Search

Many firms want to target their customers with a sequence of marketing actions, rather than just a single action. We interpret sequential targeting problems as a Markov Decision Process (MDP), which can be solved using a range of Reinforcement Learning (RL) algorithms. MDPs require the construction of Markov state spaces. These state spaces summarize the current information about each customer in each time period, so that movements over time between Markov states describe customers’ dynamic paths. The Markov property requires that the states are “memoryless,” so that future outcomes depend only upon the current state, not upon earlier states. We propose a method for constructing Markov states from historical transaction data by adapting a method that has been proposed in the computer science literature. Rather than designing states in transaction space, we construct predictions over how customers will respond to a firm’s marketing actions. We then design states using these predictions, grouping customers together if their predicted behavior is similar. To make this approach computationally tractable, we adapt the method to exploit a common feature of transaction data (sparsity). As a result, a problem that faces computational challenges in many settings, becomes more feasible in a marketing setting. The method is straightforward to implement, and the resulting states can be used in standard RL algorithms. We evaluate the method using a novel validation approach. The findings confirm that the constructed states satisfy the Markov property, and are robust to the introduction of non-Markov distortions in the data.

We evaluate the consistency of two methods for estimating the effect of an economic policy: i) asking people how the policy caused them to change their behavior (reported effects); ii) inferring this change using data on behavior and differences in treatment across people (revealed-preference estimates). Both methods are widely used to measure spending caused by increases in liquidity. Using Federal stimulus payments disbursed quasi-randomly in 2008, we find larger revealed-preference estimates of spending propensities for households who report greater spending responses, and the methods produce similar average propensities. But evidence is mixed on the relationship between spending propensities and liquidity.

Households in the Nielsen Consumer Panel were surveyed about their 2008 Economic Stimulus Payment. In estimates identified by the randomized timing of disbursement, the average household׳s spending rose by 10 percent the week it received a Payment and remained high cumulating to 1.5–3.8 percent of spending over three months. These estimates imply partial-equilibrium increases in aggregate demand of 1.3 percent of consumption in the second quarter of 2008 and 0.6 percent in the third. Spending is concentrated among households with low wealth or low past income; a household׳s spending did not increase significantly when it learned about its Payment., Sloan School of Management, Northwestern University (Evanston, Ill.). Kellogg School of Management, University of Chicago. Initiative for Global Markets, Northwestern University (Evanston, Ill.). Kellogg School of Management. Zell Center for Risk Research, Harvard University. Laboratory for Applied Economics and Policy

We evaluate the consistency of two methods for estimating the effect of an economic policy: i) surveying people to report the change in their behavior caused by the policy, ii) inferring this change using (reported) actual behavior and differences in treatment across people. Both methods have been widely used to measure propensities to spend. Using Federal stimulus payments disbursed quasi-randomly over time in 2008, we find greater revealed-preference estimates of spending by households reporting greater spending and the two methods produce similar estimates of average spending. But, counterfactually, reported preferences estimates are not higher for households with lower liquidity.

This paper evaluates theoretical explanations for the propensity of households to increase spending in response to the arrival of predictable, lump-sum payments, using households in the Nielsen Consumer Panel who received $25 million in Federal stimulus payments that were distributed randomly across weeks. The pattern of spending is inconsistent with models in which identical households cycle through high and low response states as they manage liquidity. Instead, the propensity spend is a persistent household trait. This trait is unrelated to expectation errors, almost unrelated to crude measures of procrastination and self-control, moderately related to measures of sophistication and planning, and highly related to a measure of impatience.

Households in the Nielsen Consumer Panel were surveyed about their 2008 Economic Stimulus Payment. In estimates identified by the randomized timing of disbursement, the average household’s spending rose by ten percent the week it received a Payment and remained high cumulating to 1.5–3.8 percent of spending over three months. These estimates imply partial-equilibrium increases in aggregate demand of 1.3 percent of consumption in the second quarter of 2008 and 0.6 percent in the third. Spending is concentrated among households with low wealth or low past income; a household’s spending did not increase significantly when it learned about its Payment.

We explore household consumption surrounding federal tax returns filings and refunds receipt to test various theories of consumption. Because uncertainty regarding the refund is resolved at filing, precautionary savings theory predicts an increase in consumption at this date. Contrary to this prediction, we find that households generally do not increase consumption at filing. Following the receipt of the refunds, consumption of both durables and nondurables increases dramatically and then decays quickly. Our results show that households, on average, are financially constrained, exhibit myopic behavior, and do not respond to precautionary savings motives.

Under the Economic Growth and Tax Relief Reconciliation Act of 2001, most U.S. taxpayers received a tax rebate between July and September, 2001. The week in which the rebate was mailed was based on the second-to-last digit of the taxpayer's Social Security number, a digit that is effectively randomly assigned. Using special questions about the rebates added to the Consumer Expenditure Survey, we exploit this historically unique experiment to measure the change in consumption expenditures caused by receipt of the rebate and to test the Permanent Income Hypothesis and related models. We find that households spent about 20-40 percent of their rebates on non-durable goods during the three-month period in which their rebates were received, and roughly another third of their rebates during the subsequent three-month period. The implied effects on aggregate consumption demand are significant. The estimated responses are largest for households with relatively low liquid wealth and low income, consistent with liquidity constraints.

This paper evaluates the central insight of the Consumption Capital Asset Pricing Model (CCAPM) that an asset's expected return is determined by its equilibrium risk to consumption. Rather than measure the risk of a portfolio by the contemporaneous covariance of its return and consumption growth - as done in the previous literature on the CCAPM and the pattern of crosssectional returns - we measure the risk of a portfolio by its ultimate consumption risk defined as the covariance of its return and consumption growth over the quarter of the return and many following quarters. While contemporaneous consumption risk explains little of the variation in observed average returns across the Fama and French 25 portfolios, ultimate consumption risk at a horizon of three years explains a large fraction of this variation.

This paper uses the consumption Euler equation to derive a decomposition of consumption growth into four sources. These are new information and three sources of predictable consumption growth: intertemporal substitution, changes in the preferences for consumption, and incomplete markets for consumption insurance. Using data on the expenditures of households, we implement the decomposition for the average growth rate of consumption expenditures on nondurable goods in the U.S. from the beginning of 1982 to the end of 1997. Incomplete markets for trading consumption in future states lead to statistically significant and countercyclical movements in expected consumption growth: consumption growth is expected to be higher when the unemployment rate is high. The economic importance of precautionary saving rivals that of the real interest rate, but the relative importance of each source of movement in the volatility of consumption is not precisely measured.

This paper evaluates the equity premium using novel data on the consumption of luxury goods. Specifying utility as a nonhomothetic function of both luxury and basic consumption goods, we derive pricing equations and evaluate the risk of holding equity. Household survey and national accounts data mostly reflect basic consumption and therefore overstate the risk aversion necessary to match the observed equity premium. The risk aversion implied by the consumption of luxury goods is more than an order of magnitude less than that implied by national accounts data. For the very rich, the equity premium is much less of a puzzle.