Much as we would like to believe that the high citation count for this article is due to the brilliance and clarity of our argument, it is more likely that the count is due to our being in the right place (that is, the right part of the discipline) at the right time. In the 1960s and 1970s, serious quantitative analysis was used primarily in the study of American politics. But since the 1980s it has spread to the study of both comparative politics and international relations. In comparative politics we see in the 20 most cited Review articles Hibbs's (1977) and Cameron's (1978) quantitative analyses of the political economy of advanced industrial societies; in international relations we see Maoz and Russett's (1993) analysis of the democratic peace; and these studies have been followed by myriad others. Our article contributed to the methodology for analyzing what has become the principal type of data used in the study of comparative politics; a related article (Beck, Katz, and Tucker 1998), which has also had a good citation history, dealt with analyzing this type of data with a binary dependent variable, data heavily used in conflict studies similar to that of Maoz and Russett's. Thus the citations to our methodological discussions reflect the huge amount of work now being done in the quantitative analysis of both comparative politics and international relations.

Our methodology deals with the analysis of time-series–cross-section (TSCS) data. Quantitative comparative politics and international relations have leaned heavily on this type of data since the early 1990s. The early quantitative research in comparative politics, such as Cameron's (1978), used simple cross-sectional regression on typically fewer than 20 countries. With so few observations, it was impossible to tease out complicated relationships. The late 1980s saw arguments over whether results were stable if a single country was excluded. The alternative was to use more standard time-series analysis, but to examine each nation's time-series separately, as was done by Hibbs (1977). TSCS analysis pools these time series, allowing for rigorous quantitative analysis of both the temporal and the spatial dimensions of the data. TSCS data are complicated, and users of such data have been very interested in the appropriate methods for analyzing these complicated data.

Our article is also heavily cited because it deals with important issues relevant to political science research. Twenty years ago the small coterie of political methodologists taught out of standard econometrics texts. But the issues relevant to political scientists are not necessarily those relevant to economists. In our case, econometricians have developed sophisticated methods for studying panel data, which appear to look like TSCS data. Given econometricians' interests, they focused on data sets in which the cross-sectional units consisted of a huge number of survey respondents observed over a very small number of survey waves. Methods suitable for the analysis of panel data are not appropriate for TSCS data, and vice-versa. One reason our work is cited by political scientists is that as political scientists we worked on methodological issues relevant to the study of political questions.

Our article also reflects a growing interest in the finite sample properties of the estimators we all use. Econometricians tend to live and die by asymptotic properties; but we all have finite samples and sometimes what is good asymptotically is not good for finite samples. We started with a commonly used estimator that was fine asymptotically but displayed rather poor finite sample performance (at least for typical TSCS data). We showed that in typical data sets least squares, combined with a correction for standard errors, could be a huge improvement over the “asymptotically superior” estimator.

In thinking about our article, we cannot help but remember that we submitted it to the Review after it was rejected by the only other political science journal that at the time published serious methodological articles. It is good to remember that a rejection letter is not the same thing as a death sentence.

What do we wish we had done differently? The study of TSCS data has become vibrant with many ongoing debates. One point that almost everyone understands, or should, is that there is no magical method (or statistical package command) for analyzing TSCS data. The most critical issues are about specification, not about the choice (and surely not the mechanical choice) of an estimation method. Because of the context in which we were writing, we focused on estimation methods, though we always tried to move researchers away from mechanistic thinking about statistical analysis. The ongoing debates have focused our current work on more critical issues of specification. But because we now live in a world where methodological debates are commonplace, we do not worry that we missed many important issues in our article.