Transitions from Welfare to Employment: Does the Ratio between Labor Income and Social Assistance Matter?

Schneider, Hilmar; Uhlendorff, Arne Article Transitions from Welfare to Employment: Does the Ratio between Labor Income and Social Assistance Matter? Schmollers Jahrbuch – Zeitschrift für Wirtschafts- und Sozialwissenschaften. Journal of Applied Social Science Studies Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Schneider, Hilmar; Uhlendorff, Arne (2005) : Transitions from Welfare to Employment: Does the Ratio between Labor Income and Social Assistance Matter?, Schmollers Jahrbuch – Zeitschrift für Wirtschafts- und Sozialwissenschaften. Journal of Applied Social Science Studies, ISSN 1865-5742, Duncker & Humblot, Berlin, Vol. 125, Iss. 1, pp. 51-61, https://doi.org/10.3790/schm.125.1.51 This Version is available at: https://hdl.handle.net/10419/292108 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Schmollers Jahrbuch 125 (2005), 51-61 Duncker & Humblot, Berlin Transitions from Welfare to Employment: Does the Ratio between Labor Income and Social Assistance Matter? By Hilmar Schneider and Arne Uhlendorff* Abstract lt is often argued that the high level of welfare claims in Germany creates little incentive for workers with low productivity to seek for ajob. We examine the influence of the ratio between estimated potential labor income and the welfare payment level on the probability of leaving social welfare. We estimate a discrete time hazard rate model with unobserved heterogeneity using representative micro data from the German Socio Economic Panel Study (SOEP). Our results show that the ratio has a positive effect on the probability of leaving social welfare. This effect is especially relevant for house holds with a potential labor income higher than their welfare payment level. JEL Classifications: 138, 164, C41 1. Introduction In 2002, about 2.8 million people in Germany received social assistance.1 The number of recipients and the amount of income support expenditures have been rising almost continuously in the past. Why does Germany have so many welfare recipients? In the economic literature as well as in the public debate on the German welfare system, the incentive argument plays an impor tant role. lt asserts that if the difference between the amount one receives from transfers and the potential income from a regular job is too small, then it is not attractive for an individual to take the job. In this paper, we analyze this hy- * We would like to thank Peter Haan, Peter Jensen, Hans-Joachim Rudolph, Katha rina Wrohlich, the participants in the SOEP Brown Bag Seminar, the EALE Conference 2004, the IZA Summer School 2004, the IZA Brown Bag Seminar, the SOEP user conference 2004 and an anonymous referee for helpful comments and the German Science Foundation (DFG) for financial support through SPP1169 "Potential for more Flexibility on Heterogenous Labor Markets". 1 This number of welfare recipients refers to permanent transfers, the so-called Hilfe zum Lebensunterhalt. Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 52 Hilmar Schneider and Arne Uhlendorff pothesis by estimating the impact of the ratio between potential labor income and the amount of transfer payments on the transition probability from welfare to employment in Germany. Previous research indicates no influence of income variables on the dura tion on welfare in Germany. A study by Riphahn (1999) based on the German Socio-Economic Panel Study (SOEP) shows no significant influence of apre dicted real net income variable for individuals employed full-time on the exit probability out of income support. Wilde (2003) uses the low-income panel to examine how the difference between social benefits and the average income affects the probability of leaving social welfare for unskilled employees, and finds no significant effects. For our analysis, we use data from the SOEP. The data includes monthly information about the duration of social welfare of households between 1991 and 1999. We estimate a discrete-time proportional hazard rate model with competing risks and risk-specific unobserved heterogeneity. Controlling for several typical covariates, the ratio between potential labor income and the welfare level shows the expected positive effect on the prob ability of leaving social welfare for work. This effect is especially relevant for households with a potential labor income higher than their social welfare le vel. Our results are contrary to previous studies described above. However, although the incentive hypothesis cannot be rejected, this study does not un ambiguously answer whether the higher transition probability is a conse quence of higher incentives or of a higher job offer arrival rate for better educated persons. The positive effect of the ratio could be an indication of either. 2. Social Assistance in Germany The German social welfare system (Sozialhilfe) is a means-tested transfer program. To qualify for welfare, household income may not exceed a certain minimum level. The amount of social assistance is related to a basic minimum income concept depending on household size, household composition, and some adjustments to specific needs, e.g., for disabled persons. Except for a small allowance, additional income is deducted from social assistance.2 In principle, there is no time limit for welfare receipt. As a consequence, welfare payments may be regarded as a permanent alternative to labor income. Following the static labor supply theory, participation in welfare is more likely the higher the amount of benefits (e.g. Moffitt 2002). A stylised depic- 2 For a detailed description of the social assistance in Germany see e.g. Schneider/ Uhlendorff (2004). Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 Transitions from Welfare to Employment 53 tion of a utility-maximizing individual subject to a non-convex budget set is given in Figure 1. Only when the number of hours worked exceeds Q does disposable income exceed the social assistance level V and increase with net market wage rate w. Line B depicts disposable net income depending on the number of hours worked. If no social assistance existed, it would be optimal to work H hours per week. When a welfare program exists, not working may generate a higher utility level than working. I0 symbolizes an indifference curve that connects all combinations of disposable income and working hours that are equivalent to not working at all. I1, in comparison, stands for an indif ference curve related to working H hours and achieving a disposable income of wH. In the case depicted, a utility-maximizing person would only work H hours per week if she could achieve a disposable income of at least Y*, which can also be expressed in terms of an implicit minimum wage rate. Y" ...... ....... ... .. ... .. .. .·· . . : 1 •• : :-· ....... . ���·••-' ... Q H weeklyworking hours Figure 1: The Impact of Social Assistance on Work lncentives for a Stylised Budget Set In a dynamic job-search model, an individual does not stand to receive a particular wage but rather a particular distribution of wages ( e.g. Devine / Kie fer 1991). To leave a welfare program for employment requires an acceptable job offer. Wage offers are only accepted if they exceed the reservation wage. This reservation wage depends positively on the amount of social benefits. Given a certain frequency of wage offers and a certain level of welfare pay ments, one is more likely to observe exits from social assistance the higher an individual's market (or expected) wage is. We assume that the effect of the difference between the two income sources depends on the relative level of the social benefits. Therefore, households with a lower ratio between potential labor income and the amount of social welfare should have a lower hazard rate from welfare to employment.3 Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 54 Hilmar Schneider and Arne Uhlendorff Moreover, it should be taken into account that the relationship between re servation wages and actual wages is not a deterrninistic one. In practice, one may even observe actual wages falling below the replacement wage, i.e., be low the level of social assistance. In terms of our model, this means that indi viduals attribute an unmeasured high non-monetary value to work. Working for less than the replacement wage could, for exarnple, be explained as an attempt to avoid a stigma effect. Thus, unmeasured utility is another reason for the probabilistic nature of the model. If employment is viewed as a value in itself, it is reasonable to expect a non-linear relationship between the ratio of the two income sources and the hazard rate from welfare to employment. We will control for potential nonlinearities by considering three different le vels of the ratio between the potential market wage and the replacement wage: ratios lower than and equal to 1 (potential market wage below replacement wage), ratios between 1 and 1.5 (potential market wage around replacement wage), and ratios above 1.5 (potential market wage clearly above replacement wage). An alternative explanation for a lower exit probability of households with a lower ratio could be a lower arrival rate of job offers for low-skilled workers. This could be explained by a lack of job offers for low-skilled individuals. However, different low-skilled workers may have different reservation wage levels according to their household-related welfare claims. Controlling for skill level may therefore allow for a discrimination between demand effects and incentive effects. 3. Data, Variables and Methods We use SOEP waves 1992 to 2000, which provide monthly information about social welfare receipt between 1991 to 1999 (for details on the SOEP see Haisken-DeNew /Frick 2003). Excluding households in which the house hold head and, if applicable, partner are aged 61 years or older, we observe 579 uncensored or right-censored social welfare spells between January 1991 and December 1999 distributed over 455 households. These spell data are combined with several time-variant and time-invariant household and indivi dual characteristics.4 In the data, there are 386 uncensored and 193 right-censored observations. We are interested in the transition from social welfare to a situation with labor 3 We assume the income ratio to be exogenous. However, one could argue that demo graphic behavior with respect to fertility and marriage or schooling are affected by the welfare system and therefore the income ratio could be endogenous. For example, Keane and Wolpin (2002) take into account the impact of welfare benefits on the eco nornic and demographic behavior ofwomen in the U.S. 4 For detailed descriptive statistics see Schneider /Uhlendorff (2004). Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 Transitions from Welfare to Employment 55 income. Therefore, we differentiate between transitions to employment (199 cases) and alternative transitions (187 cases). A transition to employment is defined as a situation with at least one adult household member (head of the household or partner) working füll-time, both working part-time, or one per son working at least part-time in the case of single households subsequent to benefit receipt. In the following, we describe the procedures used to estimate and calculate the two income sources. 3.1 Potential Net lncome and Social Assistance Our first step is to estimate a gross market wage equation with a pooled sample using the SOEP waves 1991-1999. We apply a sample selection mod el consisting of a log-linear wage equation and an equation describing the binary choice to work or not to work, also referred to as the type II Tobit model (e.g. Wooldrige 2002).5 With this approach we control for potential endogeneity of the participation decision. Based on these estimation results, we calculate a potential monthly füll-time gross wage for each head of house hold and his / her partner. We calculate the potential net income, assuming that in the case of a partner household, the person with the higher potential income would be working, and we account for income taxes, social security contribu tions, and child and housing allowance. We observe the receipt of social assistance as a binary variable but not the corresponding amount of social assistance for each month. Therefore we cal culate the maximum amount of social assistance. This is the income (includ ing other transfers) that a household staying on welfare would receive perma nently. The empirical distribution of the ratio between the two income sources in the first month of each spell is plotted in Figure 2. The median is 1.26, i.e., for about half of the sample, expected income does not exceed their welfare bene fits by more than 25 % . 3.2 Model Specification The process of leaving welfare in favor of labor income can appropriately be modeled by a transition rate approach. According to the type of data used here, a discrete hazard rate model has to be applied (see for example Han and Hausman 1990). The duration of welfare receipt is generated by a continuous time process, but observed in months. Two potential destination states are s We estimate separate models for East and West Germany and for men and women. The results are reported in Schneider/ Uhlendorff (2004 ). Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 56 Hilmar Schneider and Arne Uhlendorff 2 ratio 4 Figure 2: Ratio between Potential Net Income and the Amount of Social Assistance, First Month of Each Welfare Spell (n = 579) considered reflecting transitions to employment and alternative translt10ns such as those into other transfer programs. We assume proportional transition rates with covariates causing proportional shifts of a so-called baseline transi tion rate and interval constant covariates. Following from this, the probability f of a transition to state r at a given interval j is given by the difference between two survivor functions6 multi plied by the share of the risk-specific transition rate at interval j related to the hazard rate at interval j. 7 (1) fr(j) = /xp(xrßr + rrj + T/r) [SU _ 1) _ SU)] . Z: exp(xq;ßq + 1q; + T/q) q=l The 'Y parameters capture the duration dependence of the baseline transition function. They may be interpreted as an interval-specific mean of the baseline transition rate, which is equivalent to a an interval-specific constant baseline transition rate. We assume the risk-specific unobserved heterogeneity compo nents ry1 and ry2 following a bivariate normal distribution. The likelihood con tribution of a spell corresponds to (2) L(ß,,,T/i,T/2) = exp(xtjßt \itj +TJ1t' exp(x2ß2 +12j +TJ2t' [cS(jl)-(2c- l)S(j)]8 � exp(xq;ßq + iqj + T/q) q=l 6 The survivor function S(j) describes the probability that a spell lasts at leastj inter vals. 7 Fora detailed discussion of the econometric model see Schneider/ Uhlendorff (2004 ). Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 Transitions from Welfare to Employment 57 whereby c1 = 1 and c2 = 1 indicate a transition to risk 1 and risk 2 in interval j, respectively, and c corresponds to the maximum of c1 and c2. In the follow ing, we will refer to this as a random effects piecewise exponential model. 4. Empirical Results We estimate the hazard rate model for competing risks with and without unobserved heterogeneity. The coefficients of the model can be interpreted with respect to the underlying continuous time proportional hazard rate. The results are reported in Table 1. The inclusion of unobserved heterogeneity does not significantly improve the model fit.9 We created three variables representing the effect of the ratio between esti mated potential labor income and welfare payment level: the first for ratios lower than and equal to 1, the second for ratios from 1 to 1.5, and the third for ratios above 1.5. 10 In models both with and without unobserved heterogeneity, the coefficients of the latter two variables are significantly positive, while the coefficient of the first is positive but not significantly different from zero. An increase in the ratio seems to be more relevant if the potential labor income exceeds the social assistance level. The coefficient of the income ratio for a ratio between 1 and 1.5 indicates that an 0.1 higher ratio goes along with a 10% higher probability of an exit to employment, while a 0.1 higher ratio for ratios above 1.5 leads to 7% higher probability of a transition to employment. Assurning households with the same welfare level, a difference in the income ratio of 0.1 stands for a difference in estimated labor income by 10% of the social welfare level. For alternative transitions, these income variables have no significant influence. Our results confirm the theoretical predictions: given a certain level of so cial welfare payments, it is more likely to observe exits from social assis tance to employment the higher an individual's (net) market wage. This is especially the case for households with an expected labor income higher than the social assistance level. Only if the household is able to improve its income through employment does the difference between the two income sources matter. s The corresponding likelihood is solved by applying Gauss-Herrnite quadrature. 9 Interpretation of the t-values of the variance parameters is not useful since ln(variance) being O means that the variance itself is equalling 1. Testing for the var iance being O would mean to test ln(variance) equalling minus infinity, which is not feasible. 10 The ratio takes on the value 1 if the potential labor income equals the welfare payment level and the value 1.5 if the potential labor income exceeds the social welfare payments by 50%. Schmollers Jahrbuch 125 (2005) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39 f f Variable 2 years 3 years 4 years 5 and more years Year of observation 1992 1993 1994 1995 1996 1997 1998 1999 December dummy January dummy East Germany At least one adult household member with vocational training Table 1 Discrete-Time Proportional Hazard Rate Models Piecewise exponential model Random effects piecewise exponential model Transitions Alternative Transitions Transitions Alternative Transitions to Employment to Employment Coefficient t-value Coef. t-value Coefficient t-value Coef. t-value -0.41 * -1.89 -0.58*** -2.60 -0.36 -1.26 -0.33 -0.99 -0.33 -1.01 -0.68** -2.04 -0.25 -0.58 -0.30 -0.58 -0.85 -1.64 -0.82* -1.76 -0.76 -1.20 -0.35 -0.52 -1.34* -1.85 -1.06* -1.76 -1.24 -1.48 -0.52 -0.64 -0.51 -0.96 0.18 0.34 -0.52 -0.97 0.13 0.22 -0.29 -0.62 0.15 0.30 -0.30 -0.61 0.09 0.17 -0.18 -0.39 0.23 0.47 -0.17 -0.36 0.24 0.44 -0.48 -1.03 -0.06 -0.13 -0.47 -0.99 -0.07 -0.12 -0.72 -1.48 0.21 0.41 -0.71 -1.45 0.18 0.33 -0.56 -1.19 0.01 0.01 -0.56 -1.18 -0.06 -0.12 -0.15 -0.32 0.02 0.05 -0.13 -0.29 0.02 0.03 -1.49*** -2.90 -0.98* -1.80 -1.50*** -2.89 -1.06* -1.80 2.32*** 16.06 3.03*** 19.34 2.34*** 14.97 3.12*** 15.84 -1.44** -2.01 -0.55 -0.93 -1.44** -2.01 -0.50 -0.84 0.43* 1.94 0.36 1.52 0.45* 1.91 0.46 1.62 0.18 0.96 -0.17 -0.93 0.18 0.93 -0.18 -0.83 U\ 00 g:; s e; � 8. 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.125.1.51 | Generated on 2023-04-04 12:34:39