Abstract
When information about the abilities of job seekers is difficult to obtain, statistical discrimination by employers may be an efficient strategy in the hiring and wage-setting process. In this article, we use a unique, longitudinal survey that follows the PISA 2000 students in their early educational and work–life careers. We find that a deviance in the PISA test scores from what one would have predicted based on easy-to-obtain observable characteristics influences the probability of succeeding in the transition from compulsory schooling to a firm-based apprenticeship significantly but in a non-symmetric way. Only those who had a test result below their predicted result have significantly lower chances of getting an apprenticeship. We also find evidence that the importance of hard-to-obtain information on ability is further revealed in the course of the apprenticeship.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
See Phelps (1972), Spence (1973), Arrow (1973), Aigner and Cain (1977). An indirect test of statistical discrimination is provided by the employer learning literature (Farber and Gibbons 1996; Altonji and Pierret 2001), wherein information on cognitive ability that is only observable to the researcher, e.g., Armed Forces Qualification Test scores (AFQT), is found to have increasing influence on wages as workers gain experience, indicating that workers’ true productivity is gradually revealed over time to the labor market.
An ordinary working contract can be terminated by the employer without giving reasons, while an apprenticeship contract can be terminated only under certain specified conditions (see http://www.lehr-vertrag.ch/).
As of 2008, Transitions from Education to Employment (TREE) is co-funded by the Swiss National Science Foundation (SNSF) and the University of Basel. From 2000 to 2007, the project has been financed and/or carried out by said SNSF, the Departments of Education of the three cantons Berne, Geneva and Ticino, the Federal Office for Professional Education and Technology (OPET), and the Swiss Federal Statistical Office (FSO).
As a consequence thereof, worker’s organizations (Travailsuisse and the Association of Commercial Employees) have published guidelines to sensitize firms for fair selection practices, namely not to place weight on applicant’s family background and not to overweight former school types (http://www.zukunftstattherkunft.ch). There has also been launched an online platform that allows preselection of applicants based on anonymized application dossiers that only include information of objective relevance (http://www.weareready.ch).
Unless the case where firms are risk averse and there is unequal variance in productivity across groups.
Farber and Gibbons (1996) only use those parts of the AFQ test that resemble very much the PISA test scores used in this article.
Siegenthaler (2011) for example analyzed the informational value for the case of the privately sold aptitude test “multicheck retail sale” and found that the test results do not improve firms’ ability to predict apprenticeship success once easy-to-obtain information provided in application dossiers (former school grades and the level of compulsory schooling) is taken into account.
The intellectual aspiration level of apprenticeship training is also relevant for the second transition, the one from the apprenticeship training into the labor market. Bertschy et al. (2009) have shown that this level affects the chances of seamless transition in a significant and causal way.
Although we define an immediate transition from school to work as a successful transition, this does not mean that all of the unsuccessful applicants would have been better off if they had succeeded immediately in finding an apprenticeship. For some of the unsuccessful candidates the delay of transition might even improve the match between their ability, expectations and the demands of their future employers.
We have to concede the fact that employers have easy-to-obtain information on the candidates like, e.g., health problems that might be correlated with the PISA results and that are not observable by the researchers. In this case, the additional information available to employers helps them to estimate the true ability more accurately then our own regressions would suggest. This carries the potential risk that the researcher would wrongfully qualify a candidate as an over- or underachiever, whereas the employer had accurately assessed the true ability based on his/her easy-to-obtain information. To minimize the risk of an unjustified classification of the individual, we use a large threshold for the creation of the dummy variables of 73 PISA points (=one proficiency level) around the predicted score. As it is unlikely that an easy-to-obtain information not available to the researchers would influence the predication to such an extend (keeping in mind that we already control for observables that are likely to be highly correlated with information that we might miss), we assume that this potential bias is negligible. As far as this assumption is testable, we do not find evidence for violations (\(p=0.784\) in a RESET test).
The Swiss national sample of ninth graders was added to the PISA study for comparisons between the country’s different language regions, as the international PISA survey only covers pupils at age 15, independent of the grade in which they are enrolled. Because many 15 year olds are already in the 9th grade, the two populations overlap but the national sample of ninth graders is better suited for the purpose of our analyses.
PISA measures competencies in points with a mean of 500 points for all participating countries and a standard deviation of 100 points. PISA reading literacy is measured by a composite test score that summarizes the results from three reading literacy scales. The “retrieving information” scale reports on students’ ability to locate information in a text. The “interpreting texts” scale report on the ability to construct meaning and draw inferences from written information. A “reflection and evaluation” scale reports on students’ ability to relate text to their knowledge, ideas, and experiences. In addition, experts have divided the scale into six different proficiency levels (very low, low, medium low, medium high, high, very high). For our analysis, we define under- and overachievers to deviate by more than one proficiency level from what one would predict. To give an impression on the difference between two adjacent proficiency levels: students proficient at level 3 (medium low) are capable of reading tasks of moderate complexity, such as locating multiple pieces of information, making links between different parts of a text, and relating it to familiar everyday knowledge. Students proficient at level 4 (medium high) are capable of difficult reading tasks, such as locating embedded information, construing meaning from nuances of language and critically evaluating a text (OECD 2001).
A replication of our estimations for half of the students that have been tested also for mathematical literacy shows qualitatively similar results to the ones presented in this article using reading literacy. Results are available from the authors upon request.
Although we have a very rich set of background variables on the students, it is still possible that the employers can collect additional information that is not observable by the researchers. If this information would be correlated with the deviation from the predicted PISA scores, an omitted variable bias would occur. However, to the extent that these unobservables are correlated with the observable characteristics the bias is minimized (see also Footnote 11).
We have imputed missing information (1.4 %) about the aspiration level of less common tracks by regressing on training duration (years), amount of vocational schooling lectures (hours per year), and an interaction term between the two. These factors strongly explain the aspiration level (\(R^{2}\) of 84 %).
PISA test-scores have not been used by the experts to assess the aspiration level.
We have also estimated models that allow for the possibility that the effect of (positive/negative) residual PISA scores is different in magnitude depending on whether the difference is within the magnitude of one PISA competence level (small) or outside one competence level (the definition for under-/overachievers). Results show that small deviations are neither significantly different from large deviations nor significantly different from zero and thus more spurious.
Estimations excluding non-cognitive variables are available upon request from the authors.
The residual in model 4 has about half of the effect size of the predicted PISA score. In the previous regressions, this relation was 1:8.5 (no significant effect of the overall residual) and 1:3.5.
References
Aigner DJ, Cain GG (1977) Statistical theories of discrimination in labor markets. Ind Labor Relat Rev 30(2):175–187
Altonji JG, Pierret CR (1997) Employer learning and the signaling value of education. In: Ohashi I, Tachibanak T (eds) Industrial relations, incentives and employment. Macmillan Press Ltd, London, pp 159–195
Altonji JG, Pierret CR (2001) Employer learning and statistical discrimination. Q J Econ 116:313–350
Arcidiacono P, Bayer P, Hizmo A (2010) Beyond signaling and human capital: education and the revelation of ability. Am Econ J Appl Econ 2(4):76–104
Arrow KJ (1973) The theory of discrimination. In: Ashenfelter O, Rees A (eds) Labour markets. Princeton University Press, Princeton, NJ
Bauer TK, Haisken-DeNew JP (2001) Employer learning and the returns to schooling. Labour Econ 8(2):161–180
Bertschy K, Cattaneo MA, Wolter SC (2009) PISA and the transition into the labour market. LABOUR 23(1):111–137
Farber HS, Gibbons R (1996) Learning and wage dynamics. Q J Econ 111(4):1007–1047
Galindo-Rueda F (2003) Employer learning and schooling-related statistical discrimination in Britain. IZA discussion paper no. 778, Institute for the Study of Labor, Bonn
Haeberlin, U, Imdorf C, Kronig W (2004) Chancenungleichheit bei der Lehrstellensuche: Der Einfluss von Schule, Herkunft und Geschlecht. NFP43 Synthesis 7. Bern/Aarau: NFP43/SKBF
Hupka S, Sacchi S, Stalder BE (2006) Herkunft oder Leistung? Analyse des Eintritts in eine zertifizierende nachobligatorische Ausbildung anhand der Daten des Jugendlaengsschnitts TREE. Working paper. Basel: TREE
Imdorf C (2006) The selection of trainees in small and medium-sized enterprises. Integration and exclusion of immigrant youth at the transitional stage between school and vocational training in Switzerland. Working paper (english version). http://www.coreched.ch/publikationen/ISA_Durban06_Imdorf.pdf
Imdorf C (2009) Die betriebliche Verwertung von Schulzeugnissen bei der Ausbildungsstellenvergabe. Empir Paedagog 23(4):392–409
Lange F (2007) The speed of employer learning. J Labor Econ 25(1):1–35
Moser U (2004) Jugendliche zwischen Schule und Berufsbildung. Eine Evaluation bei Schweizer Grossunternehmen unter Berücksichtigung des internationalen Schulleistungsvergleich PISA. h.e.p Verlag, Bern
Muehlemann S, Schweri J, Winkelmann R, Wolter SC (2007) An empirical analysis of the decision to train apprentices. LABOUR Rev Labour Econ Ind Relat 21(3):419–441
Neuenschwander MP, Malti T (2009) Selektionsprozesse beim Uebergang in die Sekundarstufe I und II. Z Erzieh 12(2):216–232
OECD (2001) Knowledge and skills for life. First results from the OECD programme for international student assessment (PISA) 2000. OECD, Paris
OECD (2002) PISA 2000 technical report. OECD, Paris
Phelps ES (1972) The statistical theory of racism and sexism. Am Econ Rev 62(4):659–661
Siegenthaler M (2011) Can a standardised aptitude test predict the training success of apprentices? Evidence from a case study in Switzerland. Emp Res Vocat Educ Train 3(2):105–128
Spence M (1973) Job market signalling. Q J Econ 87(3):355–374
Stalder BE (2011) Das intellektuelle Anforderungsniveau beruflicher Grundbildungen in der Schweiz. Ratings der Jahre 1999–2005. Institut fuer Soziologie der Universitaet Basel/TREE, Basel
Stalder BE, Schmied E (2006) Lehrvertragsaufloesungen, ihre Ursachen und Konsequenzen. Ergebnisse aus dem Projekt LEVA. Bildungsplanung und Evaluation der Erziehungsdirektion des Kantons Bern, Bern
Topel RH, Ward MP (1992) Job mobility and the careers of young men. Q J Econ 107(2):439–479
Wolter SC, Ryan P (2011) Apprenticeship. In: Hanushek EA, Machin S, Woessmann L (eds) Handbook of the economics of education, vol 3. Elsevier, Amsterdam, pp 521–576
Acknowledgments
The authors would like to thank the Transitions from Education to Employment Survey (TREE) consortium for the kind permission to use the TREE data. The authors also thank participants of the economics of education area conference of the CESifo network and especially Joop Hartog, the discussant, as well as the editor and two anonymous referees for helpful comments and advice.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Mueller, B., Wolter, S.C. The role of hard-to-obtain information on ability for the school-to-work transition. Empir Econ 46, 1447–1471 (2014). https://doi.org/10.1007/s00181-013-0709-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00181-013-0709-2