Subjective Well-Being and the Welfare State: Giving a Fish or Teaching to Fish?

The concepts of ‘subjective well-being,’ ‘happiness,’ and ‘life satisfaction’ are used interchangeably throughout this article. For the purposes of this paper, they all refer to the same underlying concept. This follows the convention used in much of the literature (e.g., Alvarez-Diaz et al. 2010; Bjornskov 2006; Bjornskov et al. 2007, 2008, 2010; Dorn et al. 2006, 2008; Easterlin 2003; Flavin et al. 2009, 2011; Pacek and Radcliff 2008a, b; Radcliff 2001).

There is some debate within the psychological literature regarding the definitions, as well as the exact relationship between, ‘needs’ and ‘wants.’ For an overview, see Diener and Lucas (2000). In this paper, the concepts of ‘need’ and ‘want’ are considered interchangeable, largely because it is argued that the fulfillment of both depends—to varying degrees—upon the material conditions of daily living.

The distinction between ‘active’ and ‘passive’—as defined in this study—turns fundamentally upon the extent to which a given labor market program provides a predominately in-kind or cash benefit. Most passive benefits provide individuals with monetary support (i.e., a cash benefit), while most active benefits—training programs, job rotation/creation schemes, and employment incentives offered to employers and the unemployed—provide individuals with various forms of non-pecuniary support (i.e., an in-kind benefit). As discussed above, many active measures contribute to the acquisition or retention of work-related skills, while passive benefits generally help individuals manage the pecuinary costs of unemployment. However, it is important to note that this does not preclude the possibility that individuals can use passive benefits to independently pursue activities for the purposes of self-improvement (e.g., vocational training, formal education, etc.) and/or reclaiming some of the psychosocial benefits of paid employment (e.g., volunteering, job shadowing, entrepreneurship, etc.). Unfortunately, comparative data on how benefit recipients spend their benefits is neither readily available nor easily gathered. The distinction between ‘active’ and ‘passive’ follows convention in the literature and does not reflect any underlying value judgments shared by the author regarding the relative ‘superiority’ of one form of policy measures or the other.

For criticisms of the Easterlin Paradox, see Stevenson and Wolfers (2008), Veenhoven and Hagerty (2006). See Easterlin (2005) for Easterlin’s reply to Veenhoven and Hagerty.

Clark and Georgellis (2012) and Lucas (2007) also find suggestive evidence that individuals fail to completely adapt to job loss.

Countries include: Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom.

Descriptive statistics for all variables presented in the “Appendix”.

Richard Svensson, Stefan Dahlberg, Staffan Kumlin, and Bo Rothstein, “The QoG Social Policy Data Set.” Unpublished manuscript. http://​www.​qog.​pol.​gu.​se/​data/​datadownloads/​qogsocialpolicyd​ata/​.

The ALMP measure covers expenditures in the following program categories: Training (2), Job rotation and job sharing (3), Employment incentives (4), Supported employment and rehabilitation (5), Direct job creation (6), and Start-up incentives (7). The PLMP measure covers expenditures in the following categories: Out-of-work income maintenance and support (8), and Early retirement (9). For more information on the construction and inclusion criteria for these policy variables, please see Eurostat (2011).

Labor market status is proxied by a series of dummy variables for the following categories: unemployed, disabled, retired, student, and homemaker. An employed individual serves as the baseline. Marital status is another dummy that assumes the value of 1 if the respondent is married, 0 if otherwise. Income status is proxied by a respondent’s answer to a survey item asking them if they are ‘struggling’, ‘coping,’ or ‘living comfortably’ on their present income. Those who replied that they were ‘living comfortably’ on their present income were recoded as being ‘Rich,’ while those who replied that they were ‘struggling’ were recoded as being ‘Poor.’ Individuals who are ‘coping’ on present income constitute the baseline reference group. Gender is proxied by a dummy indicating whether the respondent is male. Age and age-squared are continuous variables corresponding to the respondent’s reported age. Education is a continuous measure that ranges from 1 and 5, whereby higher scores indicate higher levels of educational attainment. A value of 1 indicates that the respondent attained less than lower secondary education, while a value of 5 indicates that the respondent completed tertiary education. Family status is proxied by a dummy variable that assumes the value of 1 if the respondent has at lease one child residing in the same household and 0 if otherwise. Church attendance is a continuous measure that ranges from 1 to 7. The item is reversed coded so that higher values correspond with more frequent attendance of religious services. A value of 1 indicates that the respondent never attends religious services, while a value of 7 indicates that the respondent attends such services daily.

EPL data come the OECD Indicators of Employment Protection. Values range between 0 and 6, whereby higher values indicate more restrictive EPL. See www.​oecd.​org/​employment/​protection for more specific methodological information on the construction of this metric.

Tom Clark and Drew Linzer, “Should I Use Fixed or Random Effects?” Unpublished Manuscript. http://​userwww.​service.​emory.​edu/​tclark7/​randomeffects

Ibid. Clark and Linzer, through a series of Monte Carlo simulation experiments, show that this correlation is neither a necessary nor sufficient condition for ruling out the use of a random-effects design. Instead, the authors contend that the amount of between- versus within-variability of theoretically-interesting variables, the number of units, and the average number of observations per unit condition whether the amount of correlation between unit effects and theoretically-relevant covariates poses a serious risk to the biases (if any) of random-effects estimates.

See Rabe-Hesketh and Skrondal (2008) for the formula used to find the statistics discussed here.

Note that the interaction term between active share and total LMP spending was omitted from the analysis due to multi-collinearity. This is likely caused by the inclusion of additional Level 2 (country-level) controls in the analysis.

This, however, does raise an interesting series of questions about whether passive- versus active- benefits would have the same relative utilities in developing societies. This is a potentially interesting avenue for future research.