Red Scare - Revisiting Joe McCarthy's Influence on 1950s Elections

Red Scare? Revisiting Joe McCarthy's Influence on 1950s Elections 1 Abstract: In the early 1950s, politicians apparently allowed themselves to be spectators to the anticommunist witch hunt of Senator Joe McCarthy and his supporters, leading to a particularly grim chapter in American politics. In part, they did so because they thought the public supported McCarthy. Although politicians lacked contemporary public opinion data, they apparently inferred McCarthy’s support from key Senate race outcomes. The few senators who initially stood up to McCarthy lost their reelections when McCarthy campaigned against them. In this paper, we revisit the case of McCarthy’s influence and investigate whether politicians fundamentally misinterpreted support for McCarthy. Using county- and state-level election data from across the twentieth century, we develop plausible counterfactual measures of normal electoral support to assess McCarthy’s influence on electoral outcomes. We adopt a variety of analytic strategies that lead to a single conclusion: there is little evidence that McCarthy's attacks mattered to the election outcomes. Our results imply that politicians can greatly err when interpreting the meaning of elections, and point to the importance of research on elections to help prevent such errors. Over the last 50 years, political scientists have learned a great deal about how public opinion manifests itself in the political process during periods of “normal politics.” The mass public and political actors seem to influence each other in regular and predictable ways (Page and Shapiro 1992; Zaller 1992; Stimson 2004). But not always. Sometimes demagogues rise from obscurity and capture the public’s attention. Other times, elites misread public opinion and mistakenly attempt to exert influence beyond the bounds of what the public might permit. Our paper examines a particular kind of democratic failure: politicians misperceiving public support for a public policy or person. For elected officials, accurately divining the meaning of elections is a critical part of democratic representation, but not an easy one. After all, elections are blunt instruments for the expression of political views, and the voters themselves do not define the meaning of any given election; politicians and the media take the lead in that regard (Hershey 1992). As Shamir and Shamir argue, “How the election is understood is a major normative force in public opinion and in democratic politics…it is also significant in the conduct of politics and policy implementation following the elections” (2008, 213). Throughout American history, particular interpretations of elections have had real political consequences. For instance, Ronald Reagan claimed that his landslide victory in the 1980 election reflected an affirmation of his conservative agenda, and he used his victory as a platform to enact cuts in taxes. To others, however, the election seemed more a rejection of Carter than an embrace of conservatism (Kelly 1983). In the wake of the passage of the 2010 Affordable Care Act, many Republicans sought to cast their decisive victory in the 2010 election as a rejection of “Obamacare” and the president’s broader agenda. For instance, Senate Minority Leader Mitch McConnell argued that the election result “means that we can—and should— propose and vote on straight repeal [of the ACA], repeatedly” (NBC News 2010). Such claims can help shape the political landscape. To the extent that a particular interpretation—or misinterpretation—of an election becomes conventional wisdom among the political set, serious political consequences can ensue. Given the importance of interpreting elections, a key question is: how competent are politicians at divining the public will from elections? We approach this question by reopening the case of Wisconsin Senator Joseph McCarthy.1 In the early 1950s, politicians apparently allowed themselves to be spectators to the anticommunist witch hunts conducted by McCarthy and his supporters, leading to a particularly grim chapter in American politics. Many Americans were imprisoned, thousands more lost jobs, and countless others were stigmatized as McCarthy and his supporters targeted citizens from all walks of life, including politicians, musicians, actors, and scientists (Schrecker 1998). At first, some politicians stood up to McCarthy, most notably four-term incumbent Democratic Senator Millard E. Tydings of Maryland and Democratic Senate Majority Leader Scott W. Lucas of Illinois. McCarthy had campaigned against Tydings and Lucas, and both lost their races. Many politicians and journalists interpreted these defeats as a sign of public support for McCarthyism. While members of Congress on both sides of the aisle continued to privately express concerns about McCarthyism, in public they kept silent. Elites may have correctly interpreted the defeat of politicians such as Tydings and Lucas as a signal of public support for McCarthy. But they may also have gravely erred. Since we lack 1 Few studies have examined politicians' competence at interpreting elections, though Fowler (2013) documents an interesting example in Australia. Some studies have examined a related question about the accuracy of politicians' perceptions of public opinion, finding that perceptions can be surprisingly inaccurate (e.g., Broockman and Skovron 2013). public opinion data on McCarthy during his heyday—Gallup conducted only one poll (in August 1951) on McCarthy’s favorability ratings before the 1952 election—we do not know whether the public actually supported McCarthy.2 In the decades since McCarthy’s collapse, researchers have developed a better understanding of the factors that influence senate elections, such as the importance of partisanship, presidential candidates’ coattails, and challenger quality. These factors, not McCarthy’s campaign rhetoric, may explain the high-profile defeats of politicians such as Tydings and Lucas. In effect, McCarthy’s apparent power may have been built on a fictional foundation. In our investigation of the McCarthy case, we use county- and state-level election data from across the twentieth century and a timeline of McCarthy’s campaign activities drawn from newspaper archives and political biographies to test whether he actually contributed to the defeat of Democratic incumbent senators in 1950 and 1952. Specifically, we seek to develop plausible counterfactual measures of normal electoral support to assess the extent of McCarthy’s influence on electoral outcomes, net of the critical political variables often studied by social scientists. In the pages that follow, we first describe McCarthy’s tactics against opponents in the 1950 and 1952 elections. We identify all the senate races in which he sought to influence the electoral outcome—not just those races, such as the Tydings and Lucas races, which history has deemed to be impressive demonstrations of McCarthy’s power. We next estimate statistical models to predict county-level two-party voting in all senate races from 1938 to 1964 (or 1950 to 1964, depending on the specification) and examine whether Democratic senate candidates fared worse than predicted when McCarthy intervened against them. Finally, we consider a nonmodel-based approach, comparing how the senators he targeted fared relative to other 2 See Polsby (1960, 252) for a series of Gallup questions on McCarthy’s favorability. Democratic candidates in the same states in the same years. In all these analyses, we find little evidence that McCarthy’s attacks mattered to the elections’ outcomes. Although we cannot infer from a single case the measure of politicians’ general competence at interpreting elections, our results nevertheless suggest that politicians can gravely err when attempting to determine their meaning. McCarthy’s Influence? In February 1950, during a speech to the Republican Women’s Club of Wheeling, West Virginia, Senator McCarthy—a first-term Republican from Wisconsin—declared that he had a list of the names of many communists working within the State Department. McCarthy reiterated his claim throughout the year, receiving widespread publicity. In the elections that year, he actively campaigned on behalf of six Republican senate candidates, making speeches in each of their states. For instance, according to the Washington Post, McCarthy spoke in Baltimore and Hyattsville, MD attacking Tydings for his handling of McCarthy’s charges of Communist penetration into the State Department (Ryan 1950). In addition, he rallied his financial backers to donate thousands of dollars to the Republican challenger’s campaign (Reeves 1982). These activities seemed to pay off for McCarthy; he was credited with the defeat of Tydings and Lucas (as noted above). The defeat of Tydings, in particular, seemed to amplify McCarthy’s power. Besides being a four-term senator and a two-term congressman, Tydings chaired the “Tydings Committee,” a subcommittee of the Senate Committee on Foreign Relations, which formed in February 1950 to investigate McCarthy’s claims regarding communists in the State Department. The committee ultimately produced a final report disputing those claims, provoking McCarthy's ire and ultimately leading—in the view of many—to Tydings’ defeat.3 These high-profile defeats set the stage for McCarthy’s efforts in 1952, when he campaigned on behalf of six Republican Senate candidates, most notably William Purtell, who challenged Democratic Senator William Benton of Connecticut. In Westbury, CT, McCarthy gave a speech in which he said that Senator Benton was “worth a hundred million dollars to the Kremlin” (Oshinsky 1983). McCarthy also campaigned for Barry Goldwater, who defeated the entrenched Democratic floor leader, Senator Ernest W. McFarland of Arizona. In light of these four high-profile defeats, most criticism of McCarthy by his fellow senators stopped. When one senator, Herbert Lehman from New York, took the Senate floor in an attempt to challenge McCarthy's claims of Communist subversion, McCarthy growled, “Go back to your seat, old man.” According to Joseph Alsop, Lehman looked all around the chamber, appealing for support. He was met with silence and lowered eyes. Slowly, he turned and walked [back to his seat]. The silence of the Senate that evening was a measure of the fear which McCarthy inspired in almost all politicians… Old Sen. Lehman's back, waddling off in retreat, seemed to symbolize the final defeat of decency (Alsop 1968, 8-9). As journalist William S. White put it in his election postmortem: “the Senate now knows what it then [after the 1950 election] suspected: Senator McCarthy is a very bad man to cross politically” (1953). 3 McCarthy and other Republicans attacked Tydings, claiming that his committee’s final report whitewashed allegations of communists in government. At the time, political commentators—not just politicians—attributed the electoral outcomes directly to McCarthy’s influence. The New York Times concluded, “Senator McCarthy contributed a heavy part, if not perhaps the decisive part, to the defeat of Senator Lucas of Illinois.” It went on to declare McCarthyism, “a very considerable force in the Congress,” and branded McCarthy, “by any standard, the most politically powerful first term Senator in [this] Congress” (White 1951). A headline in Newsweek proclaimed: “Democrats Fume at McCarthy, But He Has Them Terrorized” (08/20/51). Historian Thomas Reeves (1982, 345) captures the contemporary conclusions well when he writes, “Virtually everyone, across the entire political spectrum, interpreted the elections of 1950 as, to use Sec. Owen Brewster’s term, a triumph for McCarthy and his ism.” McCarthy was widely believed to be a contributor to the victories of almost a dozen Republican senators and house members (Reeves 1982, 345; White 1951, 1953).4 Likewise, Robert Caro writes, “McCarthy’s support in elections for the Senate had been a fearsome weapon” (Caro 2002, 553). In large part, McCarthy was influential because politicians and pundits believed that the public sided with McCarthy. Walter Lippmann described McCarthy's power over public opinion as like he had cast a “spell” (Freeport Journal-Standard, 3/26/54, 11). Along similar lines, Caro 4 William White lists eight senate Republicans McCarthy helped elect, according to White’s sources in the Senate. This list, of course, is potentially endogenous, because it was constructed after the election occurred and may confound actual McCarthy influence with assumed McCarthy influence in particular elections. The McCarthy intervention measure we describe below, which is based on speeches, excludes three of White’s eight cases. However, even if we treat White’s list as an exogenous measure of influence, our main finding of no McCarthy effect remains the same (see Online Appendix 1). writes, “McCarthy's great weapon had been his mass support” (Caro 2002, 551). According to Caro, senate minority leader Lyndon Johnson failed to move against McCarthy earlier because he feared “[McCarthy] still had a dangerous and powerful hold on millions of Americans” (Caro 2002, 542). Another historian, Arthur Herman (2000, 184), wrote, “everyone on the Hill recognized he was a dangerous man to cross…because of his influence with voters.” Clearly, McCarthy developed a deep reputation as a fierce political opponent. Summing the conventional wisdom up well, an advisor to Lyndon B. Johnson recalled Johnson saying of McCarthy, “He’s the sorriest senator up here. Can’t tie his goddam shoes. But he’s riding high now, he’s got people scared to death some Communist will strangle ‘em in their sleep, and anybody who takes him on before the fevers cool—well you don’t get in a pissin’ contest with a polecat” (Caro 2002, 546). With the benefit of hindsight, we can ask if McCarthy deserves this reputation. Was he able to punish his political opponents where it hurt them most—namely at the polls on Election Day? The empirical record is mostly absent—almost no researchers have examined this question. In one of the few exceptions, Polsby (1960) examined county election returns in Connecticut and concluded that McCarthy, in fact, had little influence on the election outcome. Moreover, as some historians have noted, 1950 and 1952 were poor years for Democrats for reasons unrelated to McCarthy. For instance, speaking about the Tydings and Benton races, Griffith writes:5 5 Several historians have questioned the conventional wisdom about McCarthy’s electoral influence (though none conduct the kind of systematic analysis we do here). Oshinsky (1983), for instance, notes that William Benton of Connecticut was a relative newcomer to the Senate when he was defeated in 1952, having served only a portion of an unexpired term. Griffith (1987) refers to polling conducted during the 1952 election by Louis Bean which showed that the In retrospect it appears that Tydings was far more vulnerable than many had believed at the time. The sudden reversals in the Korean War, together with the general dissatisfaction with the Democratic administration, had taken a heavy toll. Maryland tended to vote independently in national elections, moreover, despite its heavy Democratic registration. Dewey won in the State's electoral vote in 1948, just as Eisenhower would in 1952 and 1956. The Democratic percentage of the two-party vote had been in steady decline ever since 1938 (1987, 129). Besides Polsby’s analysis, however, we are unaware of other systematic attempts to assess McCarthy's impact. The case of McCarthy, it seems, remains open. Perhaps McCarthy engineered the defeat of high-profile Democrats. Or perhaps, he was just in the right place at the right time—a vocal campaigner during a time that was bad for all Democrats. To properly assess McCarthy’s influence on the elections of the early 1950s, we need to construct plausible counterfactuals: What would the election outcomes have been in the absence of McCarthy’s intervention? Strategy of Analysis and Data In the rest of this paper, we use several strategies to construct this counterfactual. We first use regression models to forecast what should have happened in these elections had McCarthy Democratic senatorial candidates ran ahead of Adlai Stevenson in nonsouthern states. He concludes that McCarthy’s power was “an illusion.” See Rovere (1959) for more evidence of McCarthy’s power being overstated. not intervened. We conduct this analysis at the county level using 1938-1964 senate elections.6 Our dependent variable is the Democratic senate candidate’s share of the two-party vote. A vast literature in political science identifies the factors that shape senate elections, many of which we incorporate into our model. The first is the change in year-to-year partisan tides. These arise from numerous sources, but especially from the incumbent president’s popularity and the national economy (Hibbing and Alford 1982; Lewis-Beck and Rice 1985; Wright and Berkman 1986; Abramowitz and Segal 1986; 1990; Waterman 1990; Cohen, Krassa, and Hamman 1991; Shields, Goidel and Tadlock 1995). To capture the overall shifts in public sentiment towards the parties, we include indicator variables for each year (Year effects). To capture average support for the parties within each county, we also include county indicators (County effects). By including them, we eliminate variation across counties, so that we are only explaining variation in senate outcomes within each county over time. Next, we account for the advantages accrued to incumbent senators (Abramowitz and Segal 1990; Ansolabehere and Snyder 2002) by controlling for Incumbency, which we code 1 when the incumbent is a Democrat, 0 for open seats, and -1 when the incumbent is a Republican.7 Since some incumbents are appointed or win their seats in special elections, we include a separate similarly coded variable called Special incumbents. To capture trends in the partisan tendencies of each county, 6 We conduct the analysis at the county level because doing so potentially adds precision (e.g., if a McCarthy intervention corresponds with lower vote share for the Democratic senator in all counties, we can be more certain that McCarthy influenced votes). When we conduct the analysis at the state-year level, the results are similar. 7 We get essentially the same results if we use separate indicator variables for Democratic and Republican incumbents. we control for the average Democratic vote in the previous two presidential elections (Prior presidential vote) and the previous two senate elections (Prior senate vote) (Campbell and Sumners 1990; Klarner and Buchanan 2006; Highton 2000).8 To control for challenger quality, we include separate variables for whether challengers held previous offices, including senator, governor, lesser statewide offices, state legislative offices, and local offices (Abramowitz 1988; Lublin 1994),.9 We code these variables 1 if the Democratic challenger held the office prior to running for senate, 0 if no challenger held the position, and -1 if the Republican challenger held the position (Lublin 1994). Since these variables only capture challenger quality in races with incumbents, we include an indicator for open seats (Open). We also include several other variables, including the interaction of time trends with incumbency and state. These robustness checks leave the results unchanged and appear in Online Appendix 1.10 During the period we study, southern states were unusual in that they frequently elected Democratic senators with nearly 100% of the vote. We therefore exclude southern states from 8 Our model follows Highton (2000). 9 Downloaded in August, 2010, from www.american.edu/dlublin/research/data/cdlublin.zip. Since Lublin's data begin in 1952, we collected data for 1950 primarily from politicalgraveyard.com. 10 For additional work on modeling Senate elections, see Abramowitz 1981; 1989; 2006; Adams and Squire 1997; Bardwell and Lewis-Beck 2004; Born 1991; Carson 2005; Copeland 1989; Erikson 1990; Franklin 1991; 1993; Gerber 1998; Gomez and Wilson 2003; Grofman, Brunell, and Koetzle 1998; Hibbing and Brandes 1983; Highton 2004; Hill 1997; Jacobson 2005; Kahn and Kenny 1997; Kahn 1991; Kuklinski and West 1981; Miller 1990; Ragsdale and Rusk 1995; Segura and Nicholson 1995; Squire 1991; 1992; Tufte 1975; Westlye 1983. our analysis (though analyses that include the South are essentially the same as those presented here).11 The period we examine, 1938-1964, also includes an unusual period of Democratic dominance during the presidencies of Franklin D. Roosevelt and Harry S. Truman. To ensure that this period does not unduly influence the results, we also examine McCarthy's effect using only 1950-1964 elections. Finally, we weight our county-level analysis by the total number of votes cast in each county and cluster the standard errors at the state-year level. McCarthy Intervention Measure To determine which candidates McCarthy campaigned against, we searched biographies (e.g., Fried 1976; Reeves 1982; Griffith 1982; Caro 2002; Wicker 2006) and contemporary media accounts (e.g., White 1951; 1953) for races in which McCarthy gave at least one speech on behalf of the Republican candidate. This accounting, along with a rough measure of the extent to which state newspaper coverage of the races stressed McCarthy, is presented in Table 1. To construct the newspaper measure, we first counted all newspaper articles with last names of both major-party senate candidates published in September and October of the election year in the state. We then calculated McCarthy’s newspaper presence as the percent of articles that also 11 We classify the following states as southern: Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia. In the online appendix robustness table, we show that the results are the same when we only exclude former Confederate states. mentioned “McCarthy.”12 We also conducted the same newspaper coding in states where McCarthy did not intervene. As expected, we found little coverage of McCarthy in these states in 1950.13 [Table 1 about here] Some of the table’s findings fit reasonably well with historical accounts of McCarthy's activities. The four Democratic incumbents who McCarthy most famously intervened against and who lost their elections—Lucas, Tydings, McFarland, and Benton—show up in races where McCarthy made speeches and where he received at least some newspaper coverage. The table also highlights, however, races where McCarthy exerted a strong presence but the races themselves received much less discussion in historical accounts, including the 1950 Connecticut race, where he gave three speeches on behalf of Talbot, and the 1952 Indiana race, where over one-third of the newspaper coverage of the candidates mentioned McCarthy—the highest percentage of any race in the table.14 Admittedly, the statistics presented in Table 1 are blunt measures of influence. In particular, the 1952 newspaper coverage is much higher, largely because it contains false positives—stories about McCarthy's own primary race, which received sustained national attention. Nevertheless, they do suggest that McCarthy’s meaningful We conducted the search on Newspaper Archive in July 2012. At that time, this archive contained at least one major newspaper in about half of the states examined and smaller newspapers in the other half. 12 13 We did find some coverage of McCarthy in the newspapers in those states in 1952, but that coverage was exclusively about McCarthy’s 1952 primary race in Wisconsin. 14 McCarthy may have received so much attention in the Indiana race because the Republican incumbent, Jenner, was seen as a key McCarthy ally. involvement in the senate elections of the early 1950s extended beyond the four races that are most prominent in historical accounts. To study McCarthy's effect, we created a single indicator variable for the 12 races in which McCarthy gave at least one speech on behalf of a senate Republican candidate (McCarthy intervention). We also examined his effect on these races individually by creating separate indicator variables for each of the 12 races (e.g., MD 1950). Since McCarthy only intervened against Democrats, negative coefficients on these variables indicate that he harmed Democratic candidates. Results—model-based counterfactuals Having established the scope of McCarthy’s involvement in the senate races in 1950 and 1952, we next seek to assess the impact of McCarthy’s speeches on the electoral outcome. We begin our analysis by running a simple regression of county-level Democratic share of the twoparty vote on the McCarthy intervention variable for the 1938 to 1964 period. As Table 2 demonstrates, the estimated effect is -3.3 percentage points, implying that McCarthy lowered the vote share of Democrats he campaigned against by that amount (column 1). However, including county and year fixed effects to account for temporal and spatial electoral swings switches the sign of the effect to a positive 0.78 percentage points, suggesting that, if anything, McCarthy slightly increased the vote share of his Democratic targets by just under one percentage point, though the effect is imprecisely estimated (column 2). McCarthy’s apparent influence may thus simply have resulted from 1950 and 1952 being difficult elections for Democratic candidates. Including measures of incumbency and prior presidential and senate vote produces an even smaller estimated effect, though the direction of the effect becomes negative again, -0.50 (column 3). Since the 1938-1948 elections are from the era of Democratic dominance of national elections, as a robustness check we next rerun the analysis only on 1950-1964 senate elections (column 4). The results also remain more or less unchanged, with little sign of a McCarthy effect. To account for challenger quality, we include the prior experience variables described above.15 Unfortunately, we only have this data from 1950 to the present. Since a comparison of columns 3 and 4 indicates that the basic result was insensitive to the period, we include in column 5 the challenger quality measures, restricting the analysis to post 1950.16 The challenger quality variables flip the sign on the intervention variable to positive, but the McCarthy effect essentially remains zero. In sum, on average, McCarthy’s interventions did not appear to affect the election results of the early 1950s. [Table 2 about here] Perhaps, though, this average effect conceals important heterogeneity. As Table 1 demonstrates, McCarthy intervened in a variety of races, but he likely exercised more effort—as 15 Since McCarthy's attacks against incumbent Democrats may have attracted higher quality Republican challengers, challenger quality is potentially endogenous to McCarthy's efforts. As we note below, however, the results remain essentially the same with and without the quality controls. Furthermore, if McCarthy stimulated stronger Republican challengers, the estimated McCarthy effect we present will capture the consequences for Democratic vote share. Without controls for candidate quality, the estimated McCarthy effect captures both his indirect effects through candidate quality and his direct effect on the election itself. Nevertheless, we find no overall McCarthy effect, indicating that even if McCarthy was able to stimulate strong challengers in particular races, those actions had no effect on the election outcome. 16 The online appendix presents robustness checks using a simple dichotomy for candidate quality as well as other robustness checks. measured by the number of speeches and the press coverage—in some races than he did in others. To disaggregate the average effect, we next replace the single McCarthy indicator variable with a series of 12 indicator variables, one for each race in which he attempted to exert influence. Table 3 presents the estimates. Our analysis here mirrors the specifications in Table 2. For ease of exposition, Figure 1 graphically presents the coefficients and confidence intervals for each of the races using the most complete model (column 3 of Table 3). In some cases, McCarthy’s efforts were associated with a statistically significant Democratic underperformance in the election. In others, however, his efforts were associated with a statistically significant Democratic overperformance. In fact, in 6 of the 12 cases, McCarthy appears to have helped his Democratic targets, not hurt them, as the Democrats actually did better than expected based on the regression prediction. Consider, for instance, the 1952 Michigan senate election. There, the Republican challenger Charles E. Potter defeated the Democratic incumbent Blair Moody in a close election.17 The regression analysis indicates, however, that, based on prior vote patterns and the temporal tides of the 1952 election, Potter should have won by a wider margin than he did. [Table 3 about here] [Figure 1 about here] Even some of the races that helped solidify McCarthy’s reputation as a political force are suspect.18 Although the estimates suggest that McCarthy’s targets fared better than expected as often as they fared worse, Figure 1 shows that two of McCarthy's targets did fare especially 17 Moody was appointed to the Senate only the year before the election. 18 In fact, Democrats do slightly better in states where McCarthy received more news coverage, based on the measure in Table 1. poorly: Tydings in the 1950 Maryland race received 11.5 percentage points less vote share than expected and McFarland lost to Goldwater in the 1952 Arizona race by 10.2 percentage points more than expected. Could McCarthy have influenced the outcomes in these races, even if he failed to influence the others? Making strong inferences about any one race is difficult, but there are reasons to think he did not. In particular, idiosyncratic factors in Maryland and Arizona probably explain the apparent McCarthy effect. In the Tydings race, an unpopular Democratic governor topped the ticket, likely exerting a considerable drag on other Democrats such as Tydings, who actually outperformed the governor (see Table 4). In the McFarland race, 1952 was simply a bad year for Democrats in Arizona and McFarland actually did better than other Democrats, receiving more vote share than the Democratic presidential candidate and gubernatorial candidate (see Table 4). All told, examining individual races reinforces our conclusion that McCarthy exerted no reliable effect on his targets. [Table 4 about here] One scenario not accounted for in our analysis is that McCarthy might have changed the overall dynamics of the Senate races of 1950 and 1952.19 In particular, his attacks may have harmed Democrats nationwide, not just those he intervened against. If so, our tests may be biased against finding a McCarthy effect. To check the robustness of our findings against this threat, we estimated a modified version of the model in Table 2, column 3, but without 1950 and 1952 data—the years McCarthy intervened. We then forecast from this model Senate vote share in these years, subtracted this forecast from actual Senate vote share, and tested whether 19 We thank Robert Van Houweling and David Mayhew for bringing this alternative to our attention. McCarthy's interventions influenced this residual. Because we want to predict vote share in the 1950 and 1952 elections without using data from those years, we exclude from the forecasting model variables that rely on data from them, including prior Senate vote and, of course, year fixed effects. In their place, we included three new variables that capture year-to-year partisan swings: real disposable income growth in the election year (recoded to favor the Democratic Party); the interaction between real disposable income growth and a midterm indicator (since the economy seems to matter less during midterm years); and an indicator variable for Democratic versus Republican president during midterm years. We also limited the model to 1948 onwards because income growth during World War II was unusual. We present this model in the Online Appendix 2. Column 1 presents the results of the general forecasting model. Column 2 tests for the McCarthy effect on residualized Senate vote in 1950 and 1952. It regresses the residuals from column 1 on the indicator variable for McCarthy intervention. Consistent with our earlier analyses, we find no sign of a McCarthy effect. In fact, the Democratic senators he intervened against did better (0.09%) than expected, not worse. Our main finding is therefore robust to excluding 1950 and 1952 elections. All told, our analysis suggests that McCarthy’s power was illusionary. Results—other races as counterfactuals To estimate McCarthy's influence, we thus far have relied on a model-based approach. We have compared how McCarthy's targets fared to how the models predict they should have fared. This approach comes with a cost; if the models are wrong, the estimates are likely biased. Fortunately, we can also use a non-model-based approach to answer the same question. Specifically, we can compare how the candidates targeted by McCarthy fared relative to other Democratic candidates in the same state who were not targeted by him (for example, Democrats running for lieutenant governor or for the House). We can use these outcomes as counterfactuals in place of the model-based extrapolations we used in the previous section. State-level downballot races, such as lieutenant governor or attorney general, are especially useful counterfactuals because most voters probably know little about the candidates in these races, and so vote totals probably reflect general support for the parties, not the particular candidates. Using this non-model-based approach to estimate McCarthy's influence yields the same conclusion: we find little sign that McCarthy's attacks mattered. Table 4 shows these results. It presents comparisons with the Democratic vote average for state-level down-ballot races (where available: lieutenant governor, attorney general, secretary of state, and state treasurer), presidential elections for the 1952 races, US House races, and gubernatorial races (where available). Consider the Lucas race in 1950, one of McCarthy's key targets in that year. Although Lucas lost the race with 46% of the vote, the average state-level down-ballot Democrat did even worse, garnering only 44% of the vote. Furthermore, reading across the row, we see that Lucas did no worse than the average Democratic house candidate (46%) and three points worse than the Democratic gubernatorial candidate (49%). Each row in the table provides as many as four possible counterfactuals—including president in 1952. The table’s second to last column shows the average percent Democratic vote for all the available counterfactual scenarios, which in this case was 46%, identical to Lucas’s 46%, and the final column shows the difference between these (i.e., 0%). By averaging this last column across all states, we can compare the average electoral support for Senators targeted by McCarthy to the average for other Democratic candidates in the same states. In the third row from the bottom of this column—the row labeled “ave”—we see that McCarthy's targets actually did slightly better than other Democrats, beating the average by one percentage point. This effect varies somewhat by year: -1% in 1950 and 3% in 1952. But the underlying message is clear: we see little sign that McCarthy’s campaigning mattered. One concern with the results presented in Table 4 is that McCarthy may have hurt all Democrats in these states, not just the Democrats he targeted. We explored and rejected a related concern about a general McCarthy effect at the national level in our regression-based analysis above. In the second-to-last row of Table 4, we address this related concern using our nonmodel-based approach by showing the average Democratic vote for each category of races in states where McCarthy failed to intervene. Surprisingly, it shows that the Democratic senators McCarthy targeted did better than Democratic senators in other states, 49% versus 47%, respectively. (Although, non-Senate Democratic candidates in other states generally fared somewhat better than did Democrats in the states he targeted). In sum, we do not see a clear sign that McCarthy exerted a general anti-Democratic effect on the states where he intervened. Discussion Although many prominent politicians at the time knew McCarthy had fabricated his explosive allegations, they feared him and failed to stand up to him. Their fears may have stemmed from their interpretation of the 1950 and 1952 elections. Seeing Lucas and Tydings fall to McCarthy in 1950 cowed many into silence. McFarland’s and Benton’s defeats in 1952 further deepened their reticence. In this paper, we investigated whether this fear of McCarthy was well founded. Our analyses found no evidence that McCarthy reliably influenced the outcomes of the 1950 and 1952 senate elections. McCarthy's reputation thus appears to have been undeserved, arising from selective attention and a failure to appreciate voters’ anti-Democrat mood. These findings are robust. They hold using model-based estimates of Senate election outcomes. They hold when we simply compare outcomes to other elections held in the same state. Finally, they hold when we estimate the models without election data from 1950 and 1952, treating the elections in which McCarthy intervened as out-of-sample predictions. Although some candidates McCarthy targeted did worse than expected, just as many candidates did better (see Figure 1). The McCarthy case therefore shows that politicians can greatly err in their interpretation of elections. These findings speak not simply to the particular case of McCarthy, but also to the larger question of government responsiveness in a representative democracy. In a system where elections yield ambiguous outcomes subject to multiple readings, politicians’ interpretive competence is critical. The politicians of the 1950s seem to have misread McCarthy’s campaign strength, crediting him with more political influence than he actually had. Their subsequent fear of McCarthy and failure to stand up to him may have generated a self-fulfilling prophesy that awarded him undue power and helped create, in Robert Sherrill’s (1983, 11) words, “the most influential demagogue the United States has ever produced.”