefore we cannot study any population changes across time.The Greek LFS data are non-experimental.

Description of the variables
We now defi ne the complete list of variables together with their coding values that we use in the model.

Explanatory variables
The reference category of each variable is underlined.Τhe working age population is between 14-65 years old.However, marking in SPSS the ages 14 and 65 we also include those who are 13 and 66 years old something which we want to avoid; so, we include people from 15 to 64.
The variable "participation in training course(s) in the past" fi rst appeared in the 1992 questionnaire; it means that the interviewee had completed one or more training courses.This is also an indication of the attitude towards training in Greece at the beginning of the 1990s.The duration of apprenticeship and intra-fi rm training had to be at least one year according to the questionnaire of the Greek LFS.The term "popular training" (laiki epimorphosi in Greek) means training courses intended mainly for elderly people independently of their level of education, where the curriculum includes largely courses in general knowledge.

Results for Crete
Tables 2 and 3 present the results of the logistic regression for Crete for 1988 and 1992.In both years, women, non-married individuals, people who lived either in urban areas or semi-urban areas were more likely to be unemployed than men, married people, and those in rural areas.The results are in accordance with the family strategies and the gender roles in traditional families in this region, as well as with the unequal opportunities and discrimination against women by companies.The higher probability of women in relation to men could also be explained by the fact that women often join the labour market earlier in relation to men.Compulsory military service and further education (not a likely explanation anymore) were the major reasons for men's delay in entering the labour market.Extended family protection, with a view to preparation for entry into the labour market, applies to both sexes, of course.The effect of urbanisation level can be explained since in the Greek agrarian sector unemployment was not properly counted, because hidden unemployment is quite high.Also in both years, people in the age group of 15-24 years old were more likely to be unemployed than people at the age of 25 to 64.
In addition, for 1988, signifi cant differences in education have been found only between primary school graduates and university graduates indicating that primary school graduates were less likely to be unemployed than university graduates, whereas all the rest of the educational variables are statistically non-signifi cant in both 1988 and 1992.
Most importantly, none of the four types of training programmes seem to reduce the odds of unemployment.The same results on training were found for other Greek regions as well (see Rodokanakis, 2009a and2009b;Rodokanakis and Tryfonidis, 2009;Rodokanakis, 2010a and2010b;Rodokanakis and Moustaki, 2010).Also, the results of the logistic regression confi rm the conclusions of the various studies for the limited impact of vocational training in Greece (see section 6).
A O P 1 8 ( 4 ) , 2 0 1 0 , I S S N 0 5 7 2 -3 0 4 3 10.2.1 Results for Crete, 1988  Tables 4 and 5 present the results of the logistic regression in the Ionian Islands for 1988 and 1992.The results for gender and marital status are mixed.In both years, people in the age group of 15-24 years old were more likely to be unemployed than people at the age of 35 to 64; the results are mixed for the age group of 25-34.
The fi ndings concerning the urbanisation level are mixed.In 1988, people who lived in urban areas were more likely to be unemployed than those in rural areas; in this case, the effect of place of residence can be explained since unemployment was not properly counted in the Greek agrarian sector.Living in semi-urban areas was not found statistically signifi cant.In 1992, people who lived in urban or semi-urban areas were less likely to be unemployed than those in rural areas.
Concerning education, for 1988, all educational variables are statistically non-signifi cant.On the contrary, in 1992, primary school graduates were less likely to be unemployed than university graduates; also, TEI and high-school graduates were more likely to be unemployed than university degree holders.Furthermore, in the Region of the Ionian Islands, only apprenticeship was found to be statistically signifi cant; namely, the participants in this kind of training courses are more likely to be unemployed than those who have not participated in any type of training programmes.

Interaction effect among variables
Only for the 1992 sample did we fi t the interaction effects between training and urbanisation level, and between training and level of education.Interaction terms were not found to be statistically signifi cant in either region.Therefore, the variable "training" does not alter the relationship between unemployment and education, as well as unemployment and urbanisation level.In other words, the chances of fi nding a job do not change when we count training as an additional qualifi cation in relation to place of residence and level of education.

Conclusions
A signifi cant number of researchers making use of accessible data and studies to examine the potential impacts of training on employment have been referred to.In spite of being restricted to only a small number of nations, micro-economic studies of effect evaluations indicate that some programmes have managed to noticeably improve employment prospects for those taking part.On the other hand, the fi ndings include a number of programmes which appear to have had almost no effect.Programmes with fairly specifi c targeting have managed positive results and this may be due to the fact that these programmes usually take account of individual requirements.However, a number of programmes that were most widely targeted have had little impact.Lastly, to establish the ways in which programmes can be made better, more research is necessary.
According to the fi ndings of the logistic regression for the Regions of Crete and the Ionian Islands, the results for gender, marital status and place of residence are mixed.On the contrary, regarding age groups, the fi ndings are common and clear except the age group of 25-34 in the Ionian Islands.
The level of education is statistically non-signifi cant for 1988 in both regions, apart from primary school graduates in Crete, who were less likely to be unemployed than university graduates.In 1992, only in the Ionian Islands the educational level is statistically signifi cant with the exception of lyceum graduates; in Crete, all educational variables are non-signifi cant.
All training variables are statistically non-signifi cant for 1992 in both regions except the apprenticeship in the Ionian Islands but even this with negative effect as well (as already mentioned in section 10, we cannot explore training in 1988 due to the limitations of data); so, the results of the logistic regression confi rm the conclusions of the various studies for the limited impact of vocational training in Greece (see on the vocational training policies for the unemployed in Greece).
The results of educational and training variables are not compatible with the human capital theory, so the more educated and the more trained persons did not improve their position on the labour market in Crete and the Ionian Islands during the time period of the CSF-1.One explanation could be the fact that the tourist sector plays a major role in the economy of the regions under examination and so very often higher education is not necessary for the local manpower to fi nd a job.Also, the labour market of the examined regions, like most of the highly attractive Greek tourist destinations, is characterised by high levels of seasonal employment.However, in 1992 in the Ionian Islands, we observe that TEI and high-school graduates were exceptions to the above mentioned fi nding, since they were more likely to be unemployed than people with a university degree.Consequently, investigation of the subsequent years is needed in order to have a clearer picture of the 1990s given the fact that, as mentioned in the introduction, the Greek LFS micro-data are now available to researchers.
Finally, the results of the interaction effect analysis show again that training is statistically non-signifi cant in relation to both urbanisation level and level of education in both regions.
The research would merit attention of a wider international readership, since the paper does offer results that are useful for comparative research among European regions.Also, the study will be valuable to those who are interested in designing and implementing training programmes for structural change investigating the defi ciencies and ineffi ciencies which have occurred in the Greek case.
status (STA 3) (Married or divorced or widows = 1, Non-married = 0) 3) Level of education STA 8A = University graduates STA 8Α1 = MSc or PhD holders STA 8B = Technological Educational Institution (ΤΕΙ) graduates STA 8C = Lyceum (12 years of schooling) or incomplete University graduates STA 8C1 = High-school graduates (9 years: compulsory education) STA 8D = Primary school graduates or incomplete primary school or never at school.4) Urbanization level of settlement system STA 9C = Urban areas STA 9D = Semi-urban areas STA 9E = Rural areas 5) Participation in training course(s) in the past STA 26A = Apprenticeship STA 26B = Intra-fi rm training STA 26C = Continuing vocational training (CVT) STA 26D = Popular training STA 26E = Non-participation in the past in training course(s) 6) Age groups STA 40A = 15-24 years old STA 40D = 25-34 years old STA 40E = 35-44 years old STA 40C = 45-64 years old The base (or reference) categories are those with which the rest of the corresponding variables are compared.The reference categories are chosen so as to match the needs of the research.
10.3.1 Results for theIonian Islands, 1988 Malmberg-Heimonen and Vuori, 2005;Steiger, 2005; 2005;Steiger, 2005; Lechner et al., 2007Cueto and Mato, 2009Cueto and Mato, 2009-a locking-in effect of trainees is shown that may be decreasing labour mobility) the employment effects of training are mixed; namely, there are positive and negative results.Furthermore, recent research on Europe has also found that training has positive effects on employment probability, although in some cases more for specifi c age groups or areas Kluve et al., 2005;Lorentzen and Dahl, 2005 -but., 2005;Lorentzen and Dahl, 2005 -butmodest effects and only on earnings; Stenberg, 2005; Winter-Ebmer, 2006 -for men and on earnings as well; Mato and Cueto, 2008 -but no effects on earnings].

Table 1
Numbers of records eligible for analysis in the LFS samples