Un test de la hipótesis de la felicidad en la población de la tercera edad
Un test de l'hypothèse du bonheur dans la population du troisième âge
Alejandro CidI,*; Daniel FerrésII; Máximo RossiIII
IAlejandro Cid es Magíster en Economía y profesor del Departamento de Economía de la Universidad de Montevideo (Uruguay). E-mail: acid@um.edu.uy
IIDaniel Ferrés es Magíster en Economía y profesor asistente del Departamento de Economía de la Universidad de Montevideo (Uruguay). E-mail: dferres@um.edu.uy ]]>
IIIMáximo Rossi es candidato a Doctor en Economía, profesor titular (grado 5) del Departamento de Economía, Facultad de Ciencias Sociales de la Universidad de la República (Montevideo, Uruguay). E-mail: mito@decon.edu.uy
Replicated from Cuadernos de Economía, Bogotá, v.27, n.48, p. 23-45, Jan/June. 2008.
ABSTRACT
We use a rich data set that allows us to test different happiness hypotheses employing four methodological approaches. We find that older people in Uruguay have a tendency to report themselves happy when they are married, when they have higher standards of health and when they earn higher levels of income or they consider that their income is suitable for their standard of living. On the contrary, they report lower levels of happiness when they live alone and when their nutrition is insufficient. We also find that education has no clear impact on happiness. We think that our study is a contribution to the study of those factors that can explain happiness among the elderly in Latin American countries. Future work will focus on enhanced empirical analysis and in extending our study to other countries.
Keywords: Happiness, Health, Family, Censored Econometric Models, Semiparametric Methods, Treatment Evaluation. JEL: C14, C24, I10, J12.
RESUMEN
Se emplea un amplio conjunto de datos que permite evaluar de diferentes formas la hipótesis de la felicidad, empleando cuatro enfoques metodológicos. Se constata que las personas de mayor edad en Uruguay tienen una tendencia a reconocerse felices cuando están casadas, cuando tienen un buen estado de salud y si tienen altos ingresos monetarios o estiman que su ingreso es conveniente para su nivel de vida. Contrariamente, señalan niveles más bajos de felicidad cuando viven solos o cuando su nutrición es insuficiente. Se evidencia que la educación no tiene un impacto claro sobre su percepción de felicidad. Este trabajo es una contribución al estudio de los factores que pueden explicar la felicidad entre las personas de la tercera edad en los países de América Latina. El trabajo futuro se concentrará sobre un análisis empírico mejorado y sobre la expansión del estudio a otros países.
Palabras clave: felicidad, salud, familia, modelos econométricos censurados, métodos semi-paramétricos, evaluación de tratamiento. JEL: C14, C24, I10, J12.
On utilise un ensemble de données très large qui permet d'évaluer des formes différentes d'hypothèse de bonheur sous quatre approches méthodologiques différentes. On constate que les personnes les plus âgées en Uruguay ont une tendance de se reconnaitre heureux quand ils sont mariés, quand ils ont de niveaux de santé plus élèves et quand ils ont plus de revenus monétaires ou ils estiment que leur revenu est convenable pour leur niveau de vie. Au contraire, ils signalent des niveaux plus bas de bonheur quand ils vivent seuls et quand leur nutrition est insuffisante. Nous constatons aussi que l'éducation n'a aucun impact clair sur leur perception du bonheur. Ce travail est une contribution à l'étude des facteurs qui peuvent expliquer le bonheur parmis les personnes âgées dans les pays de l'Amérique Latine. Le travail futur se concentrera sur une analyse empirique améliorée et sur l'extension de l'étude à d'autres pays.
Mot clés : bonheur, santé, famille, modèles économétriques censurés, méthodes semi-paramétriques, évaluation de traitement. JEL : C14, C24, I10, J12.
Fresh interest among economists in using surveys of reported well being as a way to measure individual utility and its relation to a range of economic and social phenomena provides a new tool to understand what causes happiness.
Happiness indicators have been defined in different ways in the economics literature (see Layard, 2005 or Argyle, 2002 for a reflective summary). A large body of research on happiness in economics takes reported subjective well-being as a proxy measure for utility. Various studies are based on surveys that contain the following question: "How satisfied are you with your life?" In our work we follow this approach and we define "happiness" as satisfaction with life in general. Based on the analysis of survey data on subjective wellbeing, our work is guided by the question: "how does x affect happiness?", where x can be income, health, marital status or employment status.
It is hard to make comparisons of happiness levels accross countries. As Diener and Suh (2000) state, although researchers can empirically study quality of life and make comparisons of the subjective well-being of societies, it is unclear if we will be ever able to conclude in a definitive way that one society is better than another in terms of overall quality of life. In the end, happiness and other indicators of quality of life depend on idiosyncratic values and judgements. Still, survey data allows us to make comparisons and draw conclusions about which societies have greater subjective well-being.
Different relationships between happiness and specific variables have been explored in recent economic work. In particular, various scholars have devoted a good amount of effort trying to assess the relationship between income and happiness. This issue is particularly attractive to many people for one reason: there is vast evidence indicating that differences in income explain only a low proportion of the differences in happiness among persons. Moreover, although many countries have experienced strong increases in their per capita GDP, it is not generally true that these countries have seen average happiness rise (Blanchflower and Oswald 2004, Layard 1980). This observation is particularly true for the cases of the US, the UK, Japan and Belgium.
Along these lines, Oswald (1997) finds that happiness with life appears to be increasing in the United States throughout the 1970s, 1980s and 1990s. However the rise is so small that it seems extra income is not contributing dramatically to the quality of people's lives. Also he finds that since the early 1970s reported levels of satisfaction with life in the European countries have on average risen very slightly. In addition, Argyle (2002) provides a good summary of the "income and happiness" relationships and he obtains contradictory results. For example, he observes that rising incomes have not affected life satisfaction and that winning lotteries has negative effects for some. Also, he finds that often the rich are not happier than those with middle income. On the other hand, Argyle observes that the very poor are less happy, and richer countries have higher levels of reported happiness than poorer countries. Scholars, puzzled by these surprising observations, have worked on coming up with new hypotheses to try to explain subjective well-being. In particular, recent work has focused on testing the relevance of inequality, relative income and income aspirations when trying to understand what causes happiness.
]]> Alesina et al. (2003) studied the effect of income inequality in society on individual well-being. In their work, they found that "individuals have a lower tendency to report themselves happy when inequality is high, even after controlling for individual income". They compared results obtained for European countries and the United States.1 Interestingly, their results are clearly different across socioeconomic groups in Europe and the US. In particular, they found that in Europe the poor and those on the left of the political spectrum become unhappy as inequality grows. On the other hand, in the US, the happiness of the poor and of those on the left is uncorrelated with inequality.Frey and Stutzer (2003) tested different happiness hypotheses. In particular, they conducted an empirical test of the role of income aspirations. Their idea is based on the observation that many people compare themselves to those that are considered their others. In the past, many economists have explored this idea when trying to understand different socioeconomic phenomena.
Frey and Stutzer concluded that "the evidence presented indicates that people's well-being is better understood when their income aspirations are taken into consideration". Layard (1980) observed that happiness depends on income and status relative to what you expected it to be. An obvious problem with high income is that you get used to it. Richer people take high income for granted and cannot do without it. The same is true for status. Layard also stated that this could explain why people fight much harder against cuts in their income than for increases in their earnings.
Clark and Oswald (1994) analyzed the impact of unemployment on happiness using data from the British Household Panel Study (1991). In their work, they constructed a "caseness score" using 12 questions present in the survey. After controlling for specific individual characteristics, they utilized ordered probit estimation in order to explore the relationship between unemployment and mental well-being. They concluded that there is a strong negative relationship between these variables. Moreover, they observed that the effect of unemployment on well-being can be stronger "than any other single characteristic, including important negative ones such as divorce and separation".
Other economists have examined the relationship between happiness and different individual variables. Stack and Eshleman (1998) analyzed the relationship between marriage and happiness in a multi-country study. In particular, they observed that the positive relationship between being married and happiness indicators held for 16 of the 17 cases analyzed.
Health status is another factor that can be expected to be an especially important determinant of happiness. Gerdtham and Johannesson (1997) analyzed the relationship between happiness and health status based on data on a sample of 5,000 individuals in the Swedish adult population. In their study, they found a positive and statistically significant relationship between higher health status and happiness.
Various studies focus on the relationship between aging and happiness. Oswald (1997) and Cruz and Torres (2006) found that the relationship between happiness and age is U-shaped. While Oswald found that happiness indicators in Europe reach the minimum levels at age around 30, Cruz and Torres find that for the case of Colombia, the happiness curve decreases as it approaches to age 40; then it becomes a growing function. In our research we have focused on a sample of individuals of age 60 or more. As we will indicate in the results section, we get no robust "age" effect in our investigation.
]]> So far, most of the research on the relationship between individual characteristics and happiness has focused on industrialized countries. It is evident that factors affecting satisfaction with life may vary from region to region. The impact of income or family composition on happiness can be very much related to cultural issues. Recent studies have focused on happiness analysis in Latin America. Interestingly, Graham and Felton (2005) analyzed the effect of income inequality on happiness across Latin American countries. Their work is based on data gathered in Latinobarometro. Also, Gerstenbluth et al. (2007) studied the relationship between happiness and health in Argentina and Uruguay using the Latinobarometro survey. Bucheli (2003) focused on analyzing happiness issues among Uruguayan women in the age range between 25 and 54 and Cruz (2006) tested various happiness hypotheses among Colombians.
Our work represents a fresh attempt to understand the factors that may be related to a higher satisfaction with life in Uruguay, a Latin American country. In particular we will explore the correlation between happiness and income, family structure and health.
Correlations do not establish causation. In this sense, we understand that a crucial aspect of our future work will be related to trying to understand the way in which causality goes. A happiness function assumes that the right hand variables determine the level of the dependent variable. In the case of our study, we are aware that there may also be a reverse causation. For example, are happy people more likely to be married or is it that marriage causes happiness? In order to explore and deal with this selection bias we employ the propensity score technique.
The rest of the paper continues as follows. In section 2 we describe the data set and different happiness indicators. In section 3 we deal with multiple methodological aspects of our work. In section 4 we present the obtained results. In section 5 we present the p-score results. In section 6 we conclude.
DATA AND HAPPINESS INDICATORS
Data
Our analysis of the determinants of happiness in Uruguay relies on data from a multicountry survey called Salud, Bienestar y Envejecimiento en América Latina y el Caribe (SABE), a study sponsored by the Pan American Health Organization (PAHO)2. Since the survey is limited to the single largest city in each country, we focus on information for Montevideo (1,444 observations). SABE data was collected in 1999-2000.
Since the survey gathers information about the elderly, the sampling frame limits its scope to those 60 and older. Individuals living in institutions, such as nursing homes and mental institutions are excluded from the sample. Table 1 presents descriptive statistics of both dependent and independent variables.
]]>
Independent variables include indications of age, gender, family structure, education, health status, employment status and income. Information on these variables is present on SABE, except for income.5 The income variable is a constructed variable, obtained after extrapolating data from the Encuesta Continua de Hogares. Our approach leads to a new indicator for individual income level (see Appendix A for details) and is different from the analysis of Graham and Felton (2005) who constructed an "asset index" based on household possessions.
Table 2 presents mean values for the independent variable among the happy and the unhappy.
Happiness Indicators
]]> Our objective is to test how individual's judgment of well-being is affected by a group of individual characteristics and socioeconomic variables. We follow two paths when defining the dependent variable. Constructing two types of "happiness" indicators will allow us to conduct more robust econometric analysis about the impact of specific variables on happiness. We believe that this issue constitutes a strong aspect of our estimation approach.First, we construct a dummy variable indicating "satisfaction with life". This variable is constructed based on the following question: "In the last two weeks: have you been satisfied with your life?" Respondents can answer "yes" or "no". We use this binary variable in a probit estimation. Also we build an index of happiness based on 15 binary responses to questions related to life satisfaction (for each question, a 0 is assigned to "No" and 1 to "Yes"). Thus, this index takes the integer values from 0 to 15, where superior values mean greater life satisfaction. We used this definition of happiness when conducting OLS analysis. Finally, we expressed this index in percentage terms in order to use it in the semiparametric model.
Table 3 presents descriptive statistics about the constructed happiness indicators.
Income and Happiness
As we said, the relationship between income and happiness can be analyzed from several different points of views. Economists have focused on issues such as the relationship between (a) absolute income and happiness; (b) relative income and happiness; (c) income inequality and happiness; (d) income aspirations and happiness.8 There is sufficient evidence that absolute income, alone, does not play a substantial role in explaining happiness levels. In our work we will consider income as an independent variable, but also relative income and income aspirations.
Broadly speaking, relative income is defined as the difference between individual income and the average income for the reference group. In our work we take the following approach: we include a variable indicating the income percentile to which the respondent belongs.9 Income aspirations information is collected from the following question: "Do you think that you (and your partner) have enough money in order to cover your daily expenses?"
]]> Family and HappinessIn a context of rapid transformation of typical family structures we attempt to understand the effects of changes in family composition on happiness. In this sense, since our data set focuses on the elderly, it provides a unique opportunity to assess the long term impact of divorce and remarriage on individual happiness.
There is vast evidence about the negative impact of divorce on life satisfaction. Again, most of this evidence is reflected by data related to industrialized countries. Our dataset allows us to investigate the impact of marriage and divorce in the Latin American region. We know that our dataset restricts our attention to those that were 60 or older in 1999-2000. In issues related to moral related values, it is definitely interesting to compare our results to other studies that may contain information for younger cohorts.
Health status and Happiness
In our work we analyze the impact of health in both absolute and relative terms. In particular we constructed two different variables: one that indicates the self reported health condition and another that expresses respondents' opinions about individual health compared to other people in their age group. The intuition for taking both variables into account is that working with both absolute and relative terms will enhance our understanding of happiness levels.
ESTIMATION
We follow four different strategies because we understand that by proceeding in this way we add robustness to our analysis. We believe that each of the techniques that we use presents a potential advantage:
Ordinary Least Squares Estimation10
We run an OLS regression where a "happiness index" is the dependent variable. This particular model estimation presents a major advantage: it is very intuitive and it has a straightforward interpretation. On the downside, we are aware that the index is built based on answers to 15 questions (point values range from 0 to 15, where superior values indicate greater life satisfaction). Defined in this way, "Happiness" could be seen as a doubly censored variable which takes on the value zero and fifteen with positive probability. In other words, the dependent variable suffers from interval censoring and OLS could provide inconsistent estimators. Other shortcomings of the linear probability model are: a) predicted values for "Happiness" could be negative or greater than fifteen; b) the variance of "Happiness" is probably heteroskedastic; c) E(Happiness|x) is nonlinear.
]]> ProbitIn our study, we define a dummy variable that takes the value of 1 when individuals express satisfaction with life. Both logit and probit models are suitable to analyze the link between independent variables and the "satisfaction with life" variable. Probit may be a more appropriate choice for the case in which normal distribution of the dependent variable can be assumed.
Tobit
Because the dependent variable suffers from interval censoring, we also applied a Tobit Model. We take into account that heteroskedasticity and nonnormality result in the Tobit estimator being inconsistent.
A Semiparametric Censored Regression Model
As mentioned above, Tobit models require some specifications of the error distribution: normality and homoskedasticity. In order to relax these requirements, the semiparametric approach has been proposed in the recent economic literature to provide consistent estimates for censored data. Thus one of the advantages of semiparametric models for censored models is that estimators are consistent under weaker distributional assumptions. The attribute "semiparametric" in this model comes from the fact that the distribution of the errors given the explanatory variables does not have a known parametric form. In this work we present results for the symmetrically censored least squares (SCLS) estimator.
The symmetrically censored least squares (SCLS) approach was proposed by Powell (1986). This estimator is based on the assumption that errors are symmetrically (and independently) distributed around zero, so it is less restrictive than Tobit requirements (normally distributed and homoskedastic errors). The SCLS estimators are consistent and asymptotically normal for a wide class of symmetric error distributions with heteroskedasticity of unknown form (for a summary, see Chay and Powell, 2001, or Cameron and Trivedi, 2005).
Powell (1986) states that if the underlying error terms were symmetrically distributed about zero, and if the latent dependent variables were observable, classical least squares estimation would yield consistent estimates. But due to the censoring, the observed dependent variable y has an asymmetric distribution. Powell's approach consists in symmetrically censoring the dependent variable y (it is usually known as a "symmetric trimmed" method) so that symmetry can be restored, and then the regression coefficients can be estimated by least squares. Symmetric censoring of the dependent variable implies that observations with values above the censoring point are dropped, and this means that there could be a loss of efficiency due to the information dropped in those observations. However this problem is reduced in the present paper because a relatively large sample is used.
RESULTS
]]> Table 4 presents results for the four model estimations. We present results for men and women separately.
Obtained results indicate that:
In sum, we find that our results are pretty much in line with those obtained by other studies but in this case for a non-industrialized country. Individuals who have higher health levels, are or feel richer and are married show higher levels of satisfaction with life. We also find some evidence showing that malnutrition and living alone is negatively related to happiness.
TREATMENT EVALUATION AND MARITAL STATUS
The typical dilemma in treatment evaluation involves the inference of a causal association between the treatment and the outcome. In this paper, we pay particular attention to the effects of personal marital status on happiness. Thus, we observe (yi,xi,Di), i=1,/img/revistas/s_ceco/v1nse/.,N, where yi is the happiness index, xi represents the regressors, and Di is the treatment variable and takes the value 1 if the treatment is applied (got married) and is 0 otherwise. The impact of a hypothetical change in D on y, holding x constant, is of interest. But no individual is simultaneously observed in both states. Moreover, the sample does not come from a randomized social experiment: it comes from observational data and the assignment of individuals to the treatment and control groups is not random. Hence, we estimate the treatment effects based on propensity score. This approach is a way to reduce the bias by performing comparisons of outcomes using treated and control individuals who are as similar as possible (Becker and Ichino 2002). The propensity score is defined as the conditional probability of receiving a treatment given pre-treatment characteristics:
]]> p(X)=Pr{D=1|X}=E{D|X}
where D={0,1} is the indicator of exposure to treatment and X is the vector of pre-treatment characteristics.
The propensity score was estimated in this application using a Probit model12. Since the probability of observing two units with exactly the same value of the propensity score is in principle zero since p(X) is a continuous variable, various methods have been developed in previous literature (for a summary, see Cameron et al. 2005) to match comparison units sufficiently close to the treated units. In the present paper, after estimating p(X) we employed the Kernel Matching method.13
The tables below show the result:
]]>
In the case of men, though the "Average Effect of Treatment (got married) on the Treated" is positive at a 90 percent, the 95 percent confidence interval includes zero. In the case of women, the point estimates indicate that being married increases happiness and it is significantly different from zero. Thus, data suggest positive association between being married and happiness, especially in the case of women above the age of 59.
As we said in the beginning of this section, the matching method attempts to made comparisons between treated and control individuals who are as similar as possible. Thus, in order to gauge the goodness of the matching, we built the tables below. The similarity between the treated and control individuals can be seen in the mean comparison test (t-test) shown on the table: there is no statistically significant difference in the characteristics of the treated and control matched individuals.
CONCLUSION
We performed empirical analysis in order to test various happiness theories on a group of older people in a Latin American country. In particular, we analyzed data from Uruguay gathered by SABE.
We find that older people in Uruguay have a tendency to report themselves happy when they are married, when they have higher standards of health and when they earn higher levels of income. However, the relationship between income and happiness is far stronger in the case of women than when men are asked. When we analyze the impact of health and income on happiness we include variables indicating absolute and relative indications. Results indicate that accounting for relative positions improves our understanding of those factors affecting happiness. This implies that individuals often compare themselves with their reference groups.
Individuals report lower levels of happiness when they live alone and when their nutrition is insufficient. In the case of nutrition, we included a variable indicating malnutrition while the individual was a child and also a dummy variable signaling whether the person eats one meal a day or less. We also find that education has no clear impact on happiness.
]]> Obtained results are robust to different methodological strategies. Most observed relationships are consistent with those present in the literature. In this sense, our work is a contribution to exploring those factors that affect individual happiness in Latin American countries, with a special focus on the elderly and their particularities.Our study presents various limitations. Our future efforts will focus on three aspects: 1) to extend analyses to additional countries (Brazil, Argentina, Chile, and Mexico); 2) to incorporate additional semiparametric analysis of the relationships; and 3) to incorporate enhanced analysis of endogeneity.
FOOTNOTES
1 For the US, they present data by state.
2 The survey includes information for Argentina, Barbados, Brazil, Chile, Cuba, Mexico and Uruguay.
3 Health takes the rank of values from 2 to 8, where superior values indicate worse health.
4 Compared Health takes the values 1, 2 and 3, where superior values indicates worse health subjectively compared with other people of similar age.
5 Although SABE has an "Income" chapter, data on income is rather incomplete in the Uruguayan survey.
6 Health takes the rank of values from 2 to 8, where superior values indicate worse health.
]]> 7 Compared Health takes the values 1, 2 and 3, where superior values indicates worse health subjectively compared with other people of similar age.8 Income aspirations reflect people's perception about them having enough money for paying their daily expenses. Clearly, there is an objective, but also a subjective component in this perception.
9 We do this to avoid difficulties in defining "reference groups".
10 In the empirical application of this paper, we use robust standard errors in OLS, Probit, and Tobit models to cope with the possible existence of heteroskedasticity.
11 We only capture the effect of current marital status. Thus, our interpretation refers to whether the individual is married today or not.
12 Applied with the Stata ado file "pscore" developed by Becker and Ichino (2002).
13 This matching method was applied using the Stata ado file "psmatch2" developed by Leuven and Sianesi (2003).
14 This Mean Comparison Test (t-tests for equality of means in the treated and non-treated groups, both before and after matching) was applied using the Stata ado file "pstest" developed by Leuven and Sianesi (2003).
REFERENCES
]]>1. Alesina, A.; Di Tella, R. and MacCulloch, R. (2004). "Inequality and happiness: are Europeans and Americans different? " Journal of Public Economics, 88 (2004): 2009-2042.
2. Argyle, M. (1997). The Psychology of Religious Behaviour, Belief and Experience. London - New York: Routledge.
3. Argyle, M. (2002). The Psychology of Happiness. London - New York: Routledge.
4. Becker, S. and Ichino, A. (2002). "Estimation of average treatment effects based on propensity scores". The Stata Journal, 2 (4): 358-377.
5. Blanchflower, D. and Oswald, A. J. (2004). "Well-being over time in Britain and the United States". Journal of Public Economics, 88 (7-8): 1359-1386.
]]>6. Bucheli, M. and Rossi, M. (2003). "El grado de conformidad con la vida: evidencia para las mujeres del Gran Montevideo." Documento N° 10/03. Departamento de Economía. Facultad de Ciencias Sociales. Universidad de la República.
7. Clark, A. E. and Oswald, A. J. (1994). "Unhappiness and Unemployment". The Economic Journal, 104 (424): 648-659.
8. Cameron, A. C. and Trivedi, P. K. (2005). "Microeconometrics. Methods and Applications". Cambridge: Cambridge University Press.
9. Chay, K. Y. and Powell, J. L. (2001). "Semiparametric Censored Regression Models". Journal of Economic Perspectives. 15 (4): 29-42.
10. Cruz, J. and Torres, J. (2006) "¿De qué depende la satisfacción subjetiva de los colombianos? " Cuadernos de Economía, 25 (45): 131-154.
]]>11. Diener, E. (1985). "The Satisfaction With Life Scale". Journal of Personality Assessment, 49 (1): 71-75.
12. Diener, E. and Suh, E (Editors) (2000). Culture and Subjective Well-Being (Well Being and Quality of Life). Cambridge MA: MIT Press.
13. Frey, B. and Stutzer, A. (2002). "What can Economists learn from Happiness Research? " Journal of Economic Literature, 40 (2): 402-435.
14. Frey, B. and Stutzer, A. (2003). Testing Theories of Happiness. Institute for Empirical Research in Economics. Working Paper Series, 147. University of Zurich
15. Gerdtham, U and Johanesson, M. (1997). "The relationship between happiness, health and socio-economic factors: results based on Swedish micro data". Working Paper Series in Economics and Finance, No. 27. Department of Economics, Stockholm School of Economics, Stockholm, Sweden.
16. Gerstenbluth, M., Rossi, M. and Triunfo, P. (2007). "Felicidad y salud: Una aproximación al bienestar en el Río de la Plata". Documento de Trabajo 15/07. Departamento de Economía, Facultad de Ciencias Sociales, Universidad de la República.
17. Graham, C. and Felton, A. (2005). "Does Inequality Matter to Individual Welfare? An initial exploration based on happiness surveys from Latin America". Working Paper, 38. The Brookings Institution. Center on Social and Economic Dynamics.
18. Honoré, Bo E. and Powell, J. L. (1994). "Pairwise Difference Estimators for Censored and Truncated Regression Models". Journal of Econometrics, 64 (1-2): 241-78, September/October.
19. Layard, R. (1980). "Human Satisfactions and Public Policy". The Economic Journal, 90 (360): 737-750.
20. Layard, R. (2005). Happiness: Lessons from a New Science. London: Penguin Books.
21. Leuven, E. and Sianesi, B. (2003). "PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing". Boston College Department of Economics, Statistical Software Components.
22. Oswald, A. (1997). "Happiness and Economic Performance". The Economic Journal, 107 (445): 1815-1831.
23. Powell, J. L. (1984). "Least Absolute Deviations Estimation for the Censored Regression Model". Journal of Econometrics, 25 (3): 303-25, July.
24. Powell, J. L. (1986). "Symmetrically Trimmed Least Squares Estimation for Tobit Models". Econometrica, 54 (6): 1435-60. November.
25. Stack, S. and Eshleman, J. R. (1998). "Marital Status and Happiness: A 17-Nation Study". Journal of Marriage and the Family, 60 (2): 527-536, May.
26. Rosenbaum, P. and Rubin, D. (1983). "The central role of the propensity score in observational studies for causal effects". Biometrika, 70 (1): 41-55.
Este artículo fue recibido el 11 de diciembre de 2007 y su publicación aprobada el 19 de marzo de 2008.
In our work we deal with a major issue: a high number of no responses to income related questions in the SABE survey. In order to deal with this situation we estimated individual income using data from Encuesta Continua de Hogares (ECH, the Uruguayan household survey). We conducted different estimations for men and for women.
We regressed (the logarithm of) per capita income against a set of individual and socioeconomic variables using ECH data. Our major challenge consisted in selecting those independent variables that we could identify both in the ECH and in the SABE survey. In particular independent variables included indications of age, gender, family composition, educational level, employment status, sources of income and the ownership of different kinds of durable goods. In the case of men, our regression had an R2 of 0.67; in the case of women, R2 was 0.65.
Once we obtained the income estimations from ECH we predicted individual income for the SABE respondents. In our prediction, we utilized those coefficients obtained in our initial estimation in order to express the relationship between individual variables and income levels.
]]>