Expenditure Decentralization: Does It Make Us Happier? An Empirical Analysis Using a Panel of Countries

This paper analyzes whether fiscal decentralization of education, health, housing, social protection, recreation, culture and religion, public order and safety, and transportation have a significant effect on individual well-being. The empirical analysis is based on a non-linear hierarchical model that combines individual data (level 1) with country-level data (level 2). We match 89,584 observations from the World Value Service and the European Value Service (various years) with the average value of data recorded for 30 countries by the Government Financial Statistics (IMF). While fiscal decentralization in education and housing appears to have a negative effect on well-being, this effect is positive in the cases of health and culture and recreation. We interpret this as evidence in favor of a “selective” decentralization approach.


INTRODUCTION
According to the economic analysis perspective, decentralization is justified because it improves the efficiency of the public sector management. On the one hand, decentralization favors the consumer, since the sub-national governments know and satisfy the preferences of the citizens better (Oates 1972). On the other hand, decentralization also improves the territorial productivity of public goods and services (Oates 2005;Lockwood 2009 andWeingast 2009).
On the theoretical level, it is argued that the gains in efficiency generated by decentralization can contribute to greater economic growth, although no conclusive empirical evidence in favor of this hypothesis exists (Martínez-Vázquez J. and McNab R. 2003). Nevertheless, there is a greater consensus about the significant effects of decentralization on public spending and its indirect contributions to public well-being (Letelier 2012). Still, there are numerous arguments that deny the potential positive effects of decentralization, among which are the weakness of local bureaucracy, the implicit risk in the excessive proximity between private and public interests, and the scale economy losses in the provision of public utilities,… (Prud'homme 1995).
Though some studies have been carried out that analyze the relation between decentralization and happiness (Frey andStutzer 2000 andBjørnskov et al. 2008 has not yet been investigated. Given that decentralization in general, and fiscal decentralization (FD) in particular, is a complex phenomenon whose impact differs according to the specific area of public management Saez 2013 2015), this paper states that its effect on subjective well-being depends on the specific area that is decentralized. As previously noted in the abstract, this effect is identifed through a model one multilevel ordinal logit with I intercept random and fixed effects (Goldstein 2003;Rabe-Hesketh, Skrondal and Pickles 2005;Raudenbush and Bryk of 2002), the analysis object areas are education, health, housing, social protection, recreation, culture and religion, public order and safety, and transportation. The sample of data utilized includes 89,584 individual observations of 30 countries. From the estimations carried out, it is inferred that decentralization in the areas of recreation, culture and religion, and health have a positive effect on happiness; where as decentralization of the functions of education and housing have a negative impact.
The remainder of this document is presented in the following structure: in section 2, we review the literature that links the theory of decentralization to happiness, epigraph 3 puts forth, the theoretical framework, the methodology utilized and the data employed, while in section 4, the results of the econometric estimations are analyzed. Conclusions are presented in section 5.

LITERATURE REVIEW
The theory of subjective well-being is a field of investigation whose origins go back to the 1970`s and 1980`s (Easterlin 1974;Scitovsky 1975;Kapteyn and they Go Praag 1976;Morawetz 1977;Ng 1978;Wansbeek and Kapteyn 1983;Martin and Lichter 1983;Sirgy et al. 1985;and Headey and Krause 1988). The goal of this theory is to explain life satisfaction through the lens of ordinal utility.
The works of Praag, Frijters and Ferrer-i-Carbonell (2003) and Ferrer-i- Carbonell and Frijters (2004) have been very important in advancing the theory of subjective well-being in different spheres: personal health, family financial situation, working conditions, and leisure and free time, among others.
In terms of the economics of happiness, Frey and Stutzer (2000) differentiate three categories of exogenous factors that determine subjective well-being: i) personality and demographic factors (age, sex, marital status, level of education, ideology, religion, etc; ii) micro and macroeconomic factors (level of income, the unemployment, the inflation, etc); and iii) the institutional context (democratic state, federalism, decentralization of spending, etc). There are various studies analyzing the impact of the institutional context. Frey andStutzer (2000 and, Stutzer and Lalive (2004), Frey (2008) and Bjørnskov et al., (2008and 2010 have analyzed the effect of some institutions on happiness. Radcliff (2001) investigated the role of government ideology and other characteristics of the Welfare State. Veenhoven (2000) showed that in affluent countries, political and individual liberty have a positive effect on happiness. Moreover, in less affluent countries economic institutions and courts have a greater influence. Bjørnskov et al., (2010) concluded that in more affluent countries, political institutions have more influence on personal satisfaction. More recently, Voigt and Blume (2012) have found that a positive correlation exists between happiness and federalism.
Because the analysis of the effect of institutions on subjective well-being is still a new area of investigation, less is known about the link between decentralization and happiness. Frey and Stutzer (2000) carried out a pioneering study in which they analyzed the effects of decentralization on an interregional level in Switzerland. They concluded that institutional factors, such as government initiatives, referendums, and local autonomy, have a significant and positive effect on the satisfaction of the Swiss. Nevertheless, this effect is dependent upon the direct link that exists between the binomial democracyvoter preferences and subjective well-being. Similiar studies were carried out by Díaz-Mountain and Rodriguez-Pose (2012), who extended their analysis to every European country and studied how different powers and resources of regional and local European governments improve the level of individual satisfaction. Bjørnskov et al., (2008), making use of a more extensive database that included 60000 individual observations of 66 countries, concluded that decentralization of spending does not have a significant impact on happiness. Sujarwoto and Tampubolon (2015) found that in Indonesia a developing country, fiscal decentralization if it increases happiness, but political decentralization is not significant.
From the standpoint of the probable impact of decentralization on the performance of the State, theoretical literature has made significant contributions. On the one hand, theoretical literature shows the positive effect of decentralization on efficiency and the quality of public spending (Oates 1972).
On the other, there is an intense debate regarding the possibility of reaching greater degrees of decentralization in the context of the limited professionalization in the lower levels of government, the greater feasibility of corruption and capture of the elite brought about by the excessive proximity between the private and public interests, and, finally, the insufficient quality of democracy in developing countries (Prudhomme 1995, Inman andRubinfeld 2000;Storper 2005). In terms of the first statement, it is inferred that the impact of decentralized management and financing is framed as a trade-off between the benefits of having more realistic information about the local context and the cost of a reduction in the operation scale (Letelier and Sáez-Lozano 2013). In general, it is expected that those spending functions that require considerable coordination at the national level will be more efficiently carried out at the central or intermediate levels of government. On the contrary, the management and financing of those services recognized as being local public goods which seeing as they affect the quality of said goods require specific knowledge of their local context, should generally be considered the responsibility of local government. (Letelier and Sáez-Lozano 2013). Therefore, we can conclude that the impact of the functional decentralization on the quality of public goods and, by extension, its effect on happiness, depends on the specific functional area and/or public good object of analysis.

THEORETICAL FRAMEWORK, METHODOLOGY AND DATA
To analyze whether or not the decentralization of different spending functions influences happiness, we should begin by developing a theoretical framework and describing the methodology of explaining the happiness relation. We end this section with a presentation of the basic characteristics of the data used.

Theoretical Framework
Happiness * ij S is a continuous and latent variable that reflects the level of subjective well-being. * ij S is determined by two sets of explanatory variables: i) individual X ij , that represents the characteristics of individuals i (level 1) in the country j (level 2), and grouped Z j , in order to measure the degree of decentralization that public spending functions have undergone in each country j.
We define the following lineal relation between the endogenous and the explanatory variables as:  and λ are the coefficients to be estimated, and ij  the error term.
Given that happiness is a unobservable variable, we define it through the level of individual satisfaction S ij . The relation between * ij S and S ij, for the m-category of S ij, is:

S m 
A priori, we assume that individuals i are nested in the countries j. Therefore, we assume that   ij VAR  is different for each group j and that there is conditional independence among the observations. This assumption allows us to relax the condition of homoskedasticity. Additionally, we assume that the influence of the predictor variable is fixed in two levels of analysis and that a random term U 0mj exists, that collects the inter-group differences.
In expression (3), the model one is deduced multilevel ordinal logit with I intercept random and fixed effects 1 of m-category of S ij 2 (see the appendix): is the constant term,  lmj represents the parameters that measure the effect of X ij individual variables, λ rmj associated coefficients to Z sj variables, and U 0mj is the random effect of country.
From the review of the literature carried out in the previous section, it is inferred that the following hypothesis is feasible: Hypothesis 1. The effect of fiscal decentralization can differ according to the specific type of spending function. The net impact will depend on the differential between the benefits of more information that results from decentralization and the cost of a more reduced scale of operation.
The previous hypothesis is based on the assumption that individuals are capable of identifying the magnitude of the net profit of decentralization, since there is an observable relation between the quality of the public good in 1 There are three reasons for not specifying multilevel ordinal logit model with intercept and slope random of X ij : i) the theory does not justify the effect of X ij differs in each unit j due to unobservable factors; ii) differential effect of grouped j X ij is explained by Z j variables; and iii) the literature does not justify the inclusion of variables representing the interaction between X ij and Z j variables.
2 Upon being the endogenous ordinal variable S ij , we can specify ordinal logit o probit model. It given that we suppose fixed effects in the explanatory individual variable, we reject the option of ordinal probit model since produces inconsistent estimator.
question and its cost. The cost is implicit to the relative magnitude of local taxes.

Hypothesis 2. Only the decentralization of expense functions, which citizens
consider within the jurisdiction of sublevel governments, contributes to an increase of happiness. On the contrary, the decentralization of public goods (services), whose provision is perceived as cost efficient if provided by the central government, will result in a reduction in individual satisfaction.

Methodology
To estimate the model (4)  We also estimate two additional models: logit ordinal with I random intercept and fixed effects in the individual explanatory variable, and ordinal logit with I random intercept. In the appendix, the specification of both models is described.

The data
The database of this investigation has been built from the information supplied     The data for the endogenous and individual explanatory variables were derived from the WVS and the EVS. The source is cited in the second column of Table   2, indicating the wave of information extracted and the year in which the study was carried out. For the majority of the countries we have selected the wave of the WVS and EVS whose year of execution coincides with the last period of the time series of the decentralized spending. In the cases of Chile, Spain and the Netherlands, we selected the wave that was carried out two years after the last data regarding the decentralization of spending was published, in order to having a longer time series. The seven grouped explanatory variables reflect the medium value of the time series of decentralized spending in education, health, housing, social protection, recreation, culture and religion, public order and safety, and transportation. The FD level of spending is measured as the relation between the spending carried out by subnational (state and local) governments, and the total spending at all three levels of government (central, state and local). Table 3 shows the descriptive statistics of the explanatory variable.
The value of the intraclass correlation coefficient (ICC) of the model 1 shows that .2374 of the changeability of the subjective well-being is explained by the characteristics unrelated to country. Given the individual explanatory variable in the model 2, the variance of the random part diminishes by almost more than half; which explains why the ICC is reduced to .1286. If we compare the statisticians -2 log likelihood, Akaike information criterion (AIC) and Bayesian information criterion (BIC) of both models, we corroborate that Model 2 is appropriate for explaining level of individual satisfaction.
The incorporation of the national explanatory variable in Model 3 contributes to reduce the variance of the random part, if is compared with model 2, which explains why the statistical ICC diminishes. Given the statisticians -2 log likelihood, AIC and BIC show that Model 3 is most adequately explains happiness. In the same way, if we compare the coefficients of the individual explanatory variable from Models 2 and 3, we can confirm that they have the same sign and that the magnitude hardly differs. Therefore, Model 3 is the most adequate for explaining the effect of spending decentralization policies, given that it reflects the heterogeneousness of the countries.
The coefficients estimated for the grouped variables confirm Hypothesis 1. That is to say, the effect of decentralization differs in function of the policies, and is specific to each spending function. The spending parameters in education, health, housing and recreation, culture and religion are significant. The decentralization of spending functions in health and recreation, and culture and religion, make individuals happier, just as Hypothesis 2 predicts. On the contrary, the transfer of responsibility in the areas of in education and housing reduces citizens' satisfaction with these services.
Assuming there is no general consensus with regard to the effect of this variable (Bjørnskov et al., 2008). As the economics of happiness theory predicted, individual income levels positively influence happiness. Another socioeconomic factor that determines the satisfaction level of the population is that of unemployment: those unemployed are less happy than those employed.
As opposed to previous analyses, these findings show that the decentralization of spending in recreation, culture and religion, housing, education and health are the exogenous factors that determine subjective well-being, though the sign of its effect is specific to each spending function. Also, the magnitude of its effect is greater if it is compared with the impact of the individual variables.

CONCLUSIONS
The main objective of this paper, as presented in the beginning, was to cover a prominent "gap" within the research into the economics of happiness: analysis of the influence of public spending decentralization in relation to subjective wellbeing. We specify one multilevel ordinal logit with I random intercept and fixed effects model for two reasons: i) happiness is a latent variable, which we measure through the level of satisfaction declared by individuals, and ii) there are two types of explanatory variables, the individuals, that represents individual characteristics, and grouped, which reflect the spending decentralization in education, health, housing, social protection, recreation, culture and religion, public order and safety, and transportation. The central hypothesis of this work is that decentralization, measured through FD, has a different effect on happiness depending on the nature of the State function through which it is analyzed.
This is the first study in this scientific environment which applies the multilevel analysis that permits us to quantify the influence of the explanatory variable on subjective well-being. On an empirical level, we contribute three significant findings: i) happiness is determined, chiefly, by spending decentralization in recreation, culture and religion, housing, education, and health; ii) the decentralization of the policies governing recreation, culture and religion, and health contributes to greater satisfaction in a country's citizens; and iii) the transfer of spending responsibility in housing and education causes a decrease in subjective well-being. Of particular importance is the case of decentralization of policies regarding recreation, culture and religion, which is one of the exogenous factors that most heavily influences happiness.
In terms of future research, our results suggest that we should turn our efforts to study the feasibility of fiscal decentralization, examining the multiple dimensions of each specific area of spending. For example, education involves the administration of human resources, infrastructure maintenance and improvement, and the management of academic content and teaching methodologies. The same thing is true for health; an area in which the logistical aspects can easily be separated into other elements of public management.
The various spending items included in the category of social protection require similar analysis. The availability of specific data at the national level would make repeating this exercise worthwhile, in order to further break down the information, which may, in turn, prove to be extremely useful in the development of good public policies.
Upon being nested i individuals inside j grouped,  0mj can vary among j groups.
Therefore we can rewrite  0mj thus: (1.7) 1 exp( ) Upon supposing that the effects of X lij and U 0mj are fixed and random, respectively, the following is confirmed:   Calculations have been done using the GLAMM routine.