Researchers have increasingly realized the need to account for within-group dependence in estimating standard errors of regression parameter estimates. The usual solution is to calculate cluster-robust standard errors that permit heteroskedasticity and within-cluster error correlation, but presume that the number of clusters is large. Standard asymptotic tests can over-reject, however, with few (five to thirty) clusters. We investigate inference using cluster bootstrap-t procedures that provide asymptotic refinement. These procedures are evaluated using Monte Carlos, including the example of Bertrand, Duflo, and Mullainathan (2004). Rejection rates of 10% using standard methods can be reduced to the nominal size of 5% using our methods.

Default contribution rates for 401(k) pension plans powerfully influence choices. Potential causes include opt-out costs, procrastination, inattention, and psychological anchoring. Using realistically parameterized models, we show how the optimal default, the magnitude of the welfare effects, and the degree of normative ambiguity depend on the behavioral model, the scope of the choice domain deemed welfare-relevant, the use of penalties for passive choice, and other 401(k) plan features. While results are theory-specific, our analysis provides reasonably robust justifications for setting the default either at the highest contribution rate matched by the employer or—contrary to common wisdom—at zero. (JEL D14, D91, J26, J32)

Even among households with similar socioeconomic characteristics, saving and wealth vary considerably. Life-cycle models attribute this variation to differences in time preference rates, risk tolerance, exposure to uncertainty, relative tastes for work and leisure at advanced ages, and income replacement rates. These factors have testable implications concerning the relation between accumulated wealth and the shape of the consumption profile. Using the Panel Study of Income Dynamics and the Consumer Expenditure Survey, we find little support for these implications. The data are instead consistent with “rule of thumb,” “mental accounting,” or hyperbolic discounting theories of wealth accumulation. (JEL D1, D91, E21)

This paper examines how market prices, volume, and traders' dividend expectations respond to public information releases in laboratory markets for a long‐lived financial asset. The objective is to study deviations from the symmetric information risk‐neutral rational expectations (RE) benchmark, which predicts no trade in such settings. The results of a series of double‐auction and call markets are reported in which traders manage a portfolio of cash and asset shares over 15 rounds of trading. A public signal regarding the value of the liquidating dividend is released every third round, and traders' subjective expectations of the liquidating dividend are elicited each round as cash‐motivated forecasts. We find that, despite the public dividend signal, traders' dividend forecasts are heterogeneous. Forecasts and prices both underreact to the public signals, with prices under‐reacting more than forecasts. In general, price changes are not closely associated with public signals, and there is greater excess price volatility in double auctions than in call markets. Forty‐three percent of trades are inconsistent with the trader's forecasts, and inconsistent trades occur more frequently in the double‐auction markets. On average, approximately 10 percent of the outstanding shares are traded in each round, and trading volume is increasing in the mean absolute forecast revision and decreasing in the contemporaneous dispersion in forecasts. These results suggest that differential processing of the public signal and/or speculative trading for short‐term gain may help to explain why symmetric information RE predictions are often not supported in empirical and experimental settings. They also suggest that market reactions to public information releases may be influenced by market microstructure.

We examine legislative policy making in institutions with two empirically relevant features: agenda setting occurs in real time and the default policy evolves. We demonstrate that these institutions select Condorcet winners when they exist, provided a sufficient number of individuals have opportunities to make proposals. In policy spaces with either pork barrel or pure redistributional politics (where a Condorcet winner does not exist), the last proposer is effectively a dictator or near-dictator under relatively weak conditions. Copyright The Econometric Society 2006.

The Mismatch Between Life Insurance Holdings and Financial Vulnerabilities: Evidence from the Health and Retirement Study by B. Douglas Bernheim, Lorenzo Forni, Jagadeesh Gokhale and Laurence J. Kotlikoff. Published in volume 93, issue 1, pages 354-365 of American Economic Review, March 2003

How Much Should Americans Be Saving for Retirement? by B. Douglas Bernheim, Lorenzo Forni, Jagadeesh Gokhale and Laurence J. Koltikoff. Published in volume 90, issue 2, pages 288-292 of American Economic Review, May 2000

We investigate the impact of peer interaction on the quality of financial decision making in a laboratory experiment. Face-to-face communication with a randomly assigned peer significantly improves the quality of subsequent private decisions even though simple mimicry would have the opposite effect. We present evidence that the mechanism involves general conceptual learning (because the benefits of communication extend to previously unseen tasks), and that the most effective learning relationships are horizontal rather than vertical (because people with weak skills benefit most when their partners also have weak skills). The benefits of demonstrably effective financial education do not propagate to peers.

We examine methods for evaluating interventions designed to improve decision-making quality when people misunderstand the consequences of their choices. In an experiment involving financial education, conventional outcome metrics (financial literacy and directional behavioral responses) imply that two interventions are equally beneficial even though only one reduces the average severity of errors. We trace these failures to violations of the assumptions embedded in the conventional metrics. We propose a simple, intuitive, and broadly applicable outcome metric that properly differentiates between the interventions, and is robustly interpretable as a measure of welfare loss from misunderstanding consequences even when additional biases distort choices. (JEL D83, D91, G51, G53)