Any opinions expressed here are those of the author(s) and not those of the IIIS. All works posted here are owned and copyrighted by the author(s). Papers may only be downloaded for personal use only. Risk, Government and Globalization: International Survey Evidence 1

Divorce is the dissolution of a social tie, but it is also possible that attitudes about divorce flow across social ties. To explore how social networks influence divorce and vice versa, we utilize a longitudinal data set from the long-running Framingham Heart Study. We find that divorce can spread between friends, siblings, and coworkers, and there are clusters of divorcees that extend two degrees of separation in the network. We also find that popular people are less likely to get divorced, divorcees have denser social networks, and they are much more likely to remarry other divorcees. Interestingly, we do not find that the presence of children influences the likelihood of divorce, but we do find that each child reduces the susceptibility to being influenced by peers who get divorced. Overall, the results suggest that attending to the health of one’s friends ’ marriages serves to support and enhance the durability of one’s own relationship, and that, from a policy perspective, divorce should be understood as a collective phenomenon that extends far beyond those directly affected.

Discussion of the substantive impact of a variable on a dependent variable, especially for maximum likelihood models, requires more than reporting the sign and significance of coefficient. Substantive interpretation of MLE models is now practically required for publication in major journals. Postestimation simulation of the model’s parameters allows the analyst to calculate these substantive quantities of interest, and most importantly, it allows the analyst to reflect the degree of uncertainty around those quantities. We present the logic of post-estimation simulation and express its importance for presenting both the estimate and the precision of a substantive quantity of interest. CLARIFY (hereafter) (Tomz, Wittenberg, and King 2003; see also King, Tomz, and Wittenberg 2000) is an easy-to-use post-estimation program that simulates parameters and calculates substantive quantities of interest and the degree of uncertainty around those estimates. can currently be used for linear regression, logit and probit, ordered logit and probit, multinomial logit, count models (Poisson and negative binomial), duration models (Weibull), and seemingly unrelated regressions. Importantly, we emphasize that analysts can move beyond to simulate parameters of interest from a model, and that researchers estimating models outside of the canned models should almost always execute post-estimation parameter simulation.

Legal Instruments on the

This article explores the roots of white support for capital punishment in the United States. Our analysis addresses individual-level and contextual factors, paying particular attention to how racial attitudes and racial composition influence white support for capital punishment. Our findings suggest that white support hinges on a range of attitudes wider than prior research has indicated, including social and governmental trust and individualist and authoritarian values. Extending individual-level analyses, we also find that white responses to capital punishment are sensitive to local context. Perhaps most important, our results clarify the impact of race in two ways. First, racial prejudice emerges here as a comparatively strong predictor of white support for the death penalty. Second, black residential proximity functions to polarize white opinion along lines of racial attitude. As the black percentage of county residents rises, so too does the impact of racial prejudice on white support for capital punishment. States, as Max Weber famously observed, are distinguished in part by their claim to a legitimate monopoly over the use of violence within a given territory (Gerth and Mills 1946, 78). This claim is nowhere more evident, or controversial, than when the state kills an individual it has convicted of a capital crime (Sarat 2001). In a majority of countries, laws permit the state to impose lifelong incarceration, but withhold the authority to take life. At present 109 countries reject the death penalty in law or practice; 86 retain and use the death penalty, but most do not do so with great regularity (AI 2001). No country in Western Europe currently practices capital punishment; and in 1999, the U.N. Commission on Human Rights called for a worldwide moratorium on executions (Dieter 1999). Between 1990 and 2001, over 30 countries abolished this mode of pun-

The existing literature on civic engagement in the United States has focused on face-toface interactions as the key to acquiring both the desire and skills necessary to participate in civil society. However, few works have focused on the relationship between individuals' political talking networks and civic engagement on a nationwide level. One notable exception to this general oversight is Lake and Huckfeldt (1998), who present evidence suggesting a link between the frequency of interaction with political talking partners and civic engagement. Making use of public opinion data from the Cross-National Election Studies' United States study (Beck, et al., 1992), the current study extends the work of Lake and Huckfeldt in two parts. The first assesses the relationship between frequency of interaction in the political talking network and multiple indicators of civic engagement. The second addresses the relationship between attitudinal disagreement in political talking networks and civic engagement. The results suggest a positive relationship between frequency of interaction and civic engagement, and also between the amount of disagreement in the network and civic engagement. The methodological and normative implications of these relationships are discussed.

Although perceived by candidates and parties as important in affecting political outcomes, the link between spending and success in multicandidate, multiparty election campaigns remains unproven. Not only are there relatively few studies of campaign spending effects in multiparty systems, there are none examining the effect under the Single Transferable Vote (STV) electoral system. Our study examines spending effects in the Irish local elections of 1999 using STV with district magnitudes between 3 and 7 seats, contested by a median of 10 candidates in each district. Using detailed data on 1,838 candidates from 180 local constituencies and 30 councils, we provide precise estimates of the relationship between campaign spending at the candidate level in each district and electoral success, including the probability of winning. In a context where spending is miniscule relative to other contexts, and takes place under a completely different electoral system, our results echo previous studies from other contexts showing a strong effect of challenger spending and only weak effects of incumbent spending. Once allowance is made for the endogeneity of incumbent spending, however, we find a much less substantial difference between the effectiveness of spending by challengers and by incumbents, except on the marginal effect of spending on the probability of winning, where challenger spending is shown to be much more important. KEY WORDS � campaign spending � candidate elections � Ireland � STV

While the US steel industry has been in distress for decades, the “steel crisis ” of 1999-2001 was particularly acute. More than 30 steel producing and steel processing firms fell into bankruptcy between 1997 and 2001, and most of the failures occurred after President Bush took office. 1 During his presidential campaign, Bush promised steelworkers that he would not 1. For a list of bankruptcies, see USWA (2002). neglect them. As the crisis worsened, the steel industry and the United Steel Workers of America (USWA) pressed the Bush administration to make good on its campaign promise.

The Normal regression model is a close variant of the more standard least squares regression model (see Section??). Both models specify a continuous dependent variable as a linear function of a set of explanatory variables. The Normal model reports maximum likelihood (rather than least squares) estimates. The two models differ only in their estimate for the stochastic parameter σ. Syntax> z.out <- zelig(Y ~ X1 + X2, model = "normal", data = mydata)> x.out <- setx(z.out)> s.out <- sim(z.out, x = x.out) Additional Inputs In addition to the standard inputs, zelig() takes the following additional options for normal regression: ˆ robust: defaults to FALSE. If TRUE is selected, zelig() computes robust standard errors via the sandwich package (see Zeileis (2004)). The default type of robust standard error is heteroskedastic and autocorrelation consistent (HAC), and assumes that observations are ordered by time index. Examples In addition, robust may be a list with the following options:

Use probit regression to model binary dependent variables specified as a function of a set of explanatory variables. For a Bayesian implementation of this model, see Section??. Syntax> z.out <- zelig(Y ~ X1 + X2, model = "probit", data = mydata)> x.out <- setx(z.out)> s.out <- sim(z.out, x = x.out, x1 = NULL) Additional Inputs In addition to the standard inputs, zelig() takes the following additional options for probit regression: ˆ robust: defaults to FALSE. If TRUE is selected, zelig() computes robust standard errors via the sandwich package (see Zeileis (2004)). The default type of robust standard error is heteroskedastic and autocorrelation consistent (HAC), and assumes that observations are ordered by time index. In addition, robust may be a list with the following options: – method: Choose from * "vcovHAC": (default if robust = TRUE) HAC standard errors. * "kernHAC": HAC standard errors using the weights given in Andrews (1991). * "weave": HAC standard errors using the weights given in Lumley and Heagerty (1999). – order.by: defaults to NULL (the observations are chronologically ordered as in the original data). Optionally, you may specify a vector of weights (either as order.by = z, where z exists outside the data frame; or as order.by = ~z, where z is a variable in the data frame). The observations are chronologically ordered by the size of z. –...: additional options passed to the functions specified in method. See the sandwich library and Zeileis (2004) for more options. Examples Attach the sample turnout dataset:> data(turnout) Estimate parameter values for the probit regression: 1> z.out <- zelig(vote ~ race + educate, model = "probit", data = turnout)> summary(z.out) Set values for the explanatory variables to their default values.> x.out <- setx(z.out) Simulate quantities of interest from the posterior distribution.> s.out <- sim(z.out, x = x.out)> summary(s.out) Model Let Yi be the observed binary dependent variable for observation i which takes the value of