Building public support from the grassroots up: the impact of presidential travel on state-level approval.

Cohen, Jeffrey E. ; Powell, Richard J.

On September 11, 2001, as terrorists were attacking the World Trade Center ant the Pentagon, President George W. Bush was visiting a Florida school, publicizing and promoting his educational policies. Of course, a terrorist attack of that magnitude rare, but a domestic presidential trip is a common occurrence. As Figure 1 displays, from 1981 through 1999, presidents made an average of 59 visits to the states per year, aver aging nearly 79 travel days per year. (1) Furthermore, the figure shows an increase in presidential travel to the states over that period.

Presidential travel is an increasingly important aspect of the public presidency, with the public presidency being one of the signature developments of modern presidential leadership (Brace and Hinckley 1992; Hart 1987; Lammers 1982). The public presidency involves, among other things, presidential speeches to the public aimed at increasing presidential visibility and mobilizing public support for the president and his policies.

A considerable body of research has aided our understanding of the many types of presidential activities that make up the public presidency. We know, for instance, about the factors that motivate many presidential public activities, such as press conferences and political speeches (Hager and Sullivan 1994; Kernell 1986; Powell 1999). We also know that major presidential speeches lift presidential job approval, albeit briefly (Brace and Hinckley 1992, 1993; Kernell 1986; Ostrom and Simon i985; Ragsdale 1984, 1987; Simon and Ostrom 1989). In contrast, we know almost nothing about the impact of presidential trips around the nation despite their frequency.

In this study, we investigate the impact of presidential travel on state-level public approval. Not only do we steer our analysis to a topic rarely studied, we also shift our focus from national- to state-level effects. A shift in focus from the nation to the states is important because presidents often target state publics in traveling around the nation. When traveling to a state, presidents may aim to boost their own approval, generate support for local politicians running for (re)election (Cohen, Krassa, and Hamman 1991), build support for their policies among members of Congress (Cohen et al. 1991; Kernell 1986; Powell and Schloyer 2003), and enhance their own reelection prospects. The role of the Electoral College in presidential elections and the importance of states as political units in the American federal system also underscore the importance of looking at state-level effects of presidential travel. This study looks specifically at the job approval implications of presidential travel to the states.

Despite a conventional wisdom that the public activities of presidents reap approval gains, research provides a mixed assessment of the success of presidential efforts to influence public opinion. Some studies, especially those focusing on major public speeches (Brace and Hinckley 1992, 1993; Kernell 1986; Ragsdale 1984, 1987; Ostrom and Simon 1985) and campaign appearances (Cohen et al. 1991; Shaw 1999), find that presidents can affect public opinion. Other studies, however, have raised doubts about presidents' ability to gain the public's attention and influence the direction of public opinion (Edwards 2003). Given these varying assessments, domestic travel provides a useful setting for testing these competing expectations about presidential leadership of public opinion.

In the next section, we present a theoretical explanation of how presidential travel affects public opinion. Then we discuss a new data set that allows us to examine the impact of presidential travel on state-level approval from 1981 through 1999 (Beyle, Niemi, and Sigelman 2002).2 For the first time, we are thus able to match appropriate levels of analysis--travel to the states with state-level approval. (3) We then present our data analysis and conclude with some thoughts on the impact of public presidential activities on presidential approval and suggestions for future research.

A Theory of Presidential Public Opinion Strategies

Presidents commit extensive resources to maintaining and enhancing their level of support in the mass public. Some accounts contend the White House assigns more staff to such publicity operations than any other activity (Grossman and Kumar 1981; Maltese 1994; Patterson 2000). To enhance their mass support, presidents engage in a number of public activities, including nationally televised addresses, staged pseudo-events, minor speeches to specialized audiences, and domestic and international travel.

The volume of public presidential activities seems paradoxical because research suggests the uneven effectiveness of such efforts. Major public speeches and foreign trips, for example, result in only small, fleeting increases in presidential job approval. Ragsdale (1987) finds that major, nationally televised addresses produce a three-percentage point increase in presidential approval on average. Furthermore, these approval boosts are short-lived, evaporating within several weeks. Upon further examination, Brace and Hinckley (1992) find that major speeches lift presidential approval by about three points during presidents' first terms, but have no effect during second terms. Similarly, Ostrom and Simon (1985, 1989) find that presidential trips abroad do little to boost approval when controlling for other factors.

Assuming presidents are rational in their allocation of scarce resources, why would they dedicate so much in the way of time and staff for such a meager payoff?. Possibly, political scientists have been looking for such effects in the wrong places. Specifically, presidents may not view the public as an undifferentiated mass, but may be sensitive to different subgroups. While political scientists have typically tested for the impact of major speeches on the American public as a whole, most presidential public appearances are targeted at narrower, more specialized audiences (Hager and Sullivan 1994; Hart 1987; Kernell 1986; Powell 1999). To reach such audiences, presidents frequently travel outside the immediate Washington, DC area.

So why might presidents expect to receive greater payoffs from targeted visits to specific locales? Answering this question involves an understanding of presidential motivations in seeking public support and the role played by different types of media outlets in that process. First, presidents believe public support helps their ability to govern effectively and to get reelected. High levels of popular support may lead others in the policymaking process, especially Congress, to be more supportive of presidential initiatives (Cohen 1997; Edwards 1997).

Second, presidents believe that positive news coverage is necessary to build and maintain public support (Kurtz 1998). Research suggests that presidential approval rises when the public receives positive information about the president, either because news is positive (Brody 1991) or elite discourse carried in the news is supportive of the president (Zaller 1992). However, gaining positive news coverage has become more difficult for presidents in the post-Vietnam, post-Watergate era because the Washington press corps has become much more adversarial and negative in its coverage (Groeling and Kernell 1998; Grossman and Kumar 1981; Patterson 1994). Although presidents have developed numerous techniques aimed at diminishing such negativity (DiClerico 1993), the incentives for journalists to maintain an aggressive, negative tone seem to be stronger than presidential dissuasions (Patterson 1994).

Consequently, presidents direct many public activities to the local stage through travel to specific locales where they expect news coverage to be more positive (Kaniss 1991). Although presidents often bring some members of the Washington press corps with them on such trips, the brunt of news coverage is borne by local news organizations and national wire services. Wire service coverage can be very helpful for presidents because it tends to emphasize straight news rather than interpretation (Graber 1997). Even more beneficial for presidents, local press coverage of presidential trips tends to be highly positive. Local journalists often view such events as career-enhancing opportunities and they are typically less competent and/or less willing to ask presidents tough questions on controversial issues. These factors converge to produce positive news stories about presidential trips, often emphasizing the pomp, ceremony, and prestige of the office (Kaniss 1991).

Local news coverage of presidential trips also tends to be highly visible. It often garners front-page and lead-story status in the local and regional press, and may persist in the news for days. Thus, news reporting about a local presidential trip may generate positive news that reaches a large, local audience. Over time, presidents can travel around much of the nation, knitting together a sizeable audience in this fashion.

The State Presidential Job Approval Data

Public opinion data for this study come from the State Job Approval Project, (4) which collected state-level job approval ratings for presidents from a host of survey organizations. We utilize the presidential polls from 1981 through 1999 (1,002 observations). (5) However, there are several complications in using these data for our analysis that need to be addressed.

First, the collected polls use a variety of question wordings and response categories and come from a number of different polling organizations using a variety of methodologies. A question arises as to whether we can compare the level of presidential support across questions, especially approval and favorability or likeability questions. Appendix Table 1 lists the number of presidential polls by response category. The JAR data set does not provide question wordings. As the table indicates, almost all presidential poll questions appear to be approval-type questions, based upon the response categories. Only two categories use the term "favorable" (5, 37) for a total of two cases. Almost all of the other response categories use either approval terms or excellent/good variations. While we can be confident that the approval response categories are based on approval questions, we cannot be entirely certain that the "excellent/good" categories are job evaluation questions. Because of this ambiguity in mind, we entered dummy variables for each of these response categories as controls in all of our models, a total of 21 dummies.

In some instances, only positive evaluations of the president are available and "don't know" response rates also seem to vary (Beyle et al. 2002). To correct for these problems, we use the formula "(percent positive)/(percent positive + percent negative)" to measure presidential approval (Beyle et al. 2002). Similarly, Beyle et al.'s (2002) analysis found that as the number of response categories increases, the likelihood of a negative rating also increases. In sum, we need to be cognizant about the potential for house effects. To guard against this problem, our models include dummy variables to control for polling firm effects.

The JAR data set also contains many missing data points. Assuming monthly readings and fifty states, 11,400 observation points are possible, yet we have only 1,002 (8.8 percent) readings. Missing data are more common in earlier than later years, in some states than others, and in non-election years versus election years. This non-random pattern may violate standard regression assumptions. We use robust standard errors to help correct for this problem, but the missing data present a more fundamental design issue that we address by creating a baseline model of presidential job approval.

Ideally, we would want to test whether a presidential visit to a state changes presidential approval. However, the many gaps in the data make it impossible to measure approval change from one month to the next at the state level. It is possible to work around this problem by comparing the president's approval level in a particular state in the days immediately following a presidential visit to those periods when no such visit has taken place, relative to the president's expected baseline popularity in that state. Such a cross-sectional design requires that we control for other factors that affect state-level presidential approval. As we demonstrate below, our normed baseline model effectively accounts for these other sources of variation in state-level presidential job approval.

The Appendix provides some descriptive information about the job approval ratings employed in this analysis. Appendix Table 2 presents a frequency count of the number of job approval ratings by year. Trips to Maryland, Virginia, and the District of Columbia are excluded from these totals. The number of state-level presidential poll readings varies from 5 to 164 per year, with an average of nearly 50 per year. Over time, poll readings tend to increase in frequency. The average number of polls per year tops 84 during the 1990s, compared to 11 per year from 1981 to 1989.

Appendix Table 3 presents the frequency of presidential polls by state. The number of polls per state range from 1 for North Dakota to 93 for California, with an average of 19.6 per state. Although some large states possess a higher than average number of polls, some smaller states, notably Alabama, Arizona, Connecticut, and Kentucky, also show a relatively high number of polls, and an occasional large state, such as Massachusetts, has relatively few presidential polls. The number of commercial, media, and academic polling organizations within the states determines to a large extent the number of state-level presidential polls.

A Baseline Model of Presidential Job Approval

The above discussion suggests two sources of variation in state-level job approval, factors associated with conducting the polls (e.g., response categories and polling firm) and factors that emanate from the political context. To control for changes in the political context we include a normed measure of the presidential vote. Durham et al. (2003) demonstrate that the normed state-level presidential vote in the last election and the Gallup monthly national presidential approval rating effectively explain most of the variance in state-level job approval. The normed presidential vote, which measures the difference between a state's presidential vote and the national average, provides us with a sense of the basis of support for the president across the various states. The Gallup national approval rating controls for the temporal path of presidential approval. Thus, our baseline model controls for both spatial and temporal sources of variation in presidential approval. Following Durham et al. (2003) we employ these two terms, lagging the Gallup measure to insure proper causality.

To isolate the effects of presidential travel, we add several other variables in an attempt to explain as much variance in state-level presidential approval as possible. As the fit of the baseline model improves so too does our confidence that we can estimate the true impact of presidential trips on state-level approval. Thus, we add variables representing state-level economic conditions, consumer perceptions of the economy, and state-level partisanship and ideology, all of which have been found in prior research to affect public attitudes toward the president.

Although we experimented with a larger number of economic variables, we found four to be consistently related to state-level presidential approval--inflation-adjusted state income (lagged one quarter), the consumer price index (lagged one month), consumer expectations, and consumer confidence (lagged one month). The consumer expectations and confidence data come from the University of Michigan's Survey of Consumers. (6)

Last, we add controls for state partisanship and ideology, using the updated Erikson, Wright, and Maclver (1993) measure. For state partisanship, we calculate the percentage of Democratic and Republican identifiers for subsets of years (1977-1984, 1985-1992, and 1993-2001). The subsets are long enough to protect against short-term variation that may result from sampling and other error, but still remain relatively close in time to the measurement of the president's popularity. We calculate state partisanship as (percentage of the president's party identifiers)--(percentage of opposition party identifiers). Similarly, we calculate state ideology with the formula for Democratic presidents as (percentage of liberal identifiers)--(percentage of conservative identifiers), and vice versa for Republican presidents.

Table 1 presents the results of regressing state presidential approval on the political context variables, plus dummies for polling firm and response category, using robust standard errors. This baseline model explains the overwhelming proportion of the variance in state-level presidential approval, with an R2 surpassing 76 percent. These results indicate that state-level approval is, in fact, largely a function of the normed presidential vote, the prior month's national approval rating, quarterly lagged income, lagged economic confidence, current economic expectations, lagged CPI, state partisanship, and state ideology. Although not shown in Table 1, many of the polling firm and response category dummies also attain statistical significance. Given the strong [R.sup.2], the baseline control model effectively accounts for the major sources of variation in state-level presidential approval.

A Model of Presidential Job Approval in the States

Our primary question asks whether a presidential trip to a state significantly influences presidential approval, controlling for that state's expected approval of the president from the baseline model. The strong [R.sup.2] of the baseline model presents a high hurdle for assessing the effects of presidential trips.

We measure presidential visits with a dichotomous dummy variable (VISIT), which indicates the presence or absence of a presidential visit to a state during a particular month. (7) We use the lagged measure of this variable to eliminate endogeneity between presidential visits and popularity. Also, we drop Maryland, Virginia, and the District of Columbia from the estimation because of the large number of visits they receive and ambiguity about whether these are truly "local" visits. As shown in Appendix Tables 2 and 3, about 19 percent of the data points with presidential state polls were preceded by a presidential visit to the state in the previous month. Nine states for which we possess presidential poll data do not show a presidential trip to the state in the preceding month. Most of these states are ones with small numbers of state polls. (8)

Above, we also conjectured that presidential visits to the states are typically covered by the local media. We were able to get an illustrative sense of local news coverage of presidential trips by using the Lexis-Nexis news database. To do so, we used the U.S. News category, which permits one to search the news sources for each state separately. Unfortunately, it was not possible to collect data that were comparable across states because of unevenness in the inclusion of news media sources across states. For instance, some major newspapers, like the Detroit Free Press, are not included. Thus, even large states with major news organizations are not always included in the Lexis-Nexis database. Second, the earlier in time, the fewer the number of news organizations included. In the early years of the Lexis-Nexis compilation, the New York Times dominated coverage.

With these limitations in mind, we performed an exploratory examination of news coverage of presidential trips in 1995, a year we chose because it was a non-election year and also recent enough to contain relatively broad coverage in Lexis-Nexis. Then, we ran a search for news coverage in each state the president visited during the month of the visit. (9) Appendix Table 4 displays the results of this search for each state the president visited. As this table indicates, many of the presidential trips did, in fact, generate healthy doses of coverage. For example, President Clinton's 1995 trips generated a total of 34 stories in California, 37 in Florida, 12 in Georgia, 15 in Illinois, 29 in Michigan, 28 in New York, 8 in Ohio, and 15 in Texas. While 11 of the 25 states indicate no corresponding story, we have no way to determine whether this was due to the limitations of the Lexis-Nexis archive, as noted above, or the failure of the news media in these states to cover the president's trip. Nevertheless, our search suggests that presidential trips often yield local news coverage.

As shown in Table 1, regressing VISIT on presidential approval, with controls for the baseline variables, indicates that a presidential visit boosts state-level support by about 1.3 percent on average, which is statistically significant (p = .043, t = 2.03). While this effect is modest, it may be the case that longer stays in a state will have more impact on approval than shorter stays. To test this hypothesis, we reran the model with a variable measuring the total number of days the president visited the state. (10) Results (not shown) indicate that the length of stay in a state matters little, failing to reach statistical significance. All of the impact of a trip on approval is picked up by the dichotomous VISIT variable. (11)

Contextual Effects and Presidential Trips

The effects of presidential trips may vary with characteristics of presidents and the state to which they travel. In this section, we test for contextual effects of state size and the election cycle.

Large- Versus Small-State Effects of Presidential Visits

Presidents are more likely to travel to large states than small ones (the correlation between population and days in the state is .56, p = .000). Small population size reduces presidential incentives to visit. This is especially the case during election periods, when Electoral College implications become increasingly important in presidential calculations. The relative scarcity of presidential trips to small states, however, may raise the value of a trip by generating greater levels of press coverage due to the novelty of a presidential visit. Also, trips to small states may receive relatively more statewide attention because small states are likely to have fewer large media markets. For example, a presidential trip to Buffalo, NY may go unreported in the New York City area, but a trip to Casper, WY would likely generate news across the state. This may lead to greater popularity boosts in small as opposed to large states, despite general incentives to focus more attention on large states. We test this hypothesis by dividing our sample into small and large states. (12)

Table 2 reports the results of our model with dummy variables added for large and small states. Contrary to our hypothesis, visits to large states are more likely to generate approval effects than visits to small states. Although the regression coefficients for the two trip variables are similar, only the large-state variable attains statistical significance. (13) The results indicate that a trip to a large state generates about a 1.2 percent increase in approval, about the same magnitude of the undifferentiated trip variable from Table 1. (14)

Although we were surprised by these findings, on further examination it seems that several factors may account for the greater impact of presidential trips in large states than small states. Perhaps the news media in small states do not possess the resources to cover a presidential trip as intensely as media organizations in larger states, which serve larger media markets. Moreover, the major news media in many small states are often dominated by news organizations from larger states, especially for small states located near larger ones. These out-state news organizations may not cover presidential trips to neighboring small states as intensely as they report presidential trips to their home state. For example, news media based in New York or Pennsylvania may be less likely to cover trips to New Jersey or Delaware.

Election Cycle Effects

Election cycles, by influencing presidential travel decisions, might also affect whether a trip boosts the president's popularity. For instance, presidential sensitivity to public opinion may be greater in election than non-election years. Therefore, election-year travel may be geared more toward popularity effects, focusing on the electoral implications of enhanced presidential popularity. Such concerns might be muted during non-election years.

These effects may also be different in presidential and midterm election years. In midterm election years, presidents often travel in support of their party's candidates for a variety of offices (Cohen et al. 1991). In the absence of being directly challenged during midterms, presidents may accent their presidential duties and wrap themselves (and the candidates for whom they are aiding) in the aura of the office. In contrast, presidential election years pit the president against an adversary who may try to undermine the president's image, forcing the president to engage in political self-defense. Presidents may look more like candidates and less like presidents in presidential election years. Thus, we may not find as great an impact of trips on approval during presidential election years as in midterm election years, even though the impact of travel on approval is likely to be stronger in both types of election years than non-election years.

We test these hypotheses by adding to our baseline model interaction terms between a presidential visit and the three points of the electoral cycle: non-election years, midterm election years, and presidential elections years. The results from this model, shown in Table 2, confirm that trips in presidential election years do not affect presidential approval, but neither do those in midterm election years. On the other hand, trips in non-election years display a significant effect, boosting approval by about 1.4 points (p = .05, two-tailed test). (15)

Perhaps it is the case that presidents appear more presidential when they are not campaigning. When campaigning for themselves or others, presidents may look like politicians. While presidential campaign appearances may help stimulate turnout among those already likely to support the president or his allies, campaign appearances, due to their overtly partisan nature, may not result in higher job approval for the president. In contrast, when traveling around the nation during non-election years, presidents may utilize the symbolic aspects of the office to generate heightened support. As such, they look more like presidents than candidates for office, eliciting a more positive response from citizens.

Furthermore, in presidential election years, the opposition party's candidate for the presidency is also highly likely to visit the state, undermining the effects of the sitting president's visit to a state. In other words, the campaign effort of the two competing candidates may cancel each other's impact on state electorates. And as Cohen et al. (1991) argue, in midterm elections, the presidential decision calculus to visit a state may be motivated more by a desire to help his co-partisans than to boost his own popularity. Thus, the context of elections may mute the popularity effects of a presidential visit.

Conclusion

Over the past hundred years or so, public activities have become a pronounced aspect of the presidency. Even though a great deal of research has been conducted on the public presidency, almost all of it has focused on the impact of these presidential activities on national opinion. Although much presidential activity targets subsets of the nation, we have little understanding of the impact of presidential activities on these subnational publics. In this study, we shift our research focus to the states to address this concern.

Our study began with a seeming paradox: If presidents are rational in their allocation of scarce resources (time, staff, etc.), why do they commit so many resources to boosting public approval when their ability to do so by giving speeches and appearing in public seems so modest? We hypothesized presidents will enjoy greater returns from their public activity by traveling outside of the Washington, DC area where the local news reporting environment is less hostile. Using a new data set of over 1,000 poll readings at the state level across the 1980s and 1990s, we find that presidents can achieve a modest boost in state-level approval through strategically crafted public appearances. However, these effects are not uniform across all types of states and contexts. In general, we find that presidents hoping to achieve approval gains through domestic travel are more likely to be successful by focusing their efforts on large states in non-election years.

Because we detect only about a one- to two-point impact on approval from a presidential trip, some may find our results exceedingly modest and thus of little substantive interest. However, data and design limitations almost certainly contribute to the limited nature of the effects we detect. Although we possess 1,000 cases for analysis, they represent a sprinkling of data points across the time period we examined. This forced us to use a cross-sectional design, when ideally we would prefer a dynamic design that would enable us to measure the difference in approval before and after a presidential trip. Our cross-sectional approach may have attenuated the effect of presidential trips. Given these data and design restrictions, we find it remarkable that we uncovered a statistically significant effect at all.

Furthermore, we should only have expected a marginal effect on presidential approval. As an aggregate attribute, presidential approval tends to move slowly, even when major shocks hit, like positive or negative events (Brace and Hinckley 1992). Few events produce massive surges or declines in presidential approval. That is, few events stimulate rallies (Edwards and Swenson 1997). Given the inertia in presidential approval, our results are more impressive than they appear at first glance.

Finally, our measurement of presidential trips is quite blunt. When we refine our analysis to distinguish between trips to different types of states and at different points in the electoral cycle, we find that trips only affect opinion in larger states and during non-election periods. This indicates that the impact of presidential trips on public opinion varies with key aspects of the political context. A fruitful avenue for future research along the lines pursued here might be to refine the measurement of presidential trips, as well as public activities in general.

Despite these obvious limitations, we learn something new about the relationship between the president and the public from our findings. Presidents seem to more greatly affect public opinion when they appear presidential, as opposed to being a candidate for office or a supporter of their party's candidates for office. Partisan behaviors are naturally divisive; they are likely only to mobilize those already favorable to the president, as Cohen et al. (1991) find in their study of presidential appearances in midterm elections. But when presidents appear "presidential," that is, when they speak to the mass public writ broadly and when they speak to it as the nation's representative and leader, they are able to increase their level of support.

Appendix

APPENDIX TABLE 1

Number of Presidential State-Level Popularity Polls by Question
Response Category

JAR Code N Response

 1 451 Excellent, good/fair, poor
 3 275 Approve/disapprove
 5 1 Favorable/unfavorable
 6 78 Excellent, good/only fair, poor
 7 11 Excellent, good/not so good, poor
 8 60 Excellent, good/fair, poor, very poor
10 9 Excellent, pretty good/only fair, poor
12 9 Very good, good/fair, poor, very poor
13 1 Excellent, above average/average/below average, poor
16 6 Strongly approve, approve/disapprove, strongly
 disapprove
19 2 A B/C/D F rating
21 16 Excellent, pretty good/just fair, poor
22 1 Positive/negative
28 3 Excellent, pretty good/fair, poor
30 1 A 4-point scale from high {4/3} to low [2/1]
37 1 Very favorable, somewhat favorable/somewhat
 unfavorable or very unfavorable
39 2 Excellent, good/below average, poor
40 1 Fairly good/fairly bad, in between
41 2 Above average/average, below average
42 3 Excellent, very good/fair, poor
43 1 Satisfied/dissatisfied

Source: Official State Job Approval Ratings data set. See text for
details.

APPENDIX TABLE 2

State Presidential Poll Readings and State Trips, by Year, 1981-1999

 Presidential Trips to State

Year Number of Poll Readings No Yes % (a)

1981 5 5 0 0
1982 7 7 0 0
1983 7 7 0 0
1984 11 9 2 18.18
1985 11 11 0 0
1986 15 11 4 26.67
1987 16 15 1 6.25
1988 18 13 5 27.78
1989 12 10 2 16.67
1990 23 15 8 34.78
1991 48 41 7 14.58
1992 19 13 6 31.58
1993 102 87 15 14.71
1994 164 141 23 14.02
1995 124 109 15 12.10
1996 164 122 42 25.61
1997 49 41 8 16.33
1998 100 74 26 26
1999 48 31 17 35.42

APPENDIX TABLE 3

State-Level Presidential Polls and Trips, by State, 1981-1999

 Presidential Trip

State Total No Yes % (a)

Alabama 39 38 1 2.56
Alaska 2 2 0 0
Arizona 42 41 1 2.38
Arkansas 12 11 1 8.33
California 93 41 52 55.91
Colorado 13 13 0 0
Conn. 34 30 4 11.76
Delaware 3 3 0 0
Florida 37 28 9 24.32
Georgia 24 21 3 12.50
Hawaii 5 5 0 0
Idaho 6 5 1 16.67
Illinois 34 19 15 44.12
Indiana 19 17 2 10.53
Iowa 23 22 1 4.34
Kansas 7 7 0 0
Kentucky 31 24 7 22.58
Louisiana 16 15 1 6.25
Maine 4 4 0 0
Mass. 6 4 2 33.33
Michigan 32 27 5 15.63
Minnesota 16 14 2 12.50
Mississippi 12 11 1 8.33
Missouri 5 2 3 60.00
Montana 14 13 1 7.14
Nebraska 11 11 0 0
Nevada 9 9 0 0
New Hampshire 24 22 2 8.33
New Jersey 61 53 8 13.11
New Mexico 21 20 1 4.76
New York 58 24 34 58.62
North Carolina 26 24 2 7.69
North Dakota 1 1 0 0
Ohio 32 26 6 18.75
Oklahoma 10 10 0 0
Oregon 7 7 0 0
Pennsylvania 9 7 2 22.22
Rhode Island 20 19 1 5.00
South Carolina 10 10 0 0
South Dakota 4 4 0 0
Tennessee 11 9 2 18.18
Texas 36 28 8 22.22
Utah 3 3 0 0
Vermont 7 7 0 0
Washington 15 14 1 6.67
West Virginia 5 4 1 20.00
Wisconsin 22 22 0 0
Wyoming 12 11 1 8.33

(a) Percentage of state presidential polls with a presidential visit
to the state in the previous month.

APPENDIX TABLE 4

Number of Local News Stories about Presidential Trips in the
Month of the President's Trip to a State, 1995

State Stories

Arkansas 9
California 34
Colorado 18
Connecticut 2
Florida 37
Georgia 12
Hawaii 0
Illinois 15
Iowa 0
Massachusetts 0
Michigan 29
Minnesota 2
Montana 0
New Hampshire 0
New Jersey 0
New York 28
North Carolina 0
Ohio 8
Oklahoma 0
Oregon 0
Pennsylvania 1
Tennessee 1
Texas 16
Vermont 0
Wyoming 0

Source: Lexis-Nexis. See text for details.

TABLE 1

Impact of Baseline and Presidential Visits on State-Level
Job Approval, 1981-1999 (Robust Regression)

 Baseline Model

 b SE t p

Visit (lag) --
Gallup (lag) .79 .04 21.76 .000
Normed vote .32 .06 5.01 .000
St. income (lag) .08 .02 3.06 .002
Expectations .13 .04 3.43 .001
Confidence (lag) .22 .04 5.23 .000
CPI (lag) -.40 .04 -11.27 .000
St. PID .21 .06 3.74 .000
St. ideology .22 .08 2.84 .005
Constant -7.68 5.83 -1.32 .188
[R.sup.2] .77
n 929

 Visit Model

 b SE t p

Visit (lag) 1.31 .64 2.03 .043
Gallup (lag) .79 .04 21.95 .000
Normed vote .32 .06 5.02 .000
St. income (lag) .07 .02 2.94 .003
Expectations .13 .04 3.54 .000
Confidence (lag) .22 .04 5.20 .000
CPI (lag) -.40 .04 -11.35 .000
St. PID .21 .06 3.66 .000
St. ideology .21 .08 2.74 .006
Constant -6.84 5.86 -1.17 .244
[R.sup.2] .77
n 929

Source: JAR data set. The estimation also includes controls for polling
firms and response categories.

TABLE 2

Impact of Presidential Visits by Election Cycle and State Size,
1981-1999 (Robust Regression)

 Election Cycle

 b SE t P

Visit-midterm 2.18 2.49 .88 .38
Visit-pres. elec. .53 1.17 .46 .65
Visit-no election 1.40 .72 1.94 .05
Visit-large state -
Visit-small state -
Gallup (lag) .79 .04 21.93 .00
Normed vote .32 .06 5.11 .00
St. income (lag) .07 .02 2.89 .004
Expectations .14 .04 3.61 .00
Confidence (lag) .21 .04 5.19 .00
CPI (lag) -.40 .04 -11.35 .00
St. PID .20 .56 3.64 .00
St. ideology .21 .77 2.73 .01
Constant -6.49 5.77 -1.13 .26
[R.sup.2] .77
n 929

 State Size

 b SE t p

Visit-midterm --
Visit-pres. elec. --
Visit-no election --
Visit-large state 1.19 .72 1.66 .10
Visit-small state 1.67 1.30 1.28 .20
Gallup (lag) .79 .04 21.94 .00
Normed vote .32 .06 5.00 .00
St. income (lag) .07 2.00 2.95 .003
Expectations .13 .04 3.55 .00
Confidence (lag) .22 .04 5.20 .00
CPI (lag) -.40 .04 -11.41 .00
St. PID .21 .56 3.66 .00
St. ideology .21 .77 2.75 .01
Constant -6.91 5.89 -1.17 .24
[R.sup.2] .77
n 929

Source: JAR data set. The estimation also includes controls for polling
firms and response categories.

(1.) A visit to a state occurs every time a president travels outside of the Washington, DC area, appears in a state, and makes a public speech. Presidents may visit several states on one excursion. These counts exclude all public presidential appearances made within Washington, DC, Maryland, and Virginia.

(2.) Because we use state-level approval data, our level of analysis is the state. Thus, when we speak of domestic presidential travel, we are speaking of presidential visits to states. Other units of analysis are possible, such as to localities, but lacking data on approval at the local level, we cannot pursue an analysis such as the one offered here.

(3.) The only existing analysis of presidential domestic travel looks at its impact on national-level presidential approval. That research suggests that domestic travel results in lower national presidential approval during the president's first term in office, but has no effect during the second term (Brace and Hinckley 1992, 52). It is not clear why presidential travel should sometimes result in lower approval levels. Brace and Hinckley argue that presidents travel around the nation in response to negative events, but they provide no direct evidence in support of their hypothesis. We argue one possible cause of Brace and Hinckley's negative finding is that they mismatch units of measurement, state travel with national-level approval. Simon and Ostrom (1989) look at the impact of foreign travel on presidential approval and find no impact.

(4.) This data collection project was directed by Thad Beyle, Richard Niemi, and Lee Sigelman, and funded by the NSF (grant SES-9974176). We thank them for allowing us to use the data.

(5.) The data series begins with 1947, but only 34 readings exist from 1947 to 1980.

(6.) We experimented with various lags of the named variables, plus monthly unemployment data measured for the month of the poll reading, as well as for each of the three months preceding each poll (a total of four variables). Many of the economic variables are highly correlated with each other. This may induce multicollinearity, rendering the coefficient's significance tests unreliable. To determine whether a variable should remain in the baseline model, we computed F-tests comparing the [R.sup.2]'s of equations that included or excluded the variable in question. Each of the final variables in the baseline model added a significant increment to the explained variance based on the F-tests.

(7.) Data on presidential trips were collected from successive issues of the Public Papers of the Presidents of the United States series, which provides a complete record of all public presidential statements. We include any trip on which the president delivered at least one spoken public statement of at least 150 words. Thus, we do not include purely private trips such as family vacations with no public appearances.

(8.) These nine states account for 12 percent of the presidential poll readings.

(9.) In searching the states' media sources, we used the following criteria: "President Clinton w/s (within the same sentence as) speech or visit or trip or appearance or reception or spoke or attend or tour or address or campaigned."

(10.) The VISIT variable was actually calculated from the DAYS variable. For both variables, when the president did not visit the state, the variable is scored a "zero." The only difference in VISIT and DAYS is that VISIT is scored a "one" no matter how many days a president visited the state, but DAYS takes on the actual value of the number of days that a president visited the state.

(11.) To test for the robustness of these results, we reran all estimations using OLS, which essentially replicated the reported results. Furthermore, we estimated fixed effects models, identifying fixed effects for polling firm, response categories, and question wording. Reported results remained intact, with a presidential visit to a state boosting job approval by about 1.5 percentage points.

(12.) We use a cut point of 5 million people to classify small and large states. This bisects the sample into subsets of 415 and 514 cases. We tried other cut points, but the results remained essentially the same.

(13.) The table indicates a two-tailed p value of .10, but because we know direction, we should more properly use a one-tailed test, which produces a p value of .05.

(14.) The small-state coefficient may be depressed because we have not distinguished between small states that out-state media have penetrated significantly from small states with little out-state media presence. Our idea here is that when out-state media possess a large presence in a small state, they are less likely to cover a presidential trip. The lack of presidential trip coverage should result in weaker trip effects. We used the Nielsen Media Designated Market Area (DMA) map to determine how much of a state's media markets are controlled by out-state news media. Of small states, we find 17 in which indigenous media control the overwhelming percentage of the state's news media markets. Unfortunately, in our data set, we located only 10 presidential trips to these 17 states in months for which we also have state-specific presidential popularity data. We created a dummy variable for these small states with a presidential visit. When entered into various equations with and without the other state size variables we find no impact, but the lack of effect may be a function of the paucity of data to test this hypothesis.

(15.) We also investigated three-way interactions among VISIT, the election cycle, and state size by creating a dummy variable for VISIT plus large state plus non-election year and another one for VISIT plus small state plus non-election year. Our hypothesis is that the modest effects noted for the election cycle and state size might grow when both conditions are present. In neither case did these interactions prove statistically significant. When we enter the visit-large-non-election dummy, we generate b = .87, SE = .77, t = 1. 13, p = .26 (two-tailed). For the visit-small-non-election dummy we produce b = 2.45, SE = 1.82, t = 1.35, p = .18 (two-tailed). With both of these dummies entered together, results are comparable to those reported here. It is not clear why the three-way interaction would not reach significance when the election cycle and state size variables do separately, other than the three-way interaction may be segmenting the states into small subsets or the new interactions may not be discriminating among the states as well as the original state size and electoral cycle variables. In fact, when large-state and non-election cycle variables are added into the estimation together, neither reaches significance because they are highly correlated (r = .73, p = .00), indicating high multicollinearity between them. Presidents, it appears, are more likely to visit large states in non-election years compared to election years. The correlations for visits to large states in presidential and midterm election years are .27 (p = .00) and .33 (p = .00), respectively. While statistically significant, these correlations are of modest magnitude.

References

Beyle, Thad, Richard G. Niemi, and Lee Sigelman. 2002. Gubernatorial, senatorial, and state-level presidential job approval ratings: The U.S. officials job approval ratings (JAR) collection. State Politics and Policy Quarterly 2(Fall): 215-29.

Brace, Paul, and Barbara Hinckley. 1992. Follow the leader: Opinion polls and the modern presidents. New York: Harper Collins.

--. 1993. Presidential activities from Truman through Reagan: Timing and impact. Journal of Politics 55(May): 382-93.

Brody, Richard. 1991. Assessing the president. Stanford, CA: Stanford University Press.

Cohen, Jeffrey E. 1997. Presidential responsiveness and public policy-making. Ann Arbor, MI: University of Michigan Press.

Cohen, Jeffrey E., Michael A. Krassa, and John A. Hamman. 1991. The impact of presidential campaigning on midterm U.S. Senate elections. American Political Science Review 85(March): 165-78.

DiClerico, Robert E. 1993. The role of media in heightened expectations and diminished leadership capacity. In The presidency reconsidered, edited by Richard W. Waterman, 115-44. Itasca, IL: Peacock.

Durham, Brandy M., Jeffrey E. Cohen, Richard Fleisher, and Jon R. Bond. 2003. The effects of state level presidential approval on support in the Senate. Paper presented at the 2003 Midwest Political Science Association meeting, April 15 to 18, in Chicago.

Edwards, George C. III. 1997. Aligning tests with theory: Presidential approval as a source of influence with Congress. Congress and the Presidency 24(Autumn): 113-30.

--. 2003. On deaf ears: The limits of the bully pulpit. New Haven, CT: Yale University Press. Edwards, George C. III, and Tami Swenson. 1997. Who rallies? The anatomy of a rally event. Journal of Politics 59(February): 200-12.

Erikson, Robert S., Gerald C. Wright, and John P. MacIver. 1993. Statehouse democracy: Public opinion and policy in the American states. New York: Cambridge University Press.

Graber, Doris A. 1997. Mass media and American politics. 5th ed. Washington, DC: Congressional Quarterly Press.

Groeling, Tim, and Samuel Kernell. 1998. Is network news coverage of the president biased? Journal of Politics 60(November): 1063-87.

Grossman, Michael Baruch, and Martha Joynt Kumar. 1981. Portraying the president: The White House and the news media. Baltimore: Johns Hopkins University Press.

Hager, Gregory L., and Terry Sullivan. 1994. President-centered and presidency-centered explanations of presidential public activity. American Journal of Political Science 38(November): 1078-103.

Hart, Roderick P. 1987. The sound of leadership: Presidential communication in the modern age. Chicago: University of Chicago Press.

Kaniss, Phyllis. 1991. Making local news. Chicago: University of Chicago Press.

Kernell, Samuel. 1986. Going public: New strategies of presidential leadership. Washington, DC: Congressional Quarterly Press.

Kurtz, Howard. 1998. Spin cycle: How the White House and the media manipulate the news. New York: Touchstone.

Lammers, William W. 1982. Presidential attention-focusing activities. In The president and the public, edited by D. Graber, 145-71. Philadelphia: Institute for the Study of Human Affairs.

Maltese, John Anthony. 1994. Spin control: The White House Office of Communications and the management of presidential news. 2d ed. Revised. Chapel Hill, NC: University of North Carolina Press.

Ostrom, Charles W., and Dennis M. Simon. 1985. Promise and performance: A dynamic model of presidential popularity. American Political Science Review 790une): 334-58.

Patterson, Thomas. 1994. Out of order. New York: Vintage Press.

Patterson, Bradley H., Jr. 2000. The White House staff: Inside the West Wing and beyond. Washington, DC: Brookings Institution Press.

Powell, Richard J. 1999. 'Going public' revisited: Presidential speechmaking and the bargaining setting in Congress. Congress and the Presidency 26(Fall): 153-70.

Powell, Richard J., and Dean Schloyer. 2003. Public presidential appeals and congressional floor votes: Reassessing the constitutional threat. Congress and the Presidency 30(Fall): 123-38.

Ragsdale, Lyn. 1984. The politics of presidential speechmaking, 1949-1980. American Political Science Review 79(December): 971-84.

--. 1987. Presidential speechmaking and the public audience: Individual presidents and group attitudes. Journal of Politics 49(August): 704-27.

Shaw, Daron R. 1999. The effect of TV ads and candidate appearances on statewide presidential votes, 1988-96. American Political Science Review 93(June): 345-61.

Simon, Dennis, and Charles W. Ostrom, Jr. 1989. The impact of televised speeches and foreign travel on presidential approval. Public Opinion Quarterly 53(Spring): 58-82.

Zaller, John. 1992. The nature and origins of mass opinion. New York: Cambridge University Press.

Jeffrey E. Cohen is professor of political science at Fordham University and the author of several books, including Presidential Responsiveness and Public Policy-Making, as well as articles in numerous journals, including the American Political Science Review, American Journal of Political Science, and Journal of Politics.

Richard J. Powell is assistant professor of political science at the University of Maine and the author of journal articles and book chapters on the presidency, Congress, elections, and state politics, and co-author of Changing Members The Maine Legislature in the Era of Term Limits.