Household savings and consumption decisions
Title: A Mean-Variance Benchmark for Household Portfolios over the Life Cycle
Participant: Claus Munk (FI, CBS)
Description: I embed human capital in Markowitz' one-period mean-variance framework. By solving the Markowitz problem for different values of the ratio of human capital to financial wealth, I can study life-cycle effects in household portfolio decisions. The portfolio derived with this approach is very similar to the optimal portfolio derived in the much more complicated dynamic life-cycle models. An application of this simple method illustrates that young households may optimally refrain from stock investments because a house investment combined with a mortgage is more attractive from a pure investment perspective.
Title: Do Savings Mandates Cause Debt Accumulation?
Participant: Asger Lau Andersen (University of Copenhagen), Henrik Yde Andersen (ECON, CBS) and Søren Leth-Petersen (University of Copenhagen)
Description: In Denmark -- as in many other countries -- defined contribution pension schemes have gained increasing importance. In a recent study Chetty et al (2014) use population-wide Danish data with information about savings and total wealth to show that these savings mandates lead people to increase their total savings, even if they have enough financial wealth that can be adjusted to keep total savings constant. This happens because they respond passively to the savings mandates. In the short run passive responses imply that households adjust spending when they are exposed to savings mandates. However, nothing is known about the long run consequences of savings mandates. Do they lead to increased (mortgage) borrowing so that people can maintain their original consumption level, or do people reach the retirement age with higher levels of net wealth? The purpose of the present project is to answer:
Do people, who save extra because they responded passively to a savings mandate, take out more mortgage debt to maintain their original level of spending, or do they end up with more wealth when they approach retirement? To answer this question, we will make use of Danish administrative registry data with information about jobs, income, wealth and pension savings for the entire Danish population since 1980. To quantify the size of the savings mistake following a savings mandate we will use the event-study design proposed by Chetty et al. (2014). Specifically, we will consider job switches associated with a change in mandated savings rates and show that mandates generate a sharp change in savings that pass-through to total asset accumulation. Changing jobs can be associated with a change in total compensation. It is straightforward to control for this as the tax records hold detailed information about this. The design exploits the longitudinal dimension and massive size of the data, and the institutional structure where mandated retirement savings rates vary significantly across firms because of differences in collective bargaining agreements. In Chetty et al. (2014) we show that total savings rates are completely stable in the period leading up to the job switch and change sharply at the job switch. Unless preferences for savings change sharply at the job switch, this is consistent with the assumption that preferences for saving are orthogonal to job switches.
To test the hypothesis that passive decision makers target spending we measure whether savings mandates generate an increase in total accumulation of assets that can predict the uptake of mortgage debt after the job shift. Because we observe all components of the household budget constraint we are able to quantify the costs (in consumption equivalents) associated with being a passive decision maker. Another possible margin of adjustment is that people end-up retiring at a different point in time compared to people who are not exposed to savings mandates, or to have a different level of spending in retirement. Because labor market activity and complete household balance sheets are observed in the administrative data it is possible to investigate all of these alternative hypotheses.
Reference: Chetty; Friedman; Leth-Petersen; Nielsen; Olsen. 2014; Active vs. Passive Decisions and Crowd-Out in Retirement Savings: Evidence from Denmark; Quar. J. Ec. 129(3).
Title: Essays on Pension Savings (PhD project, together with Danmarks Nationalbank)
Participant: Henrik Yde Andersen (ECON, CBS and Danmarks Nationalbank)
Description: The project is comprised of three self-contained chapters concerned with the interplay between pension saving systems and household saving and debt decisions.
Chapter 1, “Do Tax Incentives for Saving in Pension Accounts Cause Debt Accumulation? Evidence from Danish Register Data", investigates if tax incentives for saving in pension accounts affect individual debt accumulation. Applying a quasi-experimental research design on a Danish 2010 policy that reduced tax incentives for saving in annuity pension schemes, we show significant substitution of savings from retirement accounts to gross debt repayments. We find that for every 1 Danish Krone that retirement savings are reduced about 30 cents go to debt repayments. Taking into account all types of savings, we find full crowd-out. Consistent with previous findings, we document that the effect is driven by a minority, about 23%, who actively rebalance their savings.
Chapter 2, “Mandatory Pension Savings and Household Debt”, asks the question if compulsory pension contributions make homeowners repay less debt or even increase borrowing. Access to individual-level information on pension contributions, bank debt and mortgages allows us to investigate exactly this. We exploit exogenous variation from job changes to identify active changes in mortgage and bank debt repayments. Conditional on increasing mandatory pension contributions when taking up a new job, we argue that decreased debt repayments can be causally linked to increased compulsory pension saving rates.
Chapter 3, “Is there a Housing Wealth Effect”, tests whether unexpected changes in house prices affect individual consumption and savings decision. To do this we use longitudinal survey data with subjective information about current and expected future house prices to calculate unanticipated house price changes. We link this information to high quality administrative records with information about savings in various financial instruments. These data makes it possible to regress spending as well as savings in different types of assets and liabilities on direct measures of anticipated and unanticipated innovations to house prices. Controlling for competing explanations we find that an unanticipated increase in housing wealth that compares to one year’s worth of income leads to an increase in spending (also measured relative to income) of 3-5 percent. The spending increase is followed by a corresponding drop in the following year suggesting that the effect is driven by spending on durable goods. The estimated effect is driven by about 8 percent of the households in the sample who actively refinance their mortgage loan and extract equity to finance the spending increase.
Title: Family Finances: Intra-Household Bargaining, Spending, and Capital Structure
Participant: Arna Olafsson (FI, CBS)
Description: Financial decisions are mostly made jointly within households and in order to understand how households determine, for instance, how much to save, how much to consume and how risky their portfolios should be, which determines how well prepared they are for retirement, we need to further our understanding of the internal financial decision-making process of households. This paper aims to test recent influential theories proposing that differences in preferences of household members lead to agency problems reflected in overspending, indebtedness, and financial fee expenses at the household level. To do so, we use comprehensive transaction-level data from individuals within households. Observing individuals within households over the life cycle gives us a unique opportunity to empirically examine how individual revealed preferences over discretionary spending and debt holdings varies by age and affect spending and indebtedness at the household level. To deal with cndogcneity, we use a fixed effects and instrumental variable approach, which hdps us tackle both self-selection and common-shocks issues. We document that the share of household income received by the spender (or debt-inclined) spouse causally increases discretionary spending ( or household debt), controlling for total household income. Moreover, we estimate individual marginal propensities to consume and link the within-household differences to debt and fee expenses at the household level. Our results arc consistent with individuals having different preferences over spending and using expensive debt and exerting their preferences with the bargaining power that income gives them. Furthermore, we document that individuals' marginal propensity to consume falls over the life cycle.
Title: Household Bargaining, Pensions, and Risk Management Schemes for Families
Participants: Jimmy Martinez-Correa and Mauricio Prado (ECON, CBS)
Description: We are interested on how couples make joint risk management choices (e.g., financial investment, insurance, and saving for pension). Such choices are very difficult to identity in Danish Registry data and that is the reason why we ran a unique laboratory experiment that allowed us to study in detail how couples make joint choices and compare them with individuals that have been randomly matched into pairs. This is a unique opportunity to study what is special about how established couples make choices. The outcome of our analysis will shed light on what makes it easier for households to make joint choices. We have preliminary interesting results that could be used, for instance, to make insurance products more appealing to households.
Our most interesting findings so far are the following:
- Classical theory of insurance tells us that with zero loading people should choose insurance as long as they are risk averse. So people should choose insurance no matter whether they are couples or not. We find that:
a. Pairs of subjects in the experiment, no matter whether they are established couples or individuals randomly matched into pairs, demand individual insurance 60% of the times with no loading. In the experiment, pairs of subjects were given the choice to face individually a risky lottery or individually insure fully or partially against the risk of the lottery. Choices had to be made jointly and in a coordinated manner, so it was not possible that in a given pair of subjects an individual chose no insurance and the other chose to insure. Subjects had to either choose to insure or not to insure in a coordinated manner.
b. However, when there is loading, we observed a significant difference between established couples and individuals randomly matched into pairs. Couples chose more insurance (49% of the times) and non-couples chose less insurance (42% of the times). This difference is statistically significant and implies that couples are more risk averse than randomly matched pairs.
- An equal-split pricing mechanism makes couples behave more “rational” in the classical sense explained in finding #1. Couples tend to choose more insurance in the zero loading treatment when they are offered the equal splitting pricing. This equal splitting mechanism is a simple pricing mechanism that sums the price of insurance for each individual and then divides this total price by two. This means that the price for each individual is exactly half of the total price to insure both persons. When couples are not offered this equal splitting pricing, and instead are offered individual fair pricing to each subject, they tend to choose to insure less. Therefore, there seems to be something special about this equal-splitting mechanism that makes couples behave more rationally and insure more when they are expected to do so according to theory. The equal splitting pricing makes, in some sense, the risk-averse households more “rational.”
We are continuing to analyze the data, and soon we will merge our experimental data to the registry data. This will allow us to study in more depth what are the profiles of households, according to demographics and other registry variables, that are important to explain the preliminary results above.
Title: Housing Decisions under Divorce
Participants: Marcel Fisher (FI, CBS) and Natalia Khorunzhina (ECON, CBS)
Description: We investigate the role of divorce as a risk factor that affects housing decisions and long-term financial consequences in a realistically calibrated life-cycle model. Divorces result in a reduction of homeownership rates, household net worth and increase the risk of poverty during retirement. Given the significant increase in divorces over the last couple of decades, our model proposes new challenges to existing pension systems.
The Consumption-Savings Puzzle and Intra-Household Dynamics
Title: The Consumption-Savings Puzzle and Intra-Household Dynamics
Participant: Arna Olafsson (FI, CBS)
Description: In this paper, we use a very accurate panel of individual spending and income to recover information on expenditure around retirement and how the effects differ by gender. The longitudinal nature of our data allows us to estimate fixed effects models, which help me tackle both self-selection and commonshocks issues. Our results show that total expenditure decreases at the time of retirement and that there is substantial heterogeneity in the retirement response across consumption categories where expenditure on ready-madefood and vehicles is reduced the most. However, we find that the overall decline in spending post-retirement is largely attributable to wives retiring rather than husbands. Households spend less on ready-made-food and more in pharmacies when husbands retire, which is consistent with them retiring due to worse health and a reduction in work-related expenses like ready-madefood during work time. However, when wives retire the household spends less on most consumption categories, which is consistent with women being responsible for a larger share of housework and therefore the expenditure in these categories is only affected when they retire. When looking at single individuals, the response to retirement is much more similar among men and women.
Title: The Pension Gender Gap: Understanding Both Its Composition and Changes over Time
Participants: Moira Daly and Fane Groes (ECON, CBS)
Description: In the first project, we will explore the pension savings behavior of men and women in order to decompose the “pension gap”. Recently the pension gap has been a topic of discussion in the Danish media; Nordea and PFA document that women have saved 64 to 80 percent of what men have. To our knowledge, the determinants of this gap have not been fully investigated. Is this “pension gap” solely explained by lower lifetime earnings or do women tend to manage their savings differently than men? How are these decisions affected by household composition?
In the second project, we will study the effects of a major change in legislation in 2007 that affected how pensions are split between husbands and wives in the event of divorce. We will determine whether or not this change in legislation altered the savings behaviour of married men and women, and if so, determine the immediate consequences of this change of legislation on the stability of families, and subsequently, on children.
Title: The Effect of Savings Commitments on Asset, Debt, and Retirement Decisions: Evidence from Mortgage Run-offs in Danish Registry Data
Participants: Steffen Andersen (FI, CBS), Philippe d’Astous (HEC Montréal), Jimmy Martinez-Correa (ECON, CBS) and Stephen H. Shore (Georgia State University)
Description: This project examines the impact of mortgage run-offs -i.e., when the term of a mortgage comes to an end and monthly payments cease -on labour income and asset accumulation. Mortgage run-offs predictably relax a saving constraint for borrowers who chose mortgage contracts that committed them to effectively save by paying down mortgage principal. We examine registry data on the universe of all Danish individuals whose mortgages were on track to run off between 1995 and 2013. These data include year-end information on labour income and the level of nearly all assets and liabilities. Therefore, we can virtually analysis the complete balance sheet of households.
Preliminary analysis shows that borrowers use 21 percent of the resources previously devoted to mortgage payments to decrease labor income, and use 31 percent to pay down other debts. The labour supply response is limited to those without substantial assets or debts prior to the run-off, while the debt reduction response is limited to (and one-for-one among) those without substantial assets but with other debt prior to the run-off.
Our continued research will shed light on how and for whom forced savings mechanisms affects economic decisions, and therefore help us understand how savings and consumption constraints affect, for instance, retirement decisions, both at the individual and household level.