In an earlier article for "The Polls" feature (Cohen 2002), I introduced a methodology to create monthly indices of foreign and economic policy presidential approval. I also described the trends in those series in that article. The advantage of using such series, especially foreign policy approval, is that we are able to evaluate the impact of foreign policy on general presidential approval better than previous research on the topic. Data limitations have led scholars to emphasize the role of economics over foreign affairs in public evaluations of presidential performance. A foreign policy public opinion series allows us to gain a better sense of the impact of foreign affairs on public thinking about the president than does the traditional research strategy. That older research strategy uses dummy variables to indicate the occurrence of a foreign policy event that is assumed or hypothesized to affect public opinion. In this article, I extend the discussion of the foreign and economic policy approval series with a special emphasis on their relationship to general presidential approval.
Here, the starting off point is a recent article by Nickelsburg and Norporth (2000), who employ quarterly foreign and economic policy approval series across a twenty-year period to investigate the comparative impacts of the two policy areas in the public mind on evaluations of the president. Their quarterly data, while providing a relatively long series to investigate, cannot easily tease out the causal relationships between foreign policy approval, economic policy approval, and general presidential approval.
One possible causal linkage is that people develop overall impressions of the president, which they then use to inform their assessment of presidential performance in these two policy areas. But public thinking may just as easily run in the opposite direction, from the specific to the general. Theories of political self-interest, as well as Mueller's (1973) notion of coalition of minorities as it applies to presidential approval, are consistent with this second way of thinking. The monthly data used here allow us to employ a granger causality test, which improves over the techniques that Nickelsburg and Norpoth use. Fundamentally, quarterly time units are too crude or too temporally wide to assess if one or more of these three series affects the others. Monthly time units, in contrast, are sliced fine enough to assess whether the three types of approval series are causally related. After assessing their causal relationships, I turn to the question of the comparative impact of economic and foreign policy approval on general presidential approval during the Clinton years.
Policy-Specific and General Presidential Approval: Granger Causality Tests
Granger causality is an especially powerful technique for assessing the causal relationship between variables. A variable, X, granger causes another variable, Y, when past values of X statistically determine current values of Y but past values of Y do not statistically determine current values of X, or more formally,
[Y.sub.t] = a + [b.sub.1][X.sub.t-1] but [X.sub.t] [not equal to] a + [c.sub.2] [Y.sub.t-1].
We can assess the causal relationship between foreign (and economic) approval and general presidential approval by regressing current values of general approval on lagged values of general approval and lagged values of foreign (economic) policy approval and also by regressing current values of foreign (economic) approval on lagged values of itself and lagged values of general approval.
Table 1 presents results of the granger test between economic and general presidential approval, while Table 2 presents the granger test between foreign and general approval. First, looking at general presidential approval, we see that general approval is a function of itself lagged one period. …