In this paper we propose a new variance estimator for OLS as well as for nonlinear estimators such as logit, probit and GMM. This variance estimator enables cluster-robust inference when there is two-way or multi-way clustering that is nonnested. The variance estimator extends the standard cluster-robust variance estimator or sandwich estimator for one-way clustering (e.g. Liang and Zeger (1986), Arellano (1987)) and relies on similar relatively weak distributional assumptions. Our method is easily implemented in statistical packages, such as Stata and SAS, that already offer cluster-robust standard errors when there is one-way clustering. The method is demonstrated by a Monte Carlo analysis for a two-way random effects model; a Monte Carlo analysis of a placebo law that extends the state-year effects example of Bertrand et al. (2004) to two dimensions; and by application to two studies in the empirical public/labor literature where two-way clustering is present.

Microeconometrics researchers have increasingly realized the essential need to account for any within-group dependence in estimating standard errors of regression parameter estimates. The typical preferred solution is to calculate cluster-robust or sandwich standard errors that permit quite general heteroskedasticity and within-cluster error correlation, but presume that the number of clusters is large. In applications with few (5-30) clusters, standard asymptotic tests can overreject considerably. We investigate more accurate inference using cluster bootstrap-t procedures that provide asymptotic refinement. These procedures are evaluated using Monte Carlos, including the much-cited differences-in-differences example of Bertrand, Mullainathan and Duflo (2004). In situations where standard methods lead to rejection rates in excess of ten percent (or more) for tests of nominal size 0.05, our methods can reduce this to five percent. In principle a pairs cluster bootstrap should work well, but in practice a wild cluster bootstrap performs better.

We study the effect of the level of inequality in society on individual well-being using a total of 123,668 answers to a survey question about ‘‘happiness’’. We find that individuals have a lower tendency to report themselves happy when inequality is high, even after controlling for individual income, a large set of personal characteristics, and year and country (or, in the case of the US, state) dummies. The effect, however, is more precisely defined statistically in Europe than in the US. In addition, we find striking differences across groups. In Europe, the poor and those on the left of the political spectrum are unhappy about inequality; whereas in the US the happiness of the poor and of those on the left is uncorrelated with inequality. Interestingly, in the US, the rich are bothered by inequality. Comparing across continents, we find that left-wingers in Europe are more hurt by inequality than left-wingers in the US. And the poor in Europe are more concerned with inequality than the poor in America, an effect that is large in terms of size but is only significant at the 10% level. We argue that these findings are consistent with the perception (not necessarily the reality) that Americans have been living in a mobile society, where individual effort can move people up and down the income ladder, while Europeans believe that they live in less mobile societies.

Abstract not found

Please do not reproduce or quote without the authors ’ permission. Comments are welcome.

This paper examines the impact of acquisitions on the subsequent innovation performance of acquiring firms in the chemicals industry. We distinguish between technological acquisitions, acquisitions in which technology is a component of the acquired firm’s assets, and nontechnologi-cal acquisitions: acquisitions that do not involve a technological component. We develop a framework relating acquisitions to firm innovation performance and develop a set of measures for quantifying the technological inputs a firm obtains through acquisitions. We find that within technological acquisitions absolute size of the acquired knowledge base enhances innovation performance, while relative size of the acquired knowledge base reduces innovation output. The relatedness of acquired and acquiring knowledge bases has a nonlinear impact on innovation output. Nontechnological acquisitions do not have a significant effect on subsequent innovation output. Copyright Ó 2001 John Wiley & Sons, Ltd. In this paper we examine the impact of acqui-sitions on the subsequent innovation performance of acquiring firms. Studying the impact of acqui-sitions on postacquisition innovation performance is important from at least three perspectives. First, this evaluation is important from the perspective of organizational learning and innovation, and helps us understand how organizations absorb and use external knowledge. Firm-level theories of technical change suggest that a firm’s inno-vativeness is an outcome of increases in its

The results of this study provide insight into why some universities generate more new companies to exploit their intellectual property than do others. We compare four different explanations for cross-institutional variation in new firm formation rates from university technology licensing offices (TLOs) over the 1994-1998 period – the availability of venture capital in the university area; the commercial orientation of university research and development; intellectual eminence; and university policies. The results show that intellectual eminence, and the policies of making equity investments in TLO start-ups and maintaining a low inventor’s share of royalties increase new firm formation. The paper discusses the implications of these results for university and public policy. 2 I.

This paper explores a class of empirical Bayes methods for leveldependent threshold selection in wavelet shrinkage. The prior considered for each wavelet coefficient is a mixture of an atom of probability at zero and a heavy-tailed density. The mixing weight, or sparsity parameter, for each level of the transform is chosen by marginal maximum likelihood. If estimation is carried out using the posterior median, this is a random thresholding procedure; the estimation can also be carried out using other thresholding rules with the same threshold. Details of the calculations needed for implementing the procedure are included. In practice, the estimates are quick to compute and there is software available. Simulations on the standard model functions show excellent performance, and applications to data drawn from various fields of application are used to explore the practical performance of the approach. By using a general result on the risk of the corresponding marginal maximum likelihood approach for a single sequence, overall bounds on

In this paper we present a technique for extending generalized linear models (GLM) to the situation where some of the predictor variables are observations from a curve or function. The technique is particularly useful when only fragments of each curve have been observed. We demonstrate, on both simulated and real world data sets, how this approach can be used to perform linear, logistic and censored regression with functional predictors. In addition, we show how functional principal components can be used to gain insight into the relationship between the response and functional predictors. Finally, we extend the methodology to apply GLM and principal components to standard missing data problems.

Abstract not found