کار حالا ، پرداخت بعد؟تجزیه و تحلیل تجربی از تجارت کردن پرداخت حقوق بازنشستگی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|25429||2013||9 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Economic Modelling, Volume 30, January 2013, Pages 835–843
We employ random effects panel data regression methodology to investigate the potential compensating differential between wages and pensions. Using data from the British Household Panel Survey (BHPS) and derived prospective pension variables as calculated by the Institute for Fiscal Studies (IFS), we find no evidence of a trade off and, indeed, some evidence of a small premium. Further analysis finds no significant differences in the results for public and private sector workers, even after controlling for sample selection bias.
According to the theory of compensating wage differentials, competition between firms equalises the overall value of employment packages for a particular type of occupation (Rosen, 1986). Thus, one would expect jobs that offered minimal fringe benefits to offer high wages, and vice versa. The theory of compensating wage differentials has been tested, with mixed results, on a variety of such benefits: health insurance (Currie and Madrian, 1999 and Olson, 2002); maternity benefits (Gruber, 1994); work-related injury and sickness insurance (Gruber and Krueger, 1991); paid vacation leave (Altonji and Usui, 2007). In contrast, relatively little empirical work has been undertaken on the compensating differential between wages and pensions, particularly on UK data. The empirical evidence that does exist is somewhat mixed—see Table A.1 (Appendix A). For North America, compelling evidence of a trade-off is found by Smith (1981), Clark and McDermed (1986), Moore (1987), Montgomery et al. (1992), Gunderson et al. (1992). Less compelling are the findings of Ehrenberg (1980), Schiller and Weiss (1980) and Bulow and Landsman (1985). No evidence of a trade-off is found by Smith and Ehrenberg (1983) or Mitchell and Pozzebon (1989) whilst significant positive relationships are found by Gustman and Steinmeier (1987), Dorsey (1989) and Even and Macpherson (1990). In terms of the UK, Inkmann (2006) finds evidence of a perfectly compensating wage differential using data from the English Longitudinal Study of Ageing (ELSA). Andrietti and Patacchini (2004) find evidence to support the implicit contract argument that male occupational pension participants employed in the private sector earn a positive wage premium only at the beginning of their career. However, once they account for the endogenous sorting of individuals into occupational pension schemes, the magnitude decreases sharply.1 In this paper we investigate the relationship between wage and pension benefits using data from the British Household Panel Survey (BHPS) and derived prospective pension rights variables as calculated by the Institute of Fiscal Studies (IFS). We find no evidence of a trade-off, even after accounting for possible sample selection bias. Indeed, our results allude to the possibility of a small premium. This paper is set out as follows: Section 2 sets out the legislative background to pension schemes in the UK whilst Section 3 discusses the theory of compensating differentials, examining in particular the remuneration trade-off in between current pay and future promised pension benefits. Section 4 outlines our data and empirical methodology. Our results are set out in Section 5 and final comments are collected in Section 6.
نتیجه گیری انگلیسی
5.1. All workers Preliminary investigation of our data revealed that most of the variables exhibited between effects that were larger than those of within effects. This would suggest that there is more variation across individuals than there is within individuals. We therefore estimated our panel regression through a random effects random model, the results of which are set out in Table A.3 (Appendix A). The results from the standard Mincer–Becker type regression Eq. (5) are set out in column (1) of Table A.3 and appear to accord with a priori expectations: higher qualification levels are associated with higher gross pay and experience increases pay at a decreasing rate. All variables are statistically significant at the one percent interval. The annual accrued pension variable, f, is included in our second regression, set out in column (2), which also controls for ethnicity, union membership and gender. The estimate of γ, the coefficient of the ratio of annual accrued present value pension rights to the gross wage, is 0.04, which would appear to reject the theory of compensating differentials. The finding of a premium between pension benefits and pay does not, however, entirely negate hypothesis (7). The theory is traditionally applied for a given level of productivity, but it is possible that fringe benefits may distort behaviour. For example, workers receiving fringe benefits may work more productively and thus earn higher wages than other workers ceteris paribus. 9 There may also be econometric reasons why the trade-off does not hold. Altonji and Usui (2007) recount the “sorry story” for compensating differential studies. More often than not, findings contradicting the theoretical expectations are explained by an omitted variable bias triggered by insufficient observable information on ability. Currie and Madrian (1999) provide further discussion on this. In our specification, the education, qualification and experience variables are all imperfect measures of general and firm specific capital. If the associated measurement error is positively correlated with the fringe benefit variable, then an OLS estimator of the fringe benefits' variable's coefficient, γ, will be biased upwards and may eventually switch sign from negative to positive. Inkmann (2006) is aware that when testing for the occupational pension compensating wage differential, there is an opportunity to test this measurement error. If the gross pension accrual variable is broken down into the component from the employer and the employee, then it may be possible to identify such an error. We follow Inkmann's (2006) approach by creating a measure of employee contributions, C, from a question in wave 11 of the BHPS that asked respondents what percentage of their salary they contribute into their employer's pension scheme. Given the obstinacy of most pension arrangements, we assume that this percentage is the same across the other waves. We then restructured our original wage equation, calculating gross salary, net of these employee contributions (Y = W − C). equation(7) View the MathML sourceln(Y+C)=β0+β1S+β2E+β3E2+γf+ε⇒lnY1+c=β0+β1S+β2E+β3E2+γf+ε⇒lnY=β0+β1S+β2E+β3E2+γf+φc+ε Turn MathJax on where c = C/Y such that the (pure) theory of compensating differentials would imply γ = ϕ = − 1. Column (3) in Table A.3 sets out the results from this regression. The estimate of γ at 0.08 is still not negative but that of φ, with a value of − 1.07, is much closer to − 1 and more in line with the theoretical prediction. 10 One would expect the relationship between wage and pension benefits to depend to some extent on the nature of the pension provision. DB pension schemes allocate more risk to the employer and the nature of their financing makes it harder to adjust the compensating wage differential for every given employee's productivity potential. In DC provision, where accrual is more transparent and where the employer does not need to worry about financing in the same way, adjusting the trade-off for each employee should be more possible. There should therefore be a greater trade-off under DB provision rather than DC provision because, ceteris paribus, DB provision provides more non-monetary risks. Wave 11 of the BHPS provides information on the type of employer pension provision. If we assume that the type of employer pension remains stable across other waves, it is then possible to decompose f into f = fDB + fDC such that: equation(8) lnY=β0+β1S+β2E+β3E2+γ1fDB+γ2fDC+φc+ε.lnY=β0+β1S+β2E+β3E2+γ1fDB+γ2fDC+φc+ε. Turn MathJax on Results from this regression are set out in column (4) of Table A.3 and continue to show a wage–pension premium for both DB and, although to a lesser extent, DC schemes. 5.2. Public sector workers An interesting and hitherto unaddressed issue is the nature of the wage–pension relation amongst public sector workers. It has long been argued that public sector employees have different preferences and work motivations to those of their private sector counterparts (see, for example, Buurman et al., 2009). Some of these differences result naturally from sectoral heterogeneities in the nature of jobs—there is often a social welfare aspect associated with public sector employment which may prove attractive to individuals who have a strong desire to help others or serve the public interest. Furthermore, employment in the public sector is generally more secure than it is in the private sector whilst public sector remuneration tends to be relatively less volatile (Bonin et al., 2007 and Clark and Postel-Vinay, 2009). As a result, one might expect individuals who select themselves into public sector employment to be more risk averse than their private sector counterparts (Bellante and Link, 1981). How these factors impact on the trade-off is illustrated in columns (5)–(8) of Table A.3 where we apply our regression analysis to a restricted sample of public sector employees. The results from our underlying Mincer regression are set out in column (5) and those from our basic wage–pension regression, which focuses only annual pension benefits, are set out in column (6). The latter suggests a larger premium at 0.06 than that found for the whole sample. Incorporating employee contributions into our analysis makes very little change to the coefficient on the pension benefit variable, but employee contributions do exhibit a much smaller trade-off than for the whole sample—see column (7). The results from decomposing the type of pension provision into DB and DC are set out in column (8) and reinforce our previous findings that DB provision is related to a slightly higher premium than DC provision. Our results for public sector should be interpreted with caution since attitudes towards risk may affect both the preferred trade-off between pensions and wages and preferred employment. Individuals who are risk averse may prefer to maintain more money in the form of a pension in order to save for an uncertain future or they may prefer to increase their current holdings of cash. And the same type of individuals may also be attracted to the security of public sector employment, ceteris paribus. To account for potential sample selection bias, we adopt the Heckman (1976) two-step estimation procedure using the inverse Mills' ratio. We follow Hersch and Viscusi (1990), Hersch and Pickton (1995), Viscusi and Hersch (2001) and Brown et al. (2006) in proxying risk averseness by a variable detailing whether or not the respondent smokes.11 The results from this analysis are set out in Table A.4. The underlying selection equation shows that public sector individuals are significantly less likely to smoke than other workers, ceteris paribus. Our results suggest that sample selection bias is an issue but the adjusted regression results do not show significant change for the coefficients of interest from the previous analysis.