http://orcid.org/0000-0002-5957-9833
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
We estimate the percentage of eligible students who do not take up their federal need-based student financial aid entitlements in a microsimulation model for the German Socio-Economic Panel Study 2002–2013. We find that about 40% of the eligible low-income students do not take up their entitlements. Non-take-up is inversely and rather inelastically related to the level of benefits. We investigate possible reasons for non-take-up, accounting for potential sample selection and endogeneity issues; and discuss policy implications.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Stefanie P. Herber http://orcid.org/0000-0002-5957-9833
Notes
1 The numbers are at present values, calculated at an interest rate of 2%, see Grave and Sinning (Citation2014).
2 Socio-Economic Panel (SOEP), data for years 1984-2013, version 30, SOEP, 2015, doi:10.5684/soep.v30.
3 Static microsimulation models use the current legal framework as a basis to study a specific population at a specific point in time (Bungers Citation1981; Fredriksen Citation1998). In contrast to this, dynamic mircosimulation models explicitly account for behavioural change over time or due to a reform. Two applications of microsimulation models in educational contexts are Steiner and Wrohlich (Citation2012) and O'Donoghue (Citation1999).
4 An extensive review of the literature is beyond the scope of this paper. Currie (Citation2004), Hernanz, Malherbet, and Pellizzari (Citation2004) and Finn and Goodship (Citation2014) provide comprehensive reviews.
5 It should be noted, however, that in contrast to BAföG most loans are supplementary and not means-tested.
6 Cadena and Keys (Citation2013) exploit that eligible US students who have to pay for room and board and live off-campus can receive a part of the interest-free Stafford loan paid in cash rather than as a credit to their university account. The authors show that if students regard different assets as non-fungible and lack self-control to limit their expenses to prevent overspending, non-take-up can be a rational reaction to avoid overspending.
7 The maximum subsidy cited here is based on the maximum monthly benefits of EUR 670, a repayment of EUR 105 a month, starting after the grace period and given an interest rate of 2%. The upfront repayment implies another implicit subsidy of the interest rate, though upfront payment is not profitable if BAföG amounts are high (Grave and Sinning Citation2014, 113).
8 Section A.1 in the appendix explains the BAföG calculation.
9 Socio-Economic Panel (SOEP), data for years 1984-2013, version 30, SOEP, 2015, doi:10.5684/soep.v30.
10 The SOEP therefore offers clear advantages over other data sets such as the EU-SILC data, which neither contain the relevant information to calculate income taxes and BAföG amounts nor allow an analysis of the determinants of non-take up.
11 Section C.3 in the Appendix provides more information and descriptive statistics.
12 The SOEP provides readily imputed income measures so that we do not lose cases due to item non-response.
13 We still exclude about 8.54% of all SOEP students for whom we do not know whether they might qualify for dependent funding because they were drawn as a new SOEP household after moving out of their parents' home. Section C.3 in the Appendix provides more information and descriptive statistics.
14 Section B.1.2 in the Appendix contains a detailed presentation and discussion of this approach.
15 Section B contains a detailed description of our approach.
16 An earlier version of this paper (Herber and Kalinowski Citation2016) contains a thorough discussion of our sample selection and variable construction procedure.
17 As our sensitivity check in Appendix C.1 shows, relaxing our restrictive assumptions substantially decreases the beta error rate. As these manual modifications do not affect the regression results, we only present the conservative results without any manual corrections. Corrected results are available upon request.
18 Although we do not find evidence for a time-trend or statistically significant differences up until 2006, we include separate year-dummies in all our regressions shown later.
19 See Bargain, Immervoll, and Viitamäki (Citation2012) for related findings with respect to social assistance.
20 In addition to the models presented in the following, we also ran various panel data models. Although the results were mostly identical, we decided in favour of cross-sectional analyses because of the small sample size, the fact that we observe students only twice on average, and the resultant low within and between variations.
21 Previous to our study, Kayser and Frick (Citation2001) and Frick and Groh-Samberg (Citation2007) used a Heckman-type approach to correct for sample selection into non-take-up of social assistance. Wilde and Kubis (Citation2005) address the issue of sample selection in a simultaneous equation model.
22 Information on the German education system and the differences between general and vocational training can be found in Secretariat of the Standing Conference of the Ministers of Education and Cultural Affairs of the Länder in the Federal Republic of Germany (Citation2017). The wage gap between having completed tertiary education and having completed vocational training in Germany is highly subject-specific, and some vocational programmes yield higher returns than university studies (Glocker and Storck Citation2014). Moreover, a large strand of the literature doubts that high school graduates are well informed about the returns on different educational options (e.g. Usher Citation2005). Systematic differences arising from an anticipated wage gap should therefore not affect our results.
23 Heckprobit models can generally be estimated without an exclusion restriction (Wooldridge Citation2010, 814). Doing so does not change our results.
24 An earlier version of this paper contains a detailed discussion of our hypotheses and expected mechanisms of non-take-up, see [author(s)].
25 See Federal Institute for Research on Building, Urban Affairs and Spatial Development (Citation2013, 3).
26 We are able to separate the level of benefits and the parents' monthly labour income because the BAföG calculation uses a special, non-deflated income measure. Owing to extensive means-testing and imposition of complex allowances and exemptions, labour income and BAföG benefits are non-linearly related. We report further robustness checks on parents' transfers in Section 5.3.
27 In more than 98% of these cases, both parents were living together either in East or West Germany.
28 The questions are worded as follows: ‘Do you generally think things over for a long time before acting—in other words, are you not impulsive at all? Or do you generally act without thinking things over for long time–in other words, are you very impulsive?’ and ‘Are you generally an impatient person, or someone who always shows great patience?’ See Vischer et al. (Citation2013) for a validation of the patience scale.
29 ‘Are you generally willing to take risks, or do you try to avoid risks?’ See Dohmen et al. (Citation2011) for a validation of this scale.
30 The gap is also robust to introducing an interaction between East German background and parents' incomes to our model, although this results in a high degree of multicollinearity. Moreover, we find no evidence that this East/West difference fades away over time (see Figure E3 in the Appendix).
31 Cadena (Citation2008) and Keys (Citation2009) show theoretically and empirically that, if a sophisticated student is sufficiently impatient and her discount function is quasi-hyperbolic at the same time, he or she will reject an interest-free loan offer in order to limit overspending during the study period.
32 The BAföG calculation proceeds on the assumption that parents need at least half of their income to support themselves. Therefore, if parents reduce their labour income by EUR 100, the maximum increase in BAföG equals EUR 50. Only 9% of the parents in our data are self-employed, i.e. may have a certain flexibility to allocate their income over time.
33 To the best of our knowledge, however, there is no adequate straightforward test for weak instruments when the dependent variable is binary, standard errors are clustered and data are weighted.
34 We report the full results in Table D2 and predicted probabilities in Table D3 in the Appendix.
35 See Hernanz, Malherbet, and Pellizzari (Citation2004) and Becker and Hauser (Citation2003, 149ff) for extensive overviews of (dis)advantages of various data sources for the analysis of non-take-up.