Take It or Leave It: (Non-)Take-Up Behavior of Social Assistance in Germany

Kayser, Hilke; Frick, Joachim R. Article Take It or Leave It: (Non-)Take-Up Behavior of Social Assistance in Germany Schmollers Jahrbuch – Zeitschrift für Wirtschaftsund Sozialwissenschaften. Journal of Applied Social Science Studies Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Kayser, Hilke; Frick, Joachim R. (2001) : Take It or Leave It: (Non-)Take-Up Behavior of Social Assistance in Germany, Schmollers Jahrbuch – Zeitschrift für Wirtschaftsund Sozialwissenschaften. Journal of Applied Social Science Studies, ISSN 1865-5742, Duncker & Humblot, Berlin, Vol. 121, Iss. 1, pp. 27-58, https://doi.org/10.3790/schm.121.1.27 This Version is available at: https://hdl.handle.net/10419/291975 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 121 (2001), 27-58 Duncker & Humblot, Berlin Take It or Leave It: (Non-)Take-Up Behavior of Social Assistance in Germany By Hilke Kayser and Joachim R. Frick* Abstract Analyzing the under-consumption of benefits in the German means-tested Social Assistance program using data from the German Socio-Economic Panel Study we confirm recent high estimates of a non-take-up rate of more than 60 percent. In light of likely measurement errors in income and in our simulation of household needs, we provide a range of estimates yielding useful boundaries for the non-take-up rate. We show that the rate varies greatly depending on the determination of eligibility. Simulation results pertaining to the amount of unclaimed benefits are qualitatively similar to those for the non-take-up rate. In our multivariate analyses on determinants of (non-)take-up behavior we find distinct differences across population groups and significant impacts of proxies for stigma, application costs and social ties. Zusammenfassung Basierend auf den Mikro-Daten des Sozio-oekonomischen Panels untersuchen wir Ausmaß und Struktur der Nichtinanspruchnahme von Sozialhilfe (Hilfe zum Lebensunterhalt) in Deutschland. Die Schätzung einer Nichtinanspruchnahmequote ("Dunkelziffer") fällt mit über 60 Prozent erwartungsgemäß höher aus als für die Höhe der nicht in Anspruch genommenen Sozialhilfebeträge. Angesichts der hohen Wahrscheinlichkeit von Messfehlern sowohl in den faktisch erhobenen Einkommensdaten als auch bei der Simulation des Bedarfseinkommens, bieten wir eine Reihe von Schätzungswerten und somit nützliche Sensitivitätsanalysen für die Nichtinanspruchnahmequote an. Es wird deutlich, wie stark die Berechnung der Inanspruchnahmequote mit der Bestimmung der Inanspruchnahmeberechtigung variiert. Multivariate Analysen offenbaren deutliche Unterschiede für verschiedene Bevölkerungsgruppen und zudem signifikante Einflüße von Indikatoren zur Messung von Stigma, Kosten der Antragstellung und insbesondere sozialen Bindungen. JEL-Classification: 138, H 31, H 53, D 31 * The authors would like to thank Edward J. Bird and Gert G. Wagner for most valuable cooperation and two anonymous referees for very helpful comments on an earlier draft. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 28 Hilke Kayser and Joachim R. Frick 1. Introduction Industrialized countries share a common principle of providing a basic safety net within the social security system to protect the poorest from falling below a certain level of economic well-being. This safety net might consist of a multitude of single programs targeted at different life events or situations (like in the USA) or a single system of social assistance (like in the Federal Republic of Germany). Notwithstanding differences, a closer look at claiming behavior across countries reveals that non-take-up of such social benefits is high, particularly for means-tested social benefits. The predominant means-tested German welfare assistance program (the so-called Sozialhilfe) is no exception to the rule (Neumann and Hertz, 1998; Bird et al., 2000; Riphahn, 2000). It is the aim of this paper to provide a better understanding of the factors that contribute to the high rate of non-take-up of welfare benefits in Germany. There are various reasons why the under-consumption of benefits must be considered a challenge that is equal in importance to the abuse of benefits. First, non-participation suggests a failure in the policy. Though at any one point in time full take-up is unrealistic because of lags and delays in claiming, the target of the policy is the full use of its benefits in order to fulfill the mandate of providing a safety net for those in need. Secondly, non-take-up by some eligible households implies a fundamental injustice when comparing non-claiming households to households in a similar economic situation who receive social benefits. Thirdly, the rejection of benefits implies that the costs of claiming exceed the anticipated benefits from the entitlement amount. To the extent that such costs result from complex schemes, poor transmission of information or similar factors they imply a failure in the design or implementation of the program. Non-take-up becomes a serious social problem if some households cannot reach the targeted income because they are - directly or indirectly - discouraged or prevented from claiming because of objective or subjective barriers.1 In this study we take another look at the non-take-up rate of the German Social Assistance program using micro-data at the household level from the Socio-Economic Panel (SOEP). Simulating the German welfare system we determine the proportion of households that does not receive benefits out of the entire population of households that are deemed eligible to receive assistance. We also extend the existing literature about the non-take-up of benefits in the German social assistance program by employing multiple regressions based on an economic participation model to estimate the deter1 For a more detailed discussion of the social and economic implications of a low take-up of social benefits see van Oorschot (1998), Hartmann (1985), and Moffitt (1983). Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 29 minants of the high non-participation rate. Given the nature of the relatively small size of representative survey samples, we provide a wide range of estimates in order to show the sensitivity of our empirical results. Understanding the magnitude of the entitlements that households are forgoing when not claiming benefits is of fiscal importance. In essence, these forgone benefits represent savings to the relevant government budget and as such policies aiming to increase participation in a particular program will increase government spending. Because entitlement levels differ between those who claim and those who do not claim, we provide a range of estimates of the fiscal savings rate from non-take-up based on our simulations. Our simulation results suggest that the estimated non-take-up rate for social assistance in Germany is rather sensitive to changes in our simulated measure of eligibility, ranging from 41.3 percent to 82 percent depending on the stringency of our eligibility criterion. With respect to unclaimed entitlements our simulations suggest a rate of unclaimed benefits out of all outstanding entitlements of 45.3 percent, a rate that can vary from around 30 percent to around 67 percent depending on how stringent an eligibility criterion we use. Furthermore, household characteristics can be associated with significantly different welfare participation rates. While the base case scenario of our multivariate analysis suggests an estimated non-take-up rate of 62.9 percent, it can be as high as 72.2 percent for households without children and as low as 19.8 percent for households with more than two children. We also find that social controls can affect participation: households that are generally pessimistic or not actively involved in religious institutions have a higher participation rate. The paper is organized as follows: After a review of the literature in section 2, we briefly describe our model of social assistance participation decision (section 3) and the German Social Assistance System (section 4). Section 5 provides information on data and methodological aspects; sections 6 and 7 show descriptive and multivariate results of our empirical work. Section 8 concludes. 2. Research Question and Literature Knowledge about non-take-up of social benefits in Germany is rather limited. Existing studies show differences in benefit receipt and in the proportion of eligible, non-claiming individuals or families out of the relevant population across various population groups.2 Somewhat less detailed is 2 Data presented by the Federal Bureau of Statistics, for example, show that the receipt of social assistance was noticeably higher in 1997 among single parents, chilSchmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 30 Hilke Kayser and Joachim R. Frick our knowledge of the non-take-up rate for various population groups. The non-take-up rate for social benefits refers to the proportion of eligible households who do not claim the benefits for which they are eligible. Empirical estimates of the non-take-up rate of social assistance in Germany per se go back to the late seventies and vary from 36 percent to 79 percent depending on the method and data used (Geissler, 1976; Bujard and Lange, 1978; Klanberg, 1979; Hauser et al., 1981; Hartmann, 1985; Hauser and Hubinger, 1993; Neumann and Hertz, 1998; Bird et al., 2000; Hauser et al., 2000; Riphahn, 2000). Of the more recent studies, Neumann and Hertz (1998) calculate a non-take-up rate for individuals of about 52 percent for 1995. They estimate the number of eligible households on the basis of SOEP data and compare that number to the number of welfare recipients based on official statistics, a method that does not obviously lead to a consistent measure of the non-take-up rate. Riphahn (2000) takes advantage of data from the 1993 Income and Expenditure Survey (EVS) to estimate a non-take up rate of 63.1 percent. Though its main strength is its sample size, the EVS has the significant shortcoming of not being as representative a sample as the SOEP.3 In their study about differences in the take up of social assistance between immigrants and native Germans, Bird et al. (2000) estimate a non-take-up rate for all households of about 60 percent for 1996. Only Bird et al. (2000) and Riphahn (2000) employ regression analyses to explain the welfare take-up rate. The international literature on determinants of welfare take-up is rather extensive, particularly with respect to welfare take-up for various programs in the United States and the United Kingdom. Craig (1991) and van Oorschot (1991, 1998) provide detailed discussions of the determinants of welfare take-up and the threshold models in particular. Much of our current understanding of the reasons for not claiming benefits for which one is eligible comes from sociological studies based on surveys of eligible individuals. An obvious advantage of survey information from eligible households is that surveys can illicit reasons for why certain characteristics are associated with a higher probability of claiming benefits. Based on such surveys, threshold models view the claiming process as a process of crossing thresholds. Not passing a well-defined threshold leads to non-take up of the benefits. For example, if households do not perdren, single adults, foreigners, and women than in the entire population (Statistisches Bundesamt, 1999). Neumann and Hertz (1998) show that in 1995 a relatively high proportion of children, women, foreigners, and larger families remain poor in the sense of not claiming benefits though having incomes below what the social assistance guidelines deem necessary. 3 Particularly, the EVS under-represents the foreign population and may do so selectively because being a voluntary quota sample probably implies that more assimilated foreign households are more likely to participate. Foreign households tend to be larger and tend to have higher needs and lower income. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 31 ceive a need for assistance, or if the benefits are not perceived as useful in meeting their needs, they will not claim. Likewise, households have to overcome a household-specific distaste for receiving public assistance before they will consider claiming benefits.4 In this paper, we extend the research by Bird et al. (2000) using 1996 data from the SOEP to simulate eligibility and obtain estimates of the non-takeup rate. Our empirical work is based on conventional utility maximizing consumer choice models according to which the refusal of benefits becomes utility maximizing if the claiming process involves costs that exceed the anticipated benefits (Moffitt, 1980, 1983; Ashenfelter, 1983; Cowell, 1986; Yaniv, 1997; Anderson and Meyer, 1997). Economic theory as well as empirical work predict that the probability of claiming increases with the level of entitlements and decreases with direct and indirect costs associated with claiming (Riphahn, 2000; Daponte et al., 1999; Blank and Ruggles, 1996; Fry and Stark, 1987; Moffitt, 1983).5 Sociological threshold models suggest that the perception of benefits and costs from claiming determine individual claiming behavior (van Oorschot 1991, 1998).6 Building upon this work, our regression model includes a variety of proxies for perceived costs and benefits as well as expectation about duration of eligibility. 4 Some of the studies also investigate the strength of the thresholds and allow for trade-offs, interconnections and the sudden impact of triggers. For example, a strong negative attitude toward benefits in general can keep a household from being aware of the existence of the benefits, but the break-up of the family may trigger changes in the attitudes toward benefits that can lead to more awareness and a higher likelihood of claiming. More recent work has added to the client-based threshold models features at the administrative level and at the scheme level that help explain non-claiming (van Oorschot, 1998). At the administrative level, giving insufficient information and advice, the use of complex forms, and handling of the claims in a way that is perceived as humiliating, for example, all contribute to higher rates of non-take up. At the scheme level, such features as vague entitlement criteria, means-testing, and whether or not the benefits supplement other income contribute to non-take up. 5 For example, research on the food stamps program in the United States, shows that the participation rate is lowest (.4 or 40%) for households who can expect to get between zero and forty dollars worth of food stamps. Participation is highest (.93 or 93%) for households who stand to receive between $203 and $600 (Daponte et al., 1999). Fry and Stark (1987) find the level of entitlements to be statistically significant in explaining participation in the British Supplementary Benefit program in a multivariate regression, a conclusion that is confirmed by Riphahn (2000) for the case of social assistance in Germany. 6 Households with a higher level of entitlement and consequently lower own income, ceteris paribus, are more likely to perceive the need for assistance, to perceive themselves as eligible for benefits, and to obtain more utility from receiving the benefits. Furthermore, because they perceive themselves more in need of assistance, they tend to be more informed about the available benefits and more likely to apply for support (van Oorschot, 1991,1998). Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 32 Hilke Kay ser and Joachim R. Frick 3. Modeling the Participation Decision The economic model underlying the empirical work in this paper it the one presented by Blank and Ruggles (1996). The unobserved participation decision of a given household i at time t, Pit, can be modeled as a comparison between the relative benefits and the relative costs of participation: T (1) Pit =f(U(YpM) - U(YnpM) - Cit(DCit,Sit), Y, Wnvd) j=t+1 Participation at any time t occurs when P¿t>0, hence when utility from participation, U(YPjt), exceeds utility from non-participation, U(Ynp¿t), minus the costs of claiming benefits, Cit(DCit, Sit), contingent on some expectations about future non-participation income, Xví=t+i Expectations about the future matter because the expectation that eligibility is only short-lived can lead to non-take-up and delayed claiming.7 In the empirical work, we employ proxies to capture a household's likely expectation about the duration of its eligibility. Dropping the subscript for period t for ease of notation, utility in either state depends on the respective incomes, income when participating, YP)¿, and income when not participating, Ynp>i in the social assistance program where: ^ Yp,i = Lp,i + Bp,i + Ynp,i ~ Lnp,i "I" NLnp i Income when participating in the social assistance program consists of the household's labor market income, Lp¿, the level of the entitlement, BP)¿, and the household's non-labor income, iVLP)¿. Income when not participating in the social assistance program consists of the household's labor market income, Lnp>¿, and non-labor income, iVLnp>¿. Additionally, we assume that LP)i and NLPyi are independent from BP)¿ since benefit receipt in essence requires that households have consumed other sources of income first. A household's costs of participating in the social assistance program, C¿, consist of the direct costs of participation in terms of monetary expenses and time such as transportation costs, DQ, as well as indirect costs such as stigma, difficulties in obtaining adequate information, uncertainty, and 7 Looking at women's participation in food stamps and in Aid to Families with Dependent Children (AFDC), Blank and Ruggles (1996) find that eligibility spells end quickly while participation spells do not, and that there is little evidence of delays in claiming by those who choose to participate. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 33 other non-pecuniary factors, Si.8 A detailed discussion of the operationalization of the variables for our empirical work follows in section V. 4. The German Social Assistance System In order to determine the non-take-up rate of eligible households in the German Social Assistance program, we first need to simulate eligibility and thus the provisions of the social assistance program. As in most countries, Germany's welfare state has historically relied heavily on the principle of social insurance rather than means-tested assistance. Until 1961, social assistance was not very important and the responsibility of local governments. In 1961, Sozialhilfe (Social Assistance, or SA) was introduced as a national, means-tested assistance program that remains today as the most important means-tested part of the German social welfare system.9 Social Assistance is divided into two branches. In this paper we will look at the more important one for our population of non-institutionalized households, Hilfe zum Lebensunterhalt (HLU) which is the SA system's provision for ongoing monthly payments to households deemed eligible on the basis of their incomes.10 For 1994, these support payments make up 82.5 percent of the system's expenditures for private households (not including expenditures on institutionalized clients - see Neuhauser, 1996, p. 635).11 8 In the existing literature, stigma enters the utility function in a variety of ways. In some models it affects an individual's utility function by lowering the benefits of welfare transfers (Yaniv, 1997; Moffitt, 1983) while in other models stigma increases the costs of take-up (Blank and Ruggles, 1996; Blundell et al., 1988). 9 There are some other means-tested programs. Certain forms of old-age pensions are available to people in circumstances of special need, and Wohngeld (housing subsidies) are based on family income, actual housing costs and the number of household members. Although less important now than in years past, the system of Lastenausgleich (equalizing the burdens of WWII) has an explicit means-testing aspect, and some of the special programs designed to help East Germans after unification have been means-tested as well. On the whole, however, these programs are not very large compared to the SA system. 10 More precisely, eligibility for social assistance is based on a "needs community (Bedarfsgemeinschaft) rather than a household. In most instances the household will be identical to the needs community, however, there are instances in which that is not the case. For example, the assumption of pooling resources within a household context results in an underestimation of otherwise eligible persons. As we describe in the simulation section, there are very few cases in which households are deemed ineligible but report receiving assistance. In our simulations, we either make these household eligible for assistance or we delete them from our analysis. Based on a 25% sample of Social Assistance recipients in Germany as well as data taken from the EVS and the Mikrozensus, Hauser et al. (2000) analyze the degree of concordance of "household" and "needs community". More than 80% of all West German needs communities are basically identical to the definition of a household in survey data. Misspecifications are somewhat more likely in needs communities with female singles and female lone parents (Hauser et al. 2000, p. 31). Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 34 Hilke Kayser and Joachim R. Frick We will explain our simulation algorithm in detail below. In short, however, each Bundesland (federal state) sets an income threshold which represents the minimum income necessary for a single adult individual to maintain a reasonably dignified existence in contemporary society - a mandate laid down by the German constitution (the so-called Eckregelsatz). This basic needs income is adjusted for families according to an equivalence scale. Additional fractions are allocated to a household with members in ongoing special circumstances, such as old age or lone parenthood according to fairly firm rules (the so-called Mehrbedarfs zuschlüge). For the most part, the HLU payments are made in cash, and they supplement the family's income up to the threshold, plus housing costs which again may be covered up to a certain threshold.12 As we will discuss in more detail below, when determining eligibility, a family's income is subject to a number of adjustments. Certain expenditures can be deducted, and in principle the receipt of social assistance requires that all income sources above a certain minimum level are used before claiming social assistance ("Nachrangigkeitsprinzip der Sozialhilfe").13 Furthermore, legislation entails leaving a certain amount of labor income to the employee, which is not taken into account when calculating eligibility. In addition to these regular monthly benefits, one-time supplements are available to help pay for special needs, such as replacing a broken furnace (the so-called einmalige Leistungen). In principle, eligibility is determined, and payments are allocated by local governments. 11 The 82.5 percent include one-time supplements (einmalige Leistungen). The remaining 17.5 percent of the SA system's expenditures for non-institutionalized clients are made up by Hilfe in besonderen Lebenslagen (HBL). Under HBL, income assistance is provided for people in certain circumstances where some expenditure is considered necessary, but the expenditure is too high for the household's current income (see Bundesministerium fur Arbeit und Sozialordnung, 1990, pp. 452 ff.). Thus, pregnant mothers may receive HBL assistance to obtain pre-natal care; disabled citizens may receive one-time help that allows integration into a certain workplace; if the main bread-winner of a household dies, temporary assistance may be given to some other household member to keep the household intact until new sources of income are found; temporarily homeless persons may be assisted with cash until some housing arrangement can be found; and persons with addictions or severe mental illness may be given one-time help to access temporary crisis services. 12 Some HLU goes to people in institutions, where the payments are made directly to the institution and not to the individual. Since we explicitly exclude institutionalized people from our data, for our purposes HLU is basically a cash assistance system. 13 The German Social Assistance Program and provisions laid out in the German civil law make it mandatory for parents to help their adult children and for adult children to support their parents before claiming social assistance benefits. Parents may not be required to support their adult child in case of pregnancy or when raising an own child up to the age of 6 years. For the most part, these provisions are not incorporated in our simulation to its full extent for lack of detail in the data. Some of the support is part of our analysis if it comes in the form of private transfers from outside the household which are included in our income measure. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 41 Table la Social Assistance (HLU) Eligibility and Take-Up in Germany 1996 (in 1,000 Households; percent in parenthesis) Criteria for Eligibility: Household Need HNi > (1 ± x) * Adjusted Household Income HY{ x = .20 x - .10 x= .05 x = .00 x = -.05 x = -.10 x = -.20 Total Number of Households (*1000) 37,287 37,287 37,287 37,287 37,287 37,287 37,287 • Not eligible for HLU 35,783 (96.0) 35,428 (95.0) 35,195 (94.4) 34,868 (93.5) 34,463 (92.4) 33,870 (90.8) 32,343 (86.7) • Eligible for HLU 1,504 (4.0) 1,859 (5.0) 2,092 (5.6) 2,420 (6.5) 2,824 (7.6) 3,417 (9.2) 4,944 (13.3) Of those: • HLU receipt 892 (2.4) 892 (2.4) 892 (2.4) 892 (2.4) 892 (2.4) 892 (2.4) 892 (2.4) • No HLU receipt 612 (1.6) 967 (2.6) 1,200 (3.2) 1,528 (4.1) 1,932 (5.2) 2,525 (6.8) 4,052 (10.9) Non-take-up rate of HLUa 41.3 52.4 57.7 63.1 68.4 73.9 82.0 a Number of households not receiving HLU (non-claimers) out of all eligible households (the socalled Dunkelziffer). Source: SOEP 1996, Authors' calculations, weighted. need threshold, HN{. This represents the most stringent eligibility rule. In the last column a household is classified as eligible if 80 percent of its income falls short of the need threshold, representing our least stringent measure. The first two rows indicate the number and respective population share of households that are or are not eligible for social assistance. Eligible households are further divided into those actually receiving HLU and those not. The ratio of the households not receiving HLU out of the total number of eligible households gives rise to the non-take-up rate (the so-called Dunkelziffer). It is possible that incomes in our sample are measured with error. On the one hand, if incomes are underreported, or if we incorrectly assign additional needs to some households during the month of the interview, then the results from columns toward the left will present a more accurate picture.27 On the other hand, there are factors that suggest using a less stringent criterion for determining eligibility from the columns toward the right. For ex27 For example, we do not know precisely the stage of an expecting mother's pregnancy so that we may incorrectly assign additional needs to an expecting mother who does not qualify at that point in time. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 42 Hilke Kayser and Joachim R. Frick ample, households may receive one-time supplements (einmalige Leistungen) for which they are eligible even if they are not eligible for social assistance.28 In that case, our simulations underestimate the number of eligible households. Furthermore, we may fail to incorporate all the components that enter the household's needs to their full extent, and administrators may be more generous in granting social assistance than our simulations prediet.29 According to our simulation results in table la, between 4.0 and 13.3 percent of the 37.3 million households in Germany in 1996 are eligible for HLU, depending on the eligibility rule. The non-take-up rate among the eligible households is 41.3 percent if the most stringent rule for eligibility is used and increases to as much as 82.0 percent of the eligible population as more households become eligible with a less stringent eligibility rule. Assuming that our simulations are essentially accurate and households are eligible if their needs exceed their income in our base case scenario (i.e. x = 0), then the non-take-up rate of social assistance is 63.1 percent. Neumann and Hertz (1998) and Riphahn (2000) refer to the proportion of those who do not receive social assistance though they are eligible - and hence are considered needy - as the shadow rate of poverty (Dunkelziffer der Armut).30 Neumann and Hertz estimate this shadow rate of poverty for 1995 to be 3.4 percent of the population of all individuals. Riphahn estimates a shadow poverty rate of 2.04 percent of all households. According to our base case scenario, we predict a slightly higher shadow poverty rate among households in Germany of 4.1 percent (the proportion of those who are eligible and have no HLU receipt in table la), but the results range from 1.6 percent to 10.9 percent if eligibility rules are changed.31 28 These one-time supplements are designed to help a household purchase expensive durable items such as a replacement for a defunct refrigerator, or winter coats for the family. 29 Hartmann (1985) points out that eligibility does not only depend on the economic situation of the household but also on the eligibility criteria determined by the law, by procedural regulations accompanying the law, and by legislation and administrative practices at the local level. Small differences in the threshold through differences in administrative practices that even the best simulation cannot be expected to pick up will therefore have a rather large impact on the estimated size of the eligible population. Hartmann (1985, pp. 174-75) also discusses various reasons for granting HLU even if the household's income exceeds the needs threshold, such as housing costs that are over-reported to the administrators. 30 Calling this number the shadow rate of poverty is not unambiguous. It assumes that households are poor if their income makes them eligible for social assistance benefits and they do not receive benefits. Particularly in light of Hartmann's (1985) finding of a great density of households right around the needs threshold the distinction between non-poor for households with incomes slightly above the needs threshold and poor for those slightly below appears at best arbitrary. 31 Riphahn's (2000) somewhat lower proportion can at least in part be explained by the differences in the data used and in the simulations. Riphahn includes fewer of the Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 43 Table lb displays the results of a similar analysis for the entitlement level. We present the simulated entitlement level for all eligible households, and separately for households who do and do not actually receive HLU. Multiplying the average entitlement levels by the corresponding number of households gives rise to the aggregate claimed and unclaimed benefits in the month of the interview. The last row then presents the proportion of the total HLU entitlements for which households are eligible that are not claimed by eligible households. Table lb Monthly Entitlement Levels by Take-Up Status in Germany 1996 (in DM / Month; Standard Deviation in parenthesis) Criteria for Eligibility: Household Need HNi > (1 ± x) * Adjusted Household Income HY{ x = .20 x=.10 a: = .05 x = .00 x = -.05 x = -.10 x = -.20 Eligible for HLU 619 (1084) 595 (1084) 585 (1073) 560 (1084) 551 (1078) 529 (1055) 518 (1008) Of those: • HLU receipt 732 (1085) 792 (1075) 819 (1071) 830 (1126) 880 (1121) 902 (1152) 959 (1236) • No HLU receipt 459 (960) 416 (894) 413 (861) 401 (842) 399 (823) 398 (795) 421 (771) Total Claimed Benefits (*1000 DM) 652,944 706,464 730,548 740,360 784,960 804,584 855,428 Total Unclaimed Benefits (*1,000 DM) 280,908 402,272 495,600 614,256 770,868 1,004,950 1,705,892 Unclaimed Benefits Rate a 30.2 36.4 40.5 45.3 49.5 55.6 66.6 a Estimated total benefits for non-claimers out of estimated total benefits for all eligible households. Source: SOEP 1996, Authors' calculations, weighted. additions to needs income than we do. To the best of our knowledge, Riphahn does not control for additional needs of pregnant women and the deductible receipt of benefits paid to families of war veterans. Furthermore, social assistance receipt is measured as receipt during at least one month in the previous year while the eligibility restrictions apply to the household's situation at the end of the year. This could lead to erroneous assignments since households whose economic situation has improved over the year may not be considered eligible even if they were in the month(s) during which they received assistance. It is thus likely that Riphahn's base case scenario underestimates both the proportion of eligible households and the proportion of hidden poor households out of all households. However, we would like to emphasize that as a proportion of all eligible households, our estimated non-take-up rate is almost identical to that found by Riphahn (63.1 percent versus 62.7 percent). Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 44 Hilke Kayser and Joachim R. Frick These descriptive results clearly confirm the hypothesis that benefits for non-claimers are drastically lower when compared to benefits for households who actually receive HLU. In our base case scenario monthly entitlements of successful claimers are about DM 830, while non-claimers forgo on average DM 401, yielding an unclaimed benefits rate of 45.3 percent. This estimate of "savings" by the authorities ranges from about 30 percent to 66 percent for the scenario with the least stringent eligibility rule. Our results confirm that claiming behavior is closely related to the relative size of the benefits to which a household is entitled, regardless of the stringency of the eligibility criterion. Table 2 summarizes the information presented in the previous table for various demographic groups, using the base case scenario. Our results suggest that older households are less likely to be eligible for HLU than households with heads that are less than fifty years of age (5.2 or 5.3 percent compared to 7.6 percent), supporting recent evidence that older households are faring rather well (Krause and Habich, 2000: 325 f.). At the same time the non-take-up rate among older households is higher than average, indicating that of the relatively low proportion of elderly households in need of assistance, a disproportionately large number chooses to leave their entitlements unclaimed.32 Voluntary poverty among elderly who renounce aid from the government is thus a concern unless this result is driven by unobserved support from outside the household such as free meals or free transportation. Other results in Table 2 indicate on the one hand that a number of demographics can be associated with a below-average non-take-up rate: more likely to claim benefits are households with a female head, single-parent households, households with children, renters, households whose head is foreign-born, and households whose head is unemployed or not employed.33 On the other hand, we find various household characteristics that can be associated with an above-average non-take-up rate: less likely to claim are single-adult households, households with several adults with or without children, households living in metropolitan areas or living in the East, as well as households whose head is employed, full-time or part-time, or whose head is retired. The unclaimed benefits rate varies in the same direction as 32 Hartmann (1985) presents survey results that illicit reasons for non-claiming by younger and older eligible individuals. Both groups cite fear of stigmatization and lack of knowledge about eligibility rules, but elder households also more predominantly cite voluntary refusal behavior or pride (Verzichthaltung) as a reason for not claiming benefits. 33 Claiming behavior in case of unemployment is likely to be influenced by characteristics of the local labor market as well. Frick (1985) supports this hypothesis by showing an increase in the positive correlation of unemployment and HLU take-up over the period 1979 to 1982. Additionally, he finds that HLU-density is significantly higher in areas with high unemployment and poor chances of re-integration in the labor market. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 45 Table 2 Social Assistance Take-Up Rates and Average Monthly Entitlement Levels by Take-Up Status for Eligible Households by Household Characteristics (1996) Characteristic Eligible Household0 NonTake-Up Rate0 Claimed Benefits (DM/ Month) Unclaimed Benefits (DM/ Month) Unclaimed Benefits Rate0 All Households 6.5 63.1 830 401 45.3 Household Headd with age >70 5.3 77.5 731 408 65.7 Household Head with age >60 5.3 77.5 705 335 62.1 Household Head with age >50 5.2 65.6 711 369 49.8 Household Head with age <50 7.6 61.6 893 422 43.1 Female Head of Household 9.3 58.0 831 396 39.7 One Child 11.8 55.0 892 488 40.1 Two Children 7.0 45.5 1106 580 30.4 More Than Two Children 16.5 35.5 1093 786 28.5 Single Parent Household 38.8 28.1 929 517 17.8 Several Adults with Child(ren) 6.5 66.8 1122 588 51.4 Several Adults without Children 2.8 73.9 912 481 59.8 Living in a Metropolitan Area 6.7 66.4 974 309 38.5 Living in a Rural Area 6.7 64.8 789 453 51.4 Living in a Rented Apartment/ House 8.9 59.4 851 383 39.7 Living in the Eastern States 7.3 66.9 669 309 48.2 Foreign Born Ethnic German Head 10.4 54.5 887 317 29.9 Foreign Born Foreign Head 14.5 52.1 1099 360 26.2 Head Works Full Time 2.2 80.7 670 441 62.8 Head Works Part Time 10.6 84.1 827 508 76.6 Head is Unemployed 25.5 38.6 781 362 22.6 Head is not Employed 26.3 42.9 1024 502 26.9 Head is Retired 6.1 80.2 528 305 70.1 a In percent of population. b Number of households not receiving HLU (non-claimers) out of all eligible households. 0 Estimated total benefits for non-claimers out of estimated total benefits for all eligible households. d The head of the household is defined as the member of the household that is the main income earner. If no main income earner can be found, we maintain the designation as head from the SOEP. Source: SOEP 1996, Authors' calculations, weighted. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 46 Hilke Kayser and Joachim R. Frick the non-take-up rate, e.g. the proportion of outstanding benefits that are saved because of non-claiming by some of the eligible households is higher among elderly households than it is among households with a younger head. Results from the cross tabulations above largely coincide qualitatively with results presented in recent studies for German welfare take-up by Neumann and Hertz (1998) and Riphahn (2000). However, cross-tabulations may provide a very inaccurate picture of what contributes to higher nontake-up rates by failing to account for interactions between the various characteristics of a household. In the next section we will estimate probit regressions to understand which household characteristics can be associated with low program participation rates when controlling for related household characteristics. 7. Non-Take up behavior: Results from Probit Regressions34 Sample means and standard deviations for the household characteristics that are included in the regressions are presented in Table 3 for households according to their eligibility status and their claiming behavior. The regression results are all based on the eligibility criterion that defines a household as eligible if the household's simulated needs exceed their adjusted income.35 In Table 4 we present results from several probit regressions for the 429 HLU-eligible households in our sample. Column one, the short list of covariates, estimates the probability of HLU take-up as a function of demographic factors such as age, education, location and family composition. In the following columns we add first social control variables, then employment status variables, and finally - as our full model - both subsets. The short regression (Column 1) provides evidence that households with higher predicted incomes have a significantly and substantially lower probability of participating in the program. Ceteris paribus, lower incomes correspond to higher benefits, so our results support the hypothesis that higher entitlements or benefits have a significant positive effect on non-take-up. Also, ceteris paribus, households with higher incomes may fail to claim benefits because they have a lower perception of need (van Oorschot, 1998), and they are more likely to be part of a social group where eligibility for social 34 Despite our interest in non-take-up, in our regression we estimate the take-up of social assistance which provides more intuitive coefficients. With non-take-up as our dependent variable a positive sign would have to be interpreted as increasing the probability of not taking up benefits - a somewhat confusing statement. Higher nontake-up in our regression reveals itself through significant negative coefficients. 35 Other eligibility criteria generate qualitatively similar results. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 47 Table 3 Summary Statistic by Eligibility and Claiming Status Characteristic Not Eligible Total I Eligible NonI Claimers | Claimers Needs Income HNi (DM) 1734 1682 1619 1790 Needs Income HNi (DM) (1717) (1791) (1755) (1824) Adjusted Household Income (DM) 3794 1278 1315 1213 Adjusted Household Income (DM) (5465) (1919) (1661) (2242) Foreign-Born Ethnic German Head (%) 0.041 0.069 0.060 0.085 Foreign-Born Ethnic German Head (%) (0.473) (0.602) (0.580) (0.634) Foreign-Born Foreign Head (%) 0.072 0.177 0.146 0.230 Foreign-Born Foreign Head (%) (0.618) (0.908) (0.867) (0.956) Total Number of Kids Age <17 0.382 0.683 0.492 1.009 Total Number of Kids Age <17 (1.852) (2.536) (2.267) (2.737) Single Parent (%) 0.018 0.167 0.074 0.325 Single Parent (%) (0.319) (0.886) (0.643) (1.063) Several Adults without Children (%) 0.413 0.174 0.203 0.123 Several Adults without Children (%) (1.174) (0.900) (0.986) (0.746) Several Adults with Child(ren) (%) 0.205 0.207 0.219 0.186 Several Adults with Child(ren) (%) (0.963) (0.963) (1.014) (0.884) Age of Household Head (Years) 49.6 44.8 45.7 43.2 (41.5) (45.6) (50.1) (37.8) No secondary education (%) 0.167 0.436 0.452 0.409 No secondary education (%) (0.888) (1.179) (1.220) (1.116) Post secondary education (%) 0.294 0.093 0.096 0.088 Post secondary education (%) (1.086) (0.691) (0.723) (0.643) Northern States of West Germany (%) 0.202 0.194 0.196 0.190 Northern States of West Germany (%) (0.957) (0.940) (0.974) (0.890) East Germany (%) 0.181 0.206 0.218 0.185 East Germany (%) (0.918) (0.961) (1.013) (0.881) Western States of West Germany (%) 0.349 0.331 0.337 0.320 Western States of West Germany (%) (1.137) (1.119) (1.159) (1.059) Metropolitan area (%) 0.190 0.198 0.208 0.181 Metropolitan area (%) (0.936) (0.948) (0.996) (0.874) Rural area (%) 0.384 0.396 0.406 0.377 Rural area (%) (1.160) (1.163) (1.204) (1.101) Renting housing unit (%) 0.610 0.856 0.805 0.943 Renting housing unit (%) (1.163) (0.835) (0.971) (0.528) Income loss since last year > = 20% (%) 0.183 0.128 0.176 0.044 Income loss since last year > = 20% (%) (0.923) (0.794) (0.935) (0.468) Pessimist (%) 0.060 0.112 0.049 0.220 Pessimist (%) (0.568) (0.750) (0.531) (0.941) Believes own behavior does not affect 0.016 0.115 0.027 0.267 course of own life (%) (0.301) (0.759) (0.394) (1.004) Has strong ties to location (%) 0.786 0.668 0.725 0.570 Has strong ties to location (%) (0.978) (1.120) (1.095) (1.124) Does never attend church or religious 0.522 0.561 0.510 0.647 meetings (%) (1.191) (1.180) (1.226) (1.085) Head works Part Time (%) 0.056 0.095 0.126 0.041 Head works Part Time (%) (0.547) (0.696) (0.815) (0.448) Head is Unemployed (%) 0.044 0.218 0.134 0.363 Head is Unemployed (%) (0.490) (0.982) (0.834) (1.092) Head is Not Employed (%) 0.046 0.238 0.162 0.369 Head is Not Employed (%) (0.501) (1.012) (0.903) (1.095) Head is Retired (%) 0.282 0.264 0.336 0.142 Head is Retired (%) (1.073) (1.049) (1.158) (0.793) Source: SOEP 1996, Authors' calculations. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 48 Hilke Kayser and Joachim R. Frick Table 4 Determinants of Social Assistance Take-up in Germany, 1996 Coefficients from Probit Regressions (standard error in parenthesis) Short List Added Social Added FuH Model Characteristic Controls Employment (1) (2) (3) (4) Intercept -0.121 0.022 1.226* -1.097 Intercept (.625) (.656) (.727) (.763) Predicted Income (*1000) -0.614** -0.584** -0.236 -0.239 Predicted Income (*1000) (.164) (.164) (.208) (.209) Foreign-Born Ethnic German Head 0.083 0.091 0.075 0.106 Foreign-Born Ethnic German Head (.232) (.240) (.236) (-244) Foreign-Born Foreign Head -0.226 -0.157 -0.161 -0.086 Foreign-Born Foreign Head (.184) (.188) (.194) (.199) Total Number of Kids Age <17 0.541** 0.521** 0.381** 0.378** Total Number of Kids Age <17 (.099) (.100) (.110) (.no) Single Parent -0.026 0.019 0.195 0.223 Single Parent (.244) (.249) (.267) (.272) Several Adults without Child(ren) 1.099** 1.077** 0.611* 0.627* Several Adults without Child(ren) (.312) (.320) (-364) (.367) Several Adults with Child(ren) -0.034 0.042 0.281 -0.193 Several Adults with Child(ren) (.265) (.268) (.281) (.285) Age of head (* 10) 0.521** 0.386 0.356 0.263 Age of head (* 10) (.265) (.274) (.287) (.294) [Age of head (*10)]2 -0.053* -0.039 -0.031 -0.023 [Age of head (*10)]2 (.027) (.028) (.029) (.030) No secondary education -0.058 -0.052 0.084 0.073 (.166) (.169) (.175) (.178) Post secondary education 0.401 0.353 0.154 0.124 Post secondary education (.280) (.283) (.298) (.301) North -0.295 -0.346 -0.232 -0.277 (.214) (.221) (.219) (.224) East -0.734** -0.837** -0.510** -0.623** (.241) (.249) (.254) (.261) West -0.078 0.122 0.007 -0.049 (.173) (.179) (.177) (.182) Metropolitan area 0.386* 0.411** 0.266 0.298 Metropolitan area (.198) (.201) (.205) (.208) Rural area -0.217 -0.197 -0.190 -0.166 (.162) (.165) (.165) (-168) Renting housing unit 0.028 -0.080 0.211 0.117 Renting housing unit (.314) (.323) (.333) (.342) Income loss since last year > = 20% -0.302 -0.327 -0.356 -0.364 Income loss since last year > = 20% (.256) (.262) (.260) (.265) Pessimist -0.473** -0.383 -(.234) -(.240) Powerless -0.431 -0.436 -(.282) -(.285) Has strong ties to location --0.164 --0.104 Has strong ties to location -(.141}, -(.144}. Does never attend church — 0.335 — 0.323 or religious meetings -(.149) -(.151) Head works Part Time - - 0.037 -0.049 --(.263) (.265) Head is Unemployed --0.627 * 0.545 * --(.228) (.234^ Head is Not Employed --0.727 * 0.681 Head is Not Employed - - (.232) (.235) Head is Retired ---0.012 0.051 - - (.303) (.311) Log likelihood (n = 429) 253.9 243.4 243.7 236.0 *: p < 0.10; **:p< 0.05 Dependent variable is coded with 1 if household receives Social Assistance, and 0 otherwise. The omitted categories point to a single-adult, native household with secondary education, living in a smaller town in the South of Germany in an apartment or house that he/she owns. He/she also works full time, has little attachment to the community, does attend church or other religious gatherings, is generally optimistic and believes that his/her actions to a certain degree affect the course of his / her life. Source: SOEP 1996, Authors' calculations. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 49 assistance is low, such that the distaste for assistance and the stigma attached to claiming assistance may be higher (Cowell, 1986). The regression results do not show any significant effect of the immigrant status variables on participation in social assistance. Both the negative sign and the failure to be significant for foreign-born foreign heads support results found by Bird et al. (2000) who suggest that there is no empirical evidence that immigrant households enter Germany in order to abuse the social welfare system. Immigrants may face higher application costs from possible language barriers and unfamiliarity with the system, and they may not claim for fear of being asked to leave Germany. In Germany, dependency on social assistance can be used as a reason to discontinue the residency permit, though that is not the case for foreign-born ethnic Germans. None of these influences can be supported empirically. A number of variables regarding the composition of the household are included for a variety of reasons. With respect to the number of adults in the household we would expect a positive impact on program participation given that resources can be pooled to facilitate the claiming process. Looking at the regression results, we find that households with several adults without children are significantly more likely to claim benefits than the reference group of single-adult households. This supports the hypothesis that application costs may be lower for a household with several adults, though it is also possible that single-adult households consist to a larger extent of older individuals with the more pronounced attitude of pride and refusal of benefits. This latter claim is supported by our finding that the age of the household head affects social assistance take-up: take-up increases with age but at a decreasing rate. Up to age 49, ceteris paribus, older heads are associated with a higher take-up rate, but for heads age 50 or older, the likelihood of claiming social assistance decreases. The regression results also show that the number of children positively impacts on a household's participation, regardless of family composition. Thus, while children may make it more time consuming and costly to apply (Blank and Ruggles, 1996), these costs are apparently outweighed by an increased perception of need (van Oorschot, 1998), possibly a higher acceptance probability on the part of program administrators that may be less disdainful of claiming households with children, or by a higher participation among households who expect to be eligible for a longer period of time. After controlling for the number of children in the household, single-parent households are no more likely to take-up benefits than single-adult households. With respect to education our results do not support the claim that households with lower educational attainment - and therefore lower earnings poSchmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 50 Hilke Kayser and Joachim R. Frick tential - experience longer eligibility spells and as such are more likely to participate. If anything, our results suggest a positive relationship between education and take-up. Possibly, a higher level of education lowers the time costs of applying and completing the application form and increases the level of awareness of the program parameters. The model suggests a lower participation rate for households with a recent income loss because households may expect their eligibility to be only temporary. This group is also more likely to contain households that are eligible for the first time which may lead them to delay the claiming decision.36 Though the negative sign for the coefficient on income loss confirms our expectation, it fails to be statistically significant. Likewise, our results cannot support the hypothesis that households in rented accommodation are more likely to participate because they expect their eligibility to be more permanent. Our empirical results strongly support participation differences across regions. East German households are significantly less likely to claim benefits than the reference group of South German households. This is in line with the argument by Neumann and Hertz (1998), that a certain distaste for receipt of social assistance has remained from the old GDR system. Moreover, take-up of means-tested housing assistance (Wohngeld) in 1996 was significantly higher among East Germans.37 With lower entitlement levels and perhaps less of a need for assistance above and beyond housing benefits, participation is less likely.38 The dummy variable for living in a metropolitan area shows the expected significant and positive impact on take-up behavior.39 This supports the hypotheses that information is more easily distributed formally and informally in more densely populated areas, and that the anonymity of living in a larger city reduces stigma. The coefficient on rural is negative though it fails to be significant. 36 Blank and Ruggles (1996) find no evidence for delayed claiming. For Aid to Families with Dependent Children (AFDC) and food stamps they find that for claiming households, participation spells start almost immediately upon becoming eligible. 37 Based on GSOEP data, Frick and Lahmann (1997) determine 6.5% of all main tenants in West Germany to receive housing assistance, whereas this rate is twice as high in East Germany (13.1%). 38 We also find that adjusted incomes in the East are on average similar to those in the rest of Germany while they deviate less. Because of the higher density of incomes in the East, we may have more households that are close to the eligibility cut-off but do not qualify. However, regressions using a less stringent eligibility criterion that should define more of those households as eligible, also generate a statistically significant coefficient on East. 39 We define as rural households those that live in towns with fewer than 20,000 inhabitants and as metropolitan those households that live in cities with more than 500,000 inhabitants. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 (Non-)Take-Up Behavior of Social Assistance in Germany 57 Appendix Table Al Instrumenting Regression for Adjusted Family Income Variable P | Std.Err. Intercept 785.99 226.55 Foreign-Born Ethnic German Head -43.27 107.64 Foreign-Born Foreign Head -144.47 79.04 Total Number of Kids Age <17 390.99 43.20 Single Parent -516.97 154.09 Several Adults without Children 1499.76 62.33 Several Adults with Child(ren) 818.23 94.17 Age of Head 873.57 92.32 Age of Head 2 -77.82 9.29 No secondary education -302.62 66.21 Post secondary education 557.57 61.70 Living in North -138.45 76.80 Living in East -676.16 68.68 Living in West -139.11 62.85 Living in metropolitan area -785.53 54.82 Living in rural area 148.95 72.16 Renting housing unit -158.74 52.82 Father without secondary education -167.24 62.27 Father with post secondary education 136.32 66.13 Mother without secondary education -113.99 55.74 Mother with post secondary education 50.13 116.93 ISCO: Science 1203.20 124.53 Management 2148.51 172.78 Office 629.00 121.56 Trade 987.07 170.31 Service 297.05 146.11 Agriculture 182.73 332.21 Manufacturing 434.12 108.73 Industry: Agriculture -135.07 332.17 Energy 284.94 216.99 Chemistry 679.46 166.26 Plastics 452.96 328.68 Stone -53.17 364.52 Metal 531.87 113.69 Wood 404.33 200.84 Textile 284.82 283.71 Food -100.28 212.04 Construction 382.23 127.09 Trade 54.19 145.45 Transportation 145.67 144.83 Banking 556.49 182.00 Other service 158.48 122.28 Non-profit -315.64 218.98 Public Sector 274.61 131.98 Adjusted R* (N=6,567) 0.374 Source: SOEP 1996, Authors' calculations. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28 58 Hilke Kayser and Joachim R. Frick Table A2 Results from the Full Model in Table 4 for Alternative Specifications Selection Bivariate Eligible if Characteristic Model Probit Model 1.05HY* > HNi & pr P pr (3 pr Intercept 0.714 0.005 -0.524 0.546 -1.263 0.122 Predicted Income --- - -0.243 0.253 Foreign-Born Ethnic German Head 0.0001 0.999 -0.001 0.996 0.151 0.563 Foreign-Born Foreign Head 0.055 0.402 -0.163 0.435 0.017 0.938 Total Number of Kids Age <17 0.074 0.008 0.199 0.038 0.406 0.001 Single Parent 0.035 0.693 0.062 0.840 0.227 0.439 Several Adults without Child(ren) 0.136 0.053 0.390 0.076 0.713 0.073 Several Adults with Child(ren) -0.092 0.229 -0.290 0.233 -0.257 0.392 Age of head (* 10) 0.066 0.415 0.208 0.423 0.327 0.308 [Age of head (*10)]2 -0.004 0.598 -0.014 0.590 -0.030 0.367 No secondary education 0.008 0.883 -0.012 0.951 0.160 0.401 Post secondary education 0.044 0.601 0.095 0.729 0.396 0.231 North -0.094 0.182 -0.294 0.197 -0.238 0.324 East -0.142 0.049 -0.458 0.048 -0.662 0.016 West -0.001 0.982 -0.032 0.859 -0.031 0.872 Metropolitan area 0.093 0.160 0.267 0.204 0.245 0.273 Rural area -0.051 0.322 -0.134 0.405 -0.165 0.355 Renting housing unit 0.007 0.941 0.111 0.736 0.211 0.557 Income loss since last year > = 20% -0.121 0.120 -0.313 0.243 -0.390 0.164 Pessimist 0.123 0.109 0.355 0.127 0.482 0.064 Powerless 0.149 0.091 0.414 0.151 0.408 0.182 Has strong ties to location -0.028 0.536 -0.090 0.528 -0.032 0.838 Does never attend church or 0.096 0.042 0.284 0.065 0.389 0.016 religious meetings Head works Part Time -0.029 0.717 -0.144 0.611 -0.169 0.552 Head is Unemployed 0.149 0.055 0.344 0.229 0.427 0.089 Head is Not Employed 0.200 0.010 0.483 0.100 0.510 0.040 Head is Retired -0.028 0.776 -0.093 0.780 0.037 0.912 lambda -0.148 0.076 - - - - Source: SOEP 1996, Authors' calculations. Schmollers Jahrbuch 121 (2001) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.121.1.27 | Generated on 2023-04-04 12:28:28