Four essays on ageing, health and labour market participation
Not peer reviewed
MetadataShow full item record
This thesis consists of five essays: An introductory essay and four essays within the topic of ageing, health and labour market participation. In the introduction I motivate the research questions, discuss how it relates to empirical economics and summarise each of the four papers.
The first project studies how a reform that changed the monetary incentives to delay retirement affect health and healthcare utilisation. The identification strategy relies on the 2011 Norwegian pension reform that increased the monetary incentives to remain employed for nearly half of the private sector workers at age 62. Before 2011, nearly half of the private sector and the entire public sector had access to early retirement (ER) pension. ER pension embodied certain aspects that could create disincentives to remain employed once reaching the ER eligibility age. First, people who retired with full ER pension continued accumulating pension points as if they had remained employed until the normal retirement age of 67. Second, the combination of earnings and ER pension faced an earnings test that proportionally reduced future pension entitlements. Instead of increasing the age at which individuals could retire with ER pensions, an important aspect of the 2011 pension reform was the introduction of flexible claiming together with employment, and the removal of the earnings test. I exploit these changes in the empirical analysis, and identify the effect of the reform by comparing potential changes in the health and employment of private sector workers, who in the absence of the reform, would have been entitled to the full ER pension, to public sector workers. Public sector workers are suitable as a comparison group, since workers in this sector experienced no change in ER pensions. I use several objective measures of health and healthcare utilisation. These are acute hospitalisations and hospital days following an acute hospitalisations, number of visits to a general practitioner (emergency room or health clinic) together with three diagnoses on cardiovascular, musculoskeletal and psychological issues, and the probability to die by age 64. The results from this paper are twofold. First, I document an average decrease in the probability of full retirement at ages 62–64, by around 10 percentage points, with a corresponding increase in the probability of remaining employed at the same ages of around 8.5 percentage points. Second, I show that these results have no clear side effect on health. The results indicate that there is a reduction in hospital days for the entire sample, and the probability of dying by age 64 for females. However, I find some indications of an increase in the probability of experiencing an acute hospitalisation for higher educated people, and I find an increase in cardiovascular issues among females. I conclude that a time-frame of 2 years leads to modest changes in objective measures of health and healthcare utilisation.
The second project investigate the short term effect of retirement on age. To identify the causal effect of retirement, we employ an regression discontinuity (RD) design. RD exploits institutional settings that determine access to a treatment. The idea is that the treatment (retirement) is determined by a running variable (age), reaching a known threshold (the statutory retirement age) that discontinuously change the probability to retire. The discontinuity gap in health at the cutoff age of 67 identifies the treatment effect. We assess the health effects of retirement at age 67, which is an important policy contribution since current retirement reforms typically aim at increasing the retirement age. We use both survey and administrative data to study the short-term effect of retirement. We belive that our health measures, collectively, will provide important insight into the multidimensional effects of retirement on health. The empirical findings of the paper show that there is a sizeable and positive effect of retirement on physical health. In contrast, we find no effect of retirement on acute hospitalisations or mortality. The results shows that while individuals, in the short term, experience a change in self-perceived heath, this does not necessarily translate into a change in more objective measures of health. We also assess the effect by socioeconomic status. Economic theory predicts that individuals with low socioeconomic status have to rely more heavily on their health as an input to the labour market compared to individuals with higher socioeconomic status. This is exactly what we found when considering the subjective measures of health; in contrast, objective of health mask no such heterogeneity. Altogether, this paper adds to a large body of literature on the relationship between retirement and health. We conclude that while retirees may regard their health better compared to those who are just below the statutory retirement age, this is only informative to the extent that it reflects self-rated health and not objective measures of health.
The third project studies a targeted policy aimed at workers aged 60 in Norway, namely the one-week extra holiday that employees aged 60 and above are entitled to by law. Until 2009, the length of vacation depended on the month of birth in the year an employee turned 60. We exploit this institutional setting in a sharp regression discontinuity design. The probability of receiving an extra week of vacation changed sharply depending on whether a worker was born in August or September. The institutional detail created a unique quasiexperimental setting: only individuals born between January and August were entitled to an extra week of vacation in the same year, whereas individuals born between September and December had to wait until the subsequent year for the extra week. We found that an increase in entitlement to vacation had no effect on sickness absence exceeding 16 days. Moreover, we found a decrease in the number of sick notes as authorised by a physician, but the effect is not robust to different specifications. The subsample estimates show that an increase in entitlement to vacation resulted in a significant decrease in the number of sick notes for women and individuals with high school as the highest level of education attained. For females, the point estimates corresponds to a reduction in sick-notes of 24%, whereas for individuals prone to sickness absence, the point estimates corresponds to a reduction of around 38%. However, turning to cause specific diagnoses of musculoskeletal, cardiovascular and psychological issues, we found no effect. Since we could not actually observe if the individuals actually used their entitled vacation in the year in question, the results are the intention-to-treat effect of being made eligible for treatment. The findings have important policy implications as a growing share of people are nearing retirement, and around 25% on disability insurance in Norway are aged 60-64. Targeted policies that adapt to the needs and preferences of employees as they get older may be of importance to mitigate this problem, but the paper questions the extent of the health-argument of increased vacation at age 60.
The fourth project study the labour market responses for individuals whose spouse experienced a health shock. Serious illness can have adverse consequences for the person with ill health and we study how the other spouse’s cope with such events. The paper’s identification strategy consists of an event study in which we assume that the event (the health shock) is difficult to predict regardless of the presence of any risk factor. We define a health shock by focusing on a particular set of outcomes that are assumed to stem from a major life event. We link these outcomes to the unique administrative data. First, we identify individuals whose spouse passed away due to ischemic heart disease, stroke or a transport accident. Second, we identify individuals whose spouse was admitted to the hospital because of an acute, non-planned admission, as a result of three conditions: myocardial infarction, stroke or congestive heart failure. We find that individuals’ whose spouse experiences a fatal health shock endure a reduction in both earnings and employment. The effect is significant and relatively high for widowers, whose income decreases by around 8%, which is persistent for the next five years after the death of their spouse. We find no effect on widows’ earnings, but both widows and widowers experience a decrease in employment. On average, widows and widowers reduce their employment by 2% and 3%, respectively. We document large flows of liquid assets after the death of a spouse, which potentially offset some of the lost earnings, but we find no clear pattern when analysing the effects by education and age. We find that a spouse’s non-fatal health shock results in no significant effect on income and employment. We do find a drop in income for spouses who experiences a non-fatal health shock, but this does not seem to affect the other spouse’s.