Richard A. Berk

Applied time series analysis for the social sciences

McCleary and Hay have made time series analysis techniques -- the Box-Jenkins or ARIMA methods -- accessible to the social scientist. Rejecting the dictum that time series analysis requires substantial mathematical sophistication, the authors take a clearly written, step-by-step approach. They describe the logic behind time series analysis, and its possible applications in impact assessment, causal modelling and forecasting, multivariate time series and parameter estimation.

Public perceptions of global warming

In this paper, we examine the way public opinion responds to the prospect of global warming. In particular, we focus on the publics “willingness to pay” in order to prevent various hypothetical climate scenarios from transpiring. To this end, fractional factorial survey methods are employed with a sample of over 600 residents of Southern California. By and large, the public is able to understand and evaluate rather complicated hypothetical climate scenarios, but the public appreciates some features of climate far better than others. In this context, the contingent valuation estimates we provide, while promising, are clearly not ready of consideration by policy makers.

The deterrent effect of arrest in incidents of domestic violence: A Bayesian analysis of four field experiments.

We examine the deterrent effect of arrest in incidents of spouse abuse. The data are from field experiments conducted in four cities o Milwaukee, Omaha, Dade County (Florida), and Colorado Springs. On the average, arrest is no more effective than other police interventions in reducing new incidents of violence. However, arrest has differential effects on subsequent violence depending on the background of the offender. The implications of the findings for social science theory and public policy are discussed. (Abstract Adapted from Source: American Sociological Review, 1992. Copyright

Varieties of Normative Consensus

The empirical study of consensus has long been impeded by lack of an adequately conceptualized appropriate measurement model. Using the study of social norms as an example, this paper presents such a model based on components of variance. A key feature of the model is an explicit acknowledgment of error variance and its role in generating phenotypical dissensus in empirical data sets. The paper distinguishes among threshold dissensus, a condition in which persons differ in their intensities of adherence to norms, segmented dissensus, in which persons differ in the norms to which they subscribe, and structureless domains, a condition in which individuation in norms applies.

Redefine Statistical Significance

We propose to change the default P-value threshold for statistical significance from 0.05 to 0.005 for claims of new discoveries.

Weighting regressions by propensity scores.

Regressions can be weighted by propensity scores in order to reduce bias. However, weighting is likely to increase random error in the estimates, and to bias the estimated standard errors downward, even when selection mechanisms are well understood. Moreover, in some cases, weighting will increase the bias in estimated causal parameters. If investigators have a good causal model, it seems better just to fit the model without weights. If the causal model is improperly specified, there can be significant problems in retrieving the situation by weighting, although weighting may help under some circumstances.

News media coverage and the epidemiology of homicide

OBJECTIVES This study assessed how newspaper coverage of homicides corresponds to the epidemiology of homicide. METHODS Stories in the Los Angeles Times about homicide (n = 2782) were compared with the homicides that occurred in Los Angeles County from 1990 through 1994 (n = 9442). The generalized linear model assessed how victim, incident, and suspect characteristics related to coverage. RESULTS Even when multiple variables were taken into account, some homicides (those with female, child, or elderly victims; those in which the suspect was a stranger to the victim; those in wealthier neighborhoods) received more coverage and others (those with Black or Hispanic victims or victims with less than a high school education; those committed with nonfirearm weapons; those in which the suspect was an intimate of the victim) received less coverage than expected. CONCLUSIONS Some homicides are more newsworthy than others. Potential implications of not providing the public with representative data are discussed.

Does arrest really deter wife battery? An effort to replicate the findings of the Minneapolis Spouse Abuse Experiment.

In this paper we try to replicate the findings from the Minneapolis Spouse Abuse Experiment (Sherman and Berk, 1984). Using longitudinal data from the criminal justice system on 783 wife-battery incidents, an ex post factor design coupled with a propensity-score analysis reveals that arrests substantially reduce the number of new incidents of wife battery. In addition, the reductions are greatest for batterers whom the police would ordinarily be especially inclined to arrest. Policy and theoretical implications are discussed. (abstract Adapted from Source: American Sociological Review, 1985. Copyright

What a Difference a Day Makes: An Empirical Study of the Impact of Shelters for Battered Women

In this article we evaluate the impact on spousal violence of shelters for battered women. Drawing on formal, hybrid theory from economics and sociology, we predict that shelters will have beneficial effects only for battery victims who are already taking control of their lives. For other women, a shelter stay may in the short run encourage retaliation. Using a two-wave panel of wife-battery victims, we find that our hypotheses are by and large supported. Some policy implications are briefly discussed.

Capitalizing on Nonrandom Assignment to Treatments: A Regression-Discontinuity Evaluation of a Crime-Control Program

Abstract Despite the enormous potential of regression-discontinuity, quasi-experimental procedures, they have to date rarely been used in the evaluation of large-scale social programs. In this article, we report an evaluation of such a program in which a regression-discontinuity analysis is employed. Through a change in legislation, the program in question extended eligibility for unemployment benefits to prisoners after their release from prison. The regression-discontinuity approach proved practical and effective; it revealed that the program cut recidivism rates by 13 percent.