Federal Reserve Bank of Minneapolis: 'Why Does Consumption Fluctuate in Old Age, How Should Government Insure It?'
The report was co-authored by
Here are the excerpts:
Abstract
In old age, consumption can fluctuate because of shocks to available resources and because health shocks affect utility from consumption. We find that even temporary drops in income and health are associated with drops in consumption and most of the effect of temporary drops in health on consumption stems from the reduction in the marginal utility from consumption that they generate. More precisely, after a health shock, richer households adjust their consumption of luxury goods because their utility of consuming them changes. Poorer households, instead, adjust both their necessary and luxury consumption because of changing resources and utility from consumption.
* * *
Introduction
To what extent are households subject to risks and to what extent are they insured by the government, themselves, their family, and their community? Given the importance of this question, it is not surprising that many papers have offered different perspectives on it, in the context of both developed and developing countries.
To answer this question, the vast majority of these papers focus on people of working age and study the effects of income shocks on consumption. The key idea is that income fluctuations that result in consumption fluctuations signal that households are imperfectly insured. While this is a very sensible approach, the population is aging, people across the world live much longer and, as they become older, health shocks assume an increasingly important role. This has two important implications. The first one is that health shocks are an additional important source of risk later in life. The second one is that health shocks might affect both resources and the marginal utility of consumption.
Disentangling the causes of consumption changes due to shocks crucially determines how we should best insure people. Consider, for instance, a person hit by a health shock that generates an income drop. Within this situation, there are two possible scenarios. In one, this person's marginal utility function does not change and any drop in consumption reflects a drop in resources. This can call for additional insurance to alleviate frictions that prevent people from smoothing out income fluctuations: this is the traditional view that has been tested. But under a different scenario, suppose that this person's health shock also reduces his or her marginal utility from consumption, for instance because the person no longer derives the same utility from traveling. This person's consumption might now fluctuate for two reasons. One pertains to the fluctuation in resources. The other one pertains to the fluctuation in the marginal utility caused by a change in heath. It could well be, for instance if this person has a lot of assets, that he or she has no problem smoothing out income fluctuations as desired, and that all fluctuations in consumption come from a change in the utility of consumption resulting from a change in health.
The implications in terms of insurance in these two scenarios are different: if the person has no change in marginal utility from consumption as a result of these health and income shocks but experiences a large drop in consumption, it is desirable to give him or her transfers to smooth consumption and thus marginal utility fluctuations.
In contrast, if health mainly reduces the marginal utility of consumption, giving resources to a person affected by a negative health shock is not an effective policy (a benevolent planner maximizing total utility would allocate less consumption to a person whose marginal utility of consumption has decreased). Thus, optimal insurance depends on why consumption fluctuates.
In this paper, we measure the effects of both income and health shocks on consumption among households over age 65 and we decompose the consumption response to a health shock into its effect on resources (which can come from changes in both income and expenses on medical goods and services, a category that we distinguish from regular consumption), and its effect on the marginal utility from consumption of goods and services.
Our analysis requires observing, for the same household, income, health, and broad-based consumption measures; and such data has been notably difficult to find. To overcome this problem and pair income and health data together with consumption data, we use the
The HRS is a longitudinal panel study that is conducted every other year and is representative of the
To compute our measure of health, we follow Blundell, Britton, Costa-Dias, and French (2017) and use the predicted value of a self-reported health index (the individuals' rating of their health status), regressed over a set of objective measures (dummies for reporting difficulties in activities of daily living and dummies for having certain health conditions, as diagnosed by a doctor). Because consumption data is at the household level, our level of analysis is the household, and we take average health of a household's members as an indicator of the health of that household.
We focus on temporary shocks, which given the frequency of our data refer to changes in health that last at most two years. This allows us to cleanly sidestep the well-known difficulty of disentangling the health and income effects of a permanent health shock from its effect on expected lifespan and bequest motives.
To identify the consumption response to these temporary health and income shocks, we rely on a statistical decomposition method a la Blundell, Pistaferri, and Preston (2008) (BPP). That is, we model the household's health and income as transitory-permanent processes, which can be represented as the sum of a permanent component that evolves as a random walk and of a transitory component that is a moving average. As we show in Appendix B, these assumptions are well supported for our age group. Our empirical strategy is to then instrument the effects of a current change in health with that of a future change in health, which correlates with the transitory part of a current health change but is uncorrelated with its permanent component.
Our method is more general than the original BPP estimator, in that we remain agnostic about how households make their consumption decision. In fact, by focusing only on the pass-through of transitory shocks, the assumption that log-consumption evolves as a random walk can be relaxed (Kaplan and Violante (2010), Commault (2020)). In addition, we do not need to assume that households know that the shock is temporary and they might thus be mistaken about the shocks' duration. Our analysis yields several important and novel findings. First, after age 65 households are subject to large temporary shocks in both income and health. In terms of magnitudes, the variance of the i.i.d. component of income explains 40% of the variance of changes in income, and the variance of the i.i.d. component of health explains 33% of the variance of changes in health (after we detrend all variables from the effect of observed demographic characteristics). The bulk of these shocks cannot be attributed to measurement error for two reasons: first, the HRS has been documented to be of excellent quality /1 and, second, we find that these transitory shocks have a significant impact on households' decision variables.
Second, these transitory shocks to income and health are correlated with each other, and this correlation is statistically significant, confirming that even short-run changes in health affect the resources available to households. Their magnitude implies that a one standard-deviation decrease in health is associated with a 2% decrease in current income.
Third, income shocks affect non-durable consumption but not medical expenses. Our estimated average pass-through of a transitory income shock to consumption is 0.11. This means that a one unit increase in the transitory component of income (corresponding to a 100% increase, that is a doubling, of current income) is associated with an 11% increase in current non-durable consumption. This increase is roughly homogeneous across consumption good categories, with leisure activities responding more strongly. In addition, a positive transitory income shock raises the consumption of (all categories of) necessities among low-wealth households and the consumption of luxuries among high-wealth households.
Fourth, the effect of health shocks are concentrated on leisure activities, car maintenance, and out-of-pocket medical expenses. Our estimates of the pass-through coefficients imply that a one standard deviation decrease in current health is associated with a 9% decrease in leisure activities expenses, a 4% decrease in car maintenance expenses, and a 9% increase in out-of-pocket medical expenses. Across levels of wealth, a positive health shock raises the consumption of necessities and possibly of luxuries (although the latter is less precisely estimated) among low-wealth households and raises the consumption of luxuries among high-wealth households.
To further examine the sources of these consumption responses, we then specify a structural life-cycle model and estimate the respective contributions of the changes in resources and in the utility from consumption. In our model, a household's consumption decisions are the solution of an intertemporal problem in which utility is separable in different consumption categories and medical consumption and the utility derived from each consumption category can depend on the household's current health status. We show that, in this framework, the response of a given category of consumption to a transitory health shock can be written as the sum of the effect of the health shock on resources and on the shape of the utility function during the current period, and that it is possible to separately estimate the coefficients that govern the effect of the change in resources and in health through the dependence of the utility function on health.
Our last main finding is that the effects of a health shock on consumption of non-durable goods mainly come from a change in the utility of consuming them, rather than from the effect of health on income, medical expenses, and other resources; and especially so for wealthier households. More specifically, looking into consumption composition, we find that health has only a small effect on consumption of necessary goods and that this effect comes from the impact of health on resources, rather than its effect on the utility of consuming necessities. In contrast, health shocks significantly impact the consumption of luxury goods, and this effect only comes from the utility of consuming them, rather than from a change in resources driven by a change in health. Splitting our sample into low- and high-wealth households, we find that the effect of health on necessary goods (which is generated by a change in resources) is only present among low-wealth households.
By making several contributions, our paper relates to several important strands of previous literature: the literature studying the impact of economic shocks on key economic outcomes, the literature striving to identify the effects of health on the utility function, the literature on household insurance, and the literature on old age savings and risk. We turn to discussing our paper's contributions in the context of each of these branches of the literature in the next section.
* * *
Conclusions
We show that income and health risks are pervasive in old age and that households over age 65 experience both permanent and transitory income and health shocks.
We also document that even transitory income and health shocks trigger significant consumption responses, and that, in terms of the sign of the response, a decrease in health is associated with a decrease in consumption (and vice-versa). We find important heterogeneity in these responses across consumption categories and levels of wealth. In our overall sample, a negative income shock reduces consumption across a variety of categories of goods, while a negative health shock primarily reduces expenses on car maintenance and leisure activities. More specifically, among low-wealth households, a negative income shock mainly reduces expenses on food and utilities, while a negative health shocks generates a drop in expenses on car maintenance. Among high-wealth households, both negative income and health shocks reduce expenses on leisure activities.
We also develop a life-cycle framework to determine what drives the response of consumption to a transitory health shock. We find that, in the response of total non-durable consumption both the resource channel and the shift in utility due to health status channel are significant. However, the resource channel contributes much less, while the shift in utility explains most of the response of non-durable consumption. Considering the responses of necessities and luxuries separately, we find that on average in the population, the resource channel is only significant for necessities, that is, a change in resources significantly affects the of consumption necessities (and this is driven by low-wealth households) but not that of luxuries. Contrary to that, the shift in utility channel is only significant for luxuries, that is, a change in current health significantly affects the consumption of luxuries through the shift in the utility function that it causes (and this driven by both low-wealth and high-wealth households) but not that of necessities.
* * *
References
Attanasio,
Blundell, R.,
Blundell, R., L. Pistaferri, and I. Preston (2008). "
Blundell, R. and I. Preston (1998). "Consumption Inequality and Income Uncertainty."
Blundell, Richard,
Blundell, Richard,
Braun,
Chung, YoonKyung (2013). "Chronic Health Conditions and Economic Outcomes." working paper.
Cloyne, James,
Cochrane, J. (1991). "A Simple Test of
Commault, J. (2020). "How Does Consumption Respond to a Transitory Income Shock? Reconciling Natural Experiments and Semi-Structural Methods." working paper.
Crawley, E. (2020). "In search of lost time aggregation." Economics Letters 189.C.
De Nardi, Mariacristina,
De Nardi, Mariacristina,
Dobkin, Carlos,
Evans, William N and W Kip Viscusi (1991). "Estimation of State-Dependent Utility Functions Using Survey Data." The Review of Economics and Statistics 73.1, pp. 94-104.
Fagereng, A.,
Farhi, Emmanuel and Iv'an Werning (2013). "Estate Taxation with Altruism Heterogeneity." American Economic Review 103.3, pp. 489-95.
Finkelstein, Amy, Erzo F. P. Luttmer, and
Finkelstein, Amy, Erzo F. P. Luttmer, and
French, Eric, John Jones, and
Golosov, Mikhail, Maxim Troshkin, and Aleh Tsyvinski (2016). "
Gourinchas, P-O. and
Heathcote, J., K. Storesletten, and
Hurd, Michael D. and
Kaplan, G. and
Kopecky, Karen A. and Tatyana Koreshkova (2014). "The Impact of Medical and Nursing Home Expenses on Savings."
Krueger, Dirk and
Krueger, Dirk and
Love, David A.,
Meghir, C. and L. Pistaferri (2004). "Income Variance Dynamics and Heterogeneity." Econometrica 72.1, pp. 1-32.
Meyer, Bruce and
Mian, A.,
Poterba, James,
Poterba, James M.,
Poterba, James M.,
Storesletten, Kjetil,
* * *
REPORT and FOOTNOTES: https://www.minneapolisfed.org/research/institute-working-papers/why-does-consumption-fluctuate-in-old-age-and-how-should-the-government-insure-it
Sens. Shaheen, Hassan to A.G. William Barr: Stop Efforts to Gut Affordable Care Act in Court
Kaiser Permanente: Highest Rated Again by Covered California
Advisor News
Annuity News
Health/Employee Benefits News
Life Insurance News