«Department of Economics and Center of Economic Research (SØF), Norwegian University of Science and Technology, N-7491 Trondheim, Norway. Abstract ...»
Welfare competition in Norway: norms and expenditures
Jon H. Fiva and Jørn Rattsø*
Department of Economics and Center of Economic Research (SØF), Norwegian University of
Science and Technology, N-7491 Trondheim, Norway.
The paper evaluates the empirical importance of welfare competition. Our contribution
is to separate between the welfare benefit norm decided at the political level and the actual
welfare benefit payments of a standardized person. Utilizing spatial econometric methods we find statistical significant and similar strategic interaction between local governments for both politicians (norm) and bureaucrats (actual benefits). We do not find support for an income effect representing altruism, the standard argument for welfare benefits in the theory literature. The strategic interaction and geographic pattern identified does not necessarily imply underprovision, since the grant financing of the local governments may generate overall excessive public spending.
Date: March 29, 2005 JEL classification C21, D78, H73 Keywords: Welfare benefits, strategic interaction, spatial econometrics.
* Corresponding author.
Tel.: +47 73591934; fax +47 73596954; e-mail: email@example.com.
1. Introduction Globalization with increased mobility of households and firms is often described as a threat to distribution policy. Some state this challenge in dramatic terms, with titles like 'can the welfare state survive?' and with propositions of a 'race to the bottom'. The issue has raised policy concern about fiscal decentralization and in particular about EU integration. Sinn (1994) has warned about the consequences of economic integration for welfare. A large literature initiated by Musgrave (1959) offers warnings against decentralization of the responsibility of distribution. Governments are encouraged to set fiscal variables to influence the location of households and firms when mobility is high, and the consequent fiscal competition will influence taxation and spending. The comprehensive theoretical literature is not matched by much empirical evidence.
In the area of welfare policy, countries have typically decentralized responsibilities to states and municipalities to take advantage of local knowledge and administration. The associated welfare competition may serve as a threat to the implemented welfare policy.
When taxpayers and welfare recipients are mobile it seems likely that the local governments will seek to attract wealthy households and avoid potential welfare recipients. The empirical importance of such welfare competition is addressed in a series of US studies summarized by Brueckner (2000) and in recent studies of the UK (Revelli, 2003) and Sweden (Dahlberg and Edmark, 2004). The present paper provides empirical evidence for another country with decentralized welfare policy, Norway. The contribution of the paper is to separate between welfare policy decisions and actual welfare payments. We also throw light on the importance of the local income level, the altruism effect dominating the theoretical literature.
The implementation of welfare policies includes the guidelines set by political institutions and the actual payments made by the welfare bureaucracy. This complicated line of implementation is typically overlooked in empirical studies. The US studies have concentrated on the benefit levels for AFDC (Aid to families with dependent children) and most authors (Berry et al. 2003, Figlio et al., 1998, Saavedra, 2000) have used the maximum amount given to a standardized family as the measure of benefit level. According to Peterson and Rom (1990), the maximum consists of a 'needs standard' and a share of the standard covered. States differ both in their assessment of needs and in the share financed. Actual expenditures per recipient are used for the US by Bailey and Rom (2004), Revelli (2003) for the UK, and Dahlberg and Edmark (2004) for Sweden. Average actual benefit payments 3 represent both discretion in the welfare bureaucracy and composition effects. Bailey and Rom (2004) argue that maximum benefits and average benefits are highly correlated for AFDC. In our data there are large composition effects when average benefits are compared, primarily as a result of different duration of welfare spells.
The relationship between politically determined norms and actual welfare payments to the clients represent a key linkage in the public choice literature between politics and bureaucracy. Under an assumption of rationality, the local politicians will take into account the bureaucratic behavior when norms are set. The variation in both norms and actual benefits are investigated below. The norms are set as guidelines for the administration and are specified as an amount paid to a 'standard user' (singe individual without children) per month.
The estimates of welfare competition in norms reflect the political response.
The analysis of actual welfare benefits is based on a unique dataset of computed expected benefits in each municipality based on individual characteristics, worked out and documented by Langørgen and Rønningen (2004). They utilize data for most of the grown-up population in Norway (more than 2,5 million individuals) and estimate expected welfare benefits received for comparable individuals. Welfare benefits are means-tested and based on an evaluation of the demands of each individual on a case-by-case basis. The individual demands vary and the welfare recipient population is quite heterogeneous, ranging from individuals in need for support for a few weeks till quite permanent welfare clients. The expected benefit measure is an attempt to take into account this heterogeneity.
In the analysis we investigate the possibility that the decisions of Norwegian local governments about welfare benefit levels depend on the benefit level in ‘neighboring’ municipalities.1 We apply spatial econometrics methods to estimate the strategic interaction among local governments. The starting point is a fiscal demand function where the benefit level in each municipality is dependent on benefits in neighboring municipalities as well as economic and political characteristics. The endogeneity of other municipalities’ welfare benefits is handled with instrumental variables.
Section 2 outlines the welfare competition mechanism, section 3 presents the econometric design, and the data are described in section 4. Section 5 shows and discusses
our estimated interaction models, and section 6 discusses other determinants of welfare benefits. A short summary of results and challenges for future work are dealt with in the concluding section.
2. Welfare competition mechanism Centralization or decentralization of redistribution policy is an old issue in the economics literature. Oates (1972) offers an early analysis of the role of the mobility, whereby local redistribution can chase the rich to other municipalities and attract the poor.
Orr (1976) formalizes the altruistic argument for welfare benefits, and shows that poor living in municipalities where they are a small fraction of the population are expected to receive higher welfare benefits than in municipalities where they are a large fraction. This cost effect implies that an inflow of poor people to a municipality will reduce the benefit level. Brown and Oates (1987) extend the Orr framework to include a migration function explicitly, which shows the elasticity of the number of poor with respect to the benefit level. They derive how the benefit level varies inversely with the elasticity of the migration function. The mobility of the poor is a source of inefficiency in decentralized systems. The extensive theoretical literature on mobility and redistribution is summarized by Cremer and Pestieau (2004).
A simple and attractive theoretical framework relevant for our analysis is worked out by Wheaton (2000). A fixed national welfare population is distributed among municipalities and receives municipality specific welfare benefits. The welfare population is assumed small relative to the total population, and the decisive representative voter is an employed immobile taxpayer. The municipalities differ in population size and private income level. As a reference point, if we assume that the welfare benefit decision is taken at the national level, there is no welfare migration to take into account and the relevant tax price for benefits is the share of recipients in the population. This result reproduces Orr (1976). When welfare benefits are decentralized and welfare migration is taken into account, the response of the welfare recipients is internalized in the political decision. Wheaton develops the migration story of Brown and Oates (1987) emphasizing the elasticity of the recipients with respect to the benefit level. The elasticity raises the tax price of benefits and consequently contributes to underprovision compared to the national decision.
5 The migration part of the model assumes that welfare recipients have their own evaluation of the attractiveness of each municipality, and in addition to this their utility depends on the welfare benefits received in each municipality. The likelihood that recipients locate in a specific municipality follows a logistic function. This supply side of the welfare market implies a positive relationship between benefit level and welfare recipients in the migration equilibrium. The demand side shows how the political decision about the benefit level depends on the size of the welfare recipient group, and the benefit level will be reduced when the number of recipients goes up. The decision is affected by the benefit level in all municipalities through the endogenous determination of welfare recipients. The geographic pattern of benefits and recipients will depend on the migration response of the welfare recipients. When the migration response is strong, all municipalities spend less on welfare.
The supply of welfare recipients is responsive to the benefit level, the benefit levels will vary little, but the recipient shares will vary much. The overall pattern will show small variation in benefits, but large variation in recipient shares. On the other hand, when the migration response is small, we expect a pattern with large variation in benefits and small variation in recipient shares. The mechanisms of the model are similar to the assumed moving costs in Smith (1991). When psychic moving costs vary by individual, the competitive mechanism mainly will be represented by the individual welfare recipients with low moving costs.
An alternative understanding of the equilibrium mechanism in the US studies assumes wage adjustment. Brueckner (2000) presents the adjustment mechanism based on Wildasin (1991), also discussed by Saavedra (2000) and Dahlberg and Edmark (2004). In this setup, the welfare recipients earn unskilled wage income at the labor market, and the wage response secures migration equilibrium. There is a cost effect of the number of welfare recipients, but also a wage effect. The wage adjustment may give negatively sloped reaction functions since higher benefits at a neighbor will induce outmigration and higher unskilled wage level, thereby motivating lower benefits. It seems unrealistic in our context to give such a prominent role to the unskilled wage adjustment, since most of the recipients are outside the labor market.
There is a separate literature addressing the mobility of welfare recipients. Most observers will agree with the conclusion of Meyer (2000) based on the US evidence that there is welfare induced migration, but that it is modest in magnitude. There are serious methodological challenges to identify welfare migration. Actual migration flows may be 6 small because municipalities do respond to the competitive pressure. Welfare competition may also be observed even when potential welfare migration is negligible. The performance of neighboring municipalities may give voters information to evaluate their own municipality.
Salmon (1987) discusses the argument for decentralization based on such yardstick competition, and Besley and Case (1995) offer empirical evidence. In this paper we will make no attempt to investigate the sources of welfare competition.
The Wheaton model offers more specific hypotheses about municipality characteristics, in particular the private income level. Municipalities with higher private income level (and also with higher grants) have higher marginal benefit of altruism and will set a higher benefit level and have higher share of welfare recipients. Other rationales for redistribution to the poor can be argued. The marginal benefit of redistribution may increase with income due to a desire to reduce the negative externalities attached to poverty (such as crime etc). Redistribution is not necessarily increasing with income level of the municipality.
If income level reflects the extent of poverty, social insurance may lead to higher redistribution with lower income level. The possibility of becoming poor motivates the nonpoor majority to redistribute. This leads to political economy arguments that may imply a negative relationship between income level and benefit level. Boadway and Keen (2000) give an overview of motives for and politics of redistribution. The political aspects generally imply that income distribution influences the redistribution policy. The key hypothesis was suggested by Meltzer and Richard (1981), that more inequality generates more distribution.
We include a measure of the income distribution as a control variable here.
The theoretical literature of welfare competition discussed above implies a simultaneous determination of welfare benefits and welfare recipients. We concentrate on the reduced form determination of welfare benefits. The estimated equation for welfare benefits
in municipality i under strategic interaction can be written:
3. Empirical modeling of welfare competition In the econometric literature strategic interaction is known as spatial autocorrelation.
The formal framework used for the statistical analysis of spatial autocorrelation is a so-called spatial stochastic process. We follow the most frequently used approach to formally express spatial autocorrelation and specify a functional form for the spatial stochastic process that
relates the value of the random variable at a given location to its value at other locations:2
b = α Wb + xβ + u (2)