The American Public’s Attention to Politics in Conflict and Crisis, 1880–1963 Free

Questions about when the public held what attitudes can be surprisingly difficult to answer. Surveys and polls are quintessentially obtrusive measures. They prompt people to think about issues in a way that is unlikely to occur in, or reflect, typical experience. Even if we accept the potential pitfalls to external validity, polls have not always been available—certainly not in war zones and, in any case, not until the 1930s, from which time the first systematic studies of public opinion are usually dated. They also tend to be retrospective; pollsters’ questions are limited to what they thought relevant before going out into the field. Yet public opinion, being central to many social and political processes, is important for understanding historical outcomes. It is also central for analyses seeking to examine the lived experience and worldview of those beyond the elite classes of society. It is therefore useful to have other measures of mass attitudes, besides polls, to extend and confirm traditional survey results.¹

One such alternative gauge of public feeling is the incidence of baby names. Although this measure has its own drawbacks, it is available in a variety of places and times. Moreover, naming involves real-life stakes and consequences that do not attend survey responses: Even though an unfortunate name could impose lifelong costs on a child, personal feelings exert an undeniable influence on baby-name choices. Names are thus a promising place to explore questions about what people consciously or unconsciously think in their daily lives. They can provide insight into how long news events command the public’s attention—the extent to which historically important events and circumstances insinuate themselves into contemporaneous, everyday life.²

The empirical analysis in this article focuses on the United States during the late nineteenth and early twentieth centuries, specifically on the naming of babies for presidential figures (including, at times, presidential candidates). It first shows that names can serve as a useful historical indicator, reliably reflecting major political events. For example, matters relating to the presidency have measurably affected the names chosen for children under many circumstances, such as elections and crises involving conflict and war. These choices help to illuminate how people in the past reacted to events, as well as how general political sentiment related to attitudes regarding the head of state. They also demonstrate some of the possible uses of naming practices, as often explored in anthropological and sociological settings, for addressing historical questions throughout the social sciences.³

Personal names have historically derived from many sources, from cultural trends to deep matters of parental identity. Names therefore preserve for posterity what might otherwise be relatively fleeting social concerns. Consider Figure 1, which presents data from the 1920 U.S. census about the years of birth (inferred from reported ages) of people whose first name is “Maine.” This unusual name briefly surged just as a slogan urging people to remember an exploded ship of that name emerged in the press. Some commentators doubt that foreign-policy crises can inflame public opinion to any calculable extent, but at least a few people were sufficiently moved to commemorate the sinking of the Maine in a lasting way. Even though by 1920, succeeding events—not the least of which was World War I—had supplanted the tragic deaths of hundreds of sailors in Cuba, those named “Maine” born in 1898 still manifested the earlier flicker of public concern. The relationship between political events and choice of name is not always so direct and literal, but names can reveal what is on the mind of at least some of the public at a given time and place.⁴

Name givers are, clearly, not a random sample of the general population. The ill, the unmarried, and the elderly are all less likely for various reasons to have, and thus to name, children, although they may indirectly influence the people who are selecting child names. Moreover, because naming is often perceived to be a weighty, abiding act, especially in the small families of recent decades, it can evade the vicissitudes of daily events to reflect deeper dimensions of identity. As the “Maine” example shows, however, even transient episodes can hold deep meaning for a while, and many people seem willing to name children after situations that were obscure to them just a few weeks previously. Since the names of prominent people and events tend to remain well known in the future, names based on the news of the day do not necessarily fade into obscurity.⁵

Time-sensitive factors are not the only, or even the primary, influence on names. Name choices also derive from parental ideology, ethnic background, and socioeconomic status. As a case in point, one strand of French studies explores which parents were likely to give their children “revolutionary” rather than traditional names during the late eighteenth century. Moreover, some of the deeper and demographic differences in naming practices may reflect varying sensitivity to recent events. People with a more traditional orientation are less apt to adopt fashionable, event-sensitive names than are those who are less conservative. These predictable patterns suggest that unusual choices of name reflect conscious thought, despite their emotional component. Even short-term responses inform an understanding of the psychology and sociology of public attitudes.⁶

The sometimes-ephemeral nature of names implies that annual data like that used in Figure 1 may be too blunt a tool for capturing public responses. Since public attention is likely to shift on scales much shorter than a year, data with a finer temporal grain can better pinpoint how the public reacted to news events. The case of people in the United States choosing to name their children after presidents is compelling evidence of baby names as indicators of day-to-day public sentiment. Although the person of the president may affect people’s lives less directly than do such policy outcomes as economic conditions or wars, the president is the standard bearer for personalized politics. A president’s ability to shift public opinion suggests an intimate connection with people’s thoughts. The question is when are presidents most on the minds of their people, in times of relative quiet or in times of crisis? Previous literature suggests several contexts that are likely to increase public attention to the president.⁷

One important factor is media attention. Presidential activities are widely covered, but those that attract the most reporting have the greatest import. The American media tends to favor political competitions with distinct winners and losers. The focus on presidential candidates, especially as an election approaches, is particularly intense. From the perspective of the press, campaign events have the additional advantage of set dates, allowing reporters to be dispatched in advance. But spontaneous events can also draw public attention to a president, especially during crises; threats to the nation tend to rally public support for the country and, by extension, its political leadership. Besides bringing people together behind national symbols, crises also demand immediate reaction, which plays to the executive’s strength, in contrast to the inherently more deliberative, and hence slower, legislature. The expansion of the presidential role is especially large if the crisis is military and the commander-in-chief function comes to the foreground.⁸

Knowing that some events are likely to bring (positive) public attention to the president, however, leaves several questions unanswered. How long do these sorts of events dominate public concern or linger in the memory? Are people more apt to personalize the presidency in crises, or is mere domination of media coverage more important? Baby names can help to provide insight into the answers.⁹

The source for the names used in this article is a 2011 version of the Master Death File, a central database intended to allow screening for fraud, which contains the United States Social Security Administration’s compilation of deceased holders of social-security numbers. The overwhelming majority of Americans have had these codes since the 1930s; members of a few exempt religious groups and those who died as children before the mid-1980s are the primary exceptions. Since the data include nearly all of the Americans who have been alive since World War II but died before 2011, they cover predominantly those residents of the United States born in the decades around the 1910s. The coverage is progressively less extensive for those born earlier (who often died before implementation of the Social Security Act) and those born later (who were more likely to be alive in 2011). Each year, from 1876 to 1965, at least 100,000 people in the file were born, providing a large enough sample to observe even uncommon names and maintain a roughly constant number of total childbirths per day within any given year.¹⁰

From these data, the dependent variable is the number of people born on a day whose given name matched that of a president (or related person of interest). Determining matches requires some judgment calls. The analysis treats middle names as first names when they are the more commonly in use; for instance, (Thomas) Woodrow Wilson is “Woodrow” and (John) Calvin Coolidge is “Calvin.” Likewise, for better comparability of presidents, only official names, rather than such nicknames as “Cal” or “Ike,” apply. Additionally, only perfect matches in spelling count: “McKinley” matches the president’s name, but “McKinlay” does not.

Many people share the name of a president at the time of their birth by mere coincidence. Presidential names are often common; duplications could derive from any number of alternative sources. The interest comes via systematic fluctuations from the background noise of those other naming sources, which are unlikely to correlate with salient, presidential moments. Thus, as in Figure 1, the key point is whether the choice of a name within the population diverged from its average baseline. To make this determination, the models below employ the methodology of the event study—a common design from economic and financial analyses. The independent variable is the number of days after a key event when an observation occurs. For example, the date of an election serves as day zero; the day before it is day negative one; the day before that one is day negative two, and so on. Conversely, the day after a death is day one; the next day after that one is day two, and so on, until every date in the frame has a number. In this analysis, the frame extends from a week before the relevant event to four weeks after it (other frames produce results similar to those reported).¹¹

To observe the effect of deaths over time, a dummy variable is created for all days zero; this indicator models the average outcome variable on the date of each event. Another dummy variable indicates days one (that is, the days immediate after each event); another indicates days two, and so on for all of the days within the frame. These are the independent variables of interest in the model. A statistically or substantively significant effect implies that a given position in the frame—say, two days after whatever event is under analysis—associates with a crop of newborns with a name count diverging from the average. This setup requires care in the selection of the events that are to qualify. Models below consider three kinds of events:

Elections offer a variety of motivations for parents giving their children political names—admiration for a candidate; partisan affiliation; or triumphalism, if the namesake candidate has emerged as victorious. Presidential elections are particularly convenient events to study in the United States since they occur on a single, well-defined day (although sometimes, as in 1876, definitive results have to wait). They also bring fame to multiple people, including losing candidates. Electoral winners and losers alike are often prominent, becoming household names, but the distinct advantages of victory are not to be denied. That is to say, elections produce not only attention to politics and to specific personalities; they also cue parents to announce their partisan loyalties, especially when successful, in their choice of babies’ names. The election data herein cover the presidential elections from 1880 to 1960.

National crises can have multiple effects. One traditional response to such crises is to rally around the executive as a symbol of the nation, but crises can also imply leadership failure, or create security or economic concerns that could distract attention from the president. The analyses in this study look at two classes of crisis—foreign policy, which is generally associated more with the president as a national symbol than domestic policy, and presidential assassinations (or attempts at them), which are the ne plus ultra of personal politics, garnering overwhelming support for a (recently shot and possibly deceased) president.

Crises present greater challenges of definition than do elections. It is not always easy to determine exactly when crises occur. Is the critical moment in an assassination the shooting or the death? When does an encroaching belligerent situation rise to the level of a crisis? This analysis denotes an assassination as the day of a shooting, even if the president died later, and a foreign-policy crisis as a formal declaration of war, plus United Nations Security Council Resolution 82, which initiated America’s involvement in the Korean War.¹²

Since analysts usually consider graphs better than tables at conveying quantitative information, the results of this study are in graphic form, showing averages and 95 percent confidence intervals. Figure 2 considers the popularity of winning candidates (left panel) and losing candidates (right panel) as demonstrated by the naming of newborns in the days around the election. Candidates have multiple names—given names and family names and, in some cases, middle names. Accordingly, the figure provides separate analyses for each of these three types of name. Because these different types are of widely varying frequency (presidential first names are, as standard given names, much more common), each class is normalized for ease of comparability; the average prevalence of a name in the two weeks before an election has a value of 100. Hence, an estimated value of 200 indicates that a name has become twice as common as at the pre-election baseline (which may include some pre-election naming of babies after the candidates).¹³

The first, middle, and family names associated with winning candidates see a surge around their election. Indeed, this boost is statistically distinguishable from the baseline level—as well as from losing candidates’ names—several days before the election, either because parents assigned names a few days after birth or because they had confidence in their candidates’ prospects. The effect is, however, strongest on Election Day itself, when victors’ surnames rocket to five times their usual frequency. Because surnames tend to be relatively uncommon as given names, estimates about surnames as given names have relatively wide confidence intervals. Yet the incidence of winners’ first names also increases dramatically on Election Day, estimated, on average, to be three times the baseline figure. Even electees’ middle names, typically only a weak association with presidents, are statistically more common on the day of the election than they were previously.¹⁴

By contrast, losing candidates’ names show far less movement. The slight upticks from the baseline that they show on Election Day are not large enough to achieve statistical significance in a traditional event-study setup. Moreover, these effects, such as they are, vanish quickly after elections. Despite their extreme familiarity, especially in the news around elections, the names of losing candidates receive relatively little recognition from name-givers. Evidently, the use of political names is not simply a by-product of familiarity; it reflects partisan triumphalism or, perhaps, a bandwagon effect.¹⁵

Figure 3 displays the reactions to crises; shootings are on the left and foreign-policy issues on the right. Because such events do not happen very often, they do not produce meaningful results for presidential surnames, which are sparse in the data, anyway (estimates have very wide confidence intervals). Hence, the figures present aggregate results for every class of name.

The shooting of incumbent presidents does not produce a spike in naming close to that seen at elections. The estimated increase in name usage never exceeds 50 percent (due, in part, to the relevant presidents having the common names of James, William, and John, which leave little room to expand). However, the increase lasts much longer. Rather than dissipating in a matter of days, the names of presidents who were victims of shootings are still appreciably above the baseline for several weeks. This timeline in part reflects the contingency that two of the three presidents lingered for some days after their shooting; the highest peak came in the aftermath of President McKinley’s death, just over a week after he was wounded. Nevertheless, it implies that popular attention to the president, and to political news, can endure for weeks.

Foreign-policy crises also cause a noticeable increase in the use of a president’s name—not, however, at the moment when war was declared but a few days later. Rallying around the president does not necessarily occur at the same time as a formal declaration of war, even though the sudden attack that launched America’s involvement in World War II (one of the four wars in the period) produced immediate support for the nation’s leader. Also notable is that the declarations of war, like the shootings, produce an increase in iterations of the president’s name that are, though statistically significant, much smaller than that associated with elections or inaugurations. Assassinations and wars may not always spur positive emotions and good will toward a president, and the preoccupations of new parents might not always permit a full awareness of a sudden outbreak of war or an assassination.

Baby names thus provide a measure of the public’s reaction to political events. More ambiguous is whether they indicate conscious approval of a chosen namesake; after all, people may not realize why a name sounds familiar or appealing when they settle on it. As noted above, the differential election effects for winning and losing candidates suggest that more than exposure is at play. Candidates may win office simply because they are more famous. Directly testing whether naming equates with public approval requires comparison with a data set containing information that clearly measures presidential approval.

Survey data can be helpful in that regard. Owing to the relatively late origins of modern polling methods, such data emerged only after the period for which the Master Death File has its fullest coverage. Consider, however, the sixty-five Gallup polls of presidential approval that the Roper Center reported for the Truman presidency. During the course of any individual poll in this set, which lasted several days, at least 4,000 children were born, of whom a certain proportion were named “Harry” or “Truman.” For example, during the first survey, from June 1 to June 5, 1945, 5,351 people in the Social Security Master Data File were born, fifteen of whom received one of the presidential names. Across the polls, the correlation between presidential approval and the share of babies receiving presidential names is 0.43, with the p statistic (on the null hypothesis that the correlation is zero) being less than 0.01. Similarly, in regressions predicting the share of babies named “Harry” or “Truman” during each poll’s duration, the favorability rating takes a positive and statistically significant coefficient (with or without year-fixed effects in the model). The implied effect is that a standard-deviation increase in presidential approval produces an approximately half-standard-deviation increase in names related to the president. Figure 4 is a graphic representation of this relationship.

Although this correlation is only moderate, it strongly suggests that the two measures capture some of the same underlying sentiment. President Truman represents a useful case because of the frequent polls during his tenure, but the Truman administration may overstate the noisiness of the birth-name measure. Because most people born during the Truman administration were still alive in 2011, hence excluded from the data herein, the number of daily observations is lower than it might be, producing a less precise measure. These excluded people, furthermore, are not a random selection; they are the segment of the population who had not lived long yet. They might have had socioeconomic or demographic characteristics that differed from those of the general population. Thus, they might have reduced the correlation between the two measures of attitudes toward the president, especially because names also show distinct linkages to parental race and education. Nonetheless, presidential names, though not a perfect analog to poll results, appear to provide a window into public sentiment.¹⁶

The primary value of having an alternative measure of public sentiment lies not in eras and contexts that had more precise, traditional measures of public opinion but in those that did not have them. For those periods, baby names provide one instrument for examining how people felt and reacted to ongoing events. They therefore deserve consideration as systematic evidence for popular reactions to otherwise unclear events.

The massacre of hundreds of Lakota Sioux at Wounded Knee, South Dakota, on December 29, 1890, was a tragic exclamation point in the already grim nineteenth-century relationship between the United States government and its indigenous people. Occurring long after the end of widespread fighting in the Sioux Wars, that bloodshed seems in retrospect particularly grotesque and gratuitous. Its costs were not merely the Native Americans and federal soldiers who were killed or wounded but also an enduring legacy of trauma and distrust. Accounts of public reaction at the time varied. The contemporary media generally framed the incident as a noble triumph for the soldiers. The headline in the next day’s New York Times—“A Fight with the Hostiles”—typified both the tone of the press and the initial perception of the incident as an encounter with armed foes—the sort of situation likely to produce a rally of patriotic support for the political leadership. The racial animus and suspicion toward the victims continued for some time. However, a simultaneous countercurrent of disquiet and outrage was evident. Within days of the bloodbath, rumors reached Washington that Wounded Knee was less a pitched battle than the slaughter of largely unarmed women and children. By the early 1890s, after a string of similar massacres, public opinion “no longer automatically took the white man’s side.”¹⁷

These accounts imply divergent attitudes among the late nineteenth-century American public, with different implications for their understanding of the policy and pressure that the government faced. These differing stories also complicate the literature about how conflicts and military atrocities affect public attitudes. Resolving whether Wounded Knee produced more approbation than revulsion, and vice versa, is thus of considerable value. Since the time of the event antedates scientific polling methods by decades, standard modern survey methodologies are not available or applicable.

The evidence presented in previous sections shows that crisis and conflict usually creates moderately enhanced presidential support; the president becomes the personification of military action. Nevertheless, attitudes toward an issue or event such as Wounded Knee and the overall evaluation of a president do not necessarily coincide. Presidential approval is a multidimensional phenomenon influenced by everything from the economy to a vague mood surrounding the country as a whole; factors completely independent of the military efforts against the Lakota may well have informed public opinion to one extent or another. Conversely, even if people feel strongly about an incident, they may not connect it directly to the president, or they may find the very idea of naming a child after a president puzzling or unseemly. But approval is tied to the salient issues of the time, especially those covered by the media. The events on the Pine Ridge Reservation attracted so much media coverage that the Harrison administration was compelled to express regret about the bloodshed within a week of its occurrence. Moreover, the reputation of a president can depend largely on the effects of conflict; even the secondary consequences of military campaigns tend to be important for presidential approval. Thus, the evaluation of President Harrison’s term in office likely had much to do with how the populace viewed Wounded Knee.¹⁸

The names that suggest positive sentiment toward Harrison and his administration are “Benjamin” and “Harrison.” In this period, around six or seven newborns received those names each day out of the approximately 2,000 daily births that eventually entered the Social Security name list. The left panel of Figure 5 gives the (smoothed) share of births featuring either of these names in the weeks around the date of Wounded Knee; the massacre is marked on the figure with a vertical dashed line (first reports of the massacre did not appear in many newspapers until the following day at the earliest).

As the figure shows, days in late December averaged, with some variations, around 2.9 people named “Benjamin” or “Harrison” per 1,000 births. At the end of the month, exactly when word of the Wounded Knee massacre was spreading, the proportion of these names dropped swiftly, remaining for most of January at a level approximately one-quarter lower than the December level. Rigorous hypothesis tests confirm the informal impression of the graph. In a regression predicting the daily rate of presidential names throughout the range shown in the figure, the predicted rate of “Benjamin” or “Harrison” births days after the massacre is 0.5 to 0.7 per 1,000 lower than days before—depending on whether the specification includes a lagged dependent variable, fixed effects according to day of the week, or other control variables (two-tailed p 0.05). Nor is this some seasonal phenomenon in which, say, “Benjamin” is always less common among January births, because of an association with particular saints’ festivals or name days. The right-side panel of Figure 5, which shows the pattern of “Benjamin” or “Harrison” births for the other three years of his term, indicates no comparable decrease in presidential names at analogous times. Regressions using data from other years finds no comparable drop in presidential names after December 29.¹⁹

Again, this finding hardly proves that the public’s general reaction to Wounded Knee was negative and anti-government. Other events might have provoked the observed pattern. Yet it points to the value of using names to gauge the sentiments of populations who may not have left behind diaries or had a direct voice in media coverage. The evidence, in this case, hints at popular views regarding the nation’s conflict with indigenous groups and the use of military force, important macrohistorical trends for which direct evidence most often comes from newspapers that are unlikely to represent public opinion accurately. Name-based evidence allows a separate vantage point from which to triangulate such material.

Although names given at birth provide a helpful window into past preoccupations and attitudes, they cannot answer every question that social scientists can raise about public opinion because they pertain only to a particular segment of the population, and they are likely to respond only to the largest shocks. Nevertheless, they are a useful complement to surveys and other traditional measures of public opinion. Naming practices are also interesting in their own right for what they reveal about past societies, individuals, and values. Child-name data is available from multiple sources in modern, bureaucratic societies. Many civic, medical, and religious institutions keep daily records of local births that are available in archives. National censuses may not keep daily records, but they often include month or year of birth from which broader trends can be computed; similar information can sometimes be gleaned from educational institutions’ registries.

This study focuses primarily on short-term naming reactions to a few types of event. It finds often substantial, but generally short-lived, jumps in the use of presidential names around major political events in the United States, be they elections or crises, though public attention was not captivated by them or the presidency for long. Further research could explore many other elaborations of this method. It could look at monthly (or, as in Figure 1, yearly) naming patterns for changes in public opinion before polls were common or scientific. Because most leaders’ last names are not common choices for first names, changes in their occurrence, one way or another, during their namesakes’ administrations probably reflected attitudes toward an incumbent, most likely in the face of scandal or geopolitical change. Variations in naming practice in general can be telling. For example, in the English-speaking world, German names—especially “Adolf” or “Adolph”—declined precipitously during the world wars.

This method is capable of broad application, both geographically and temporally. Many countries, not just the United States, have records that include birthdates as well as names that could be helpful in comparative analyses. For instance, the already digitized 1910 Norwegian census can show how names changed in the days surrounding Norwegian independence, thus providing a measure of patriotism. Similarly, such data can track the influence of events and figures outside politics: Did Jenny Lind or Florence Nightingale earn commemoration for their news-making achievements with a surge in namesakes? Celebratory names of this sort might be geographically concentrated (around Lind’s concert tour sites or Nightingale’s nursing school in London), leading to telltale spatial patterns. Furthermore, bounces in name popularity centered on political events might benefit from comparisons with, say, those surrounding religious features, like saint’s days. This methodology, like many other unconventional ones, holds promise for casting new light on a wide variety of issues.²⁰

Another step is to discern more about the parents who took political inspiration for their naming practices. Although the Master Death File does not contain the relevant data, it offers sufficient clues to link most people with records that lead, directly or indirectly, to information about families’ origins and socioeconomic status. Comparing all families with a child born on Election Day, say, and identifying the factors associated with naming children for victors could result in deep and detailed insights about identities and beliefs that would otherwise remain unknown.

The public opinion to which naming testifies can also serve as a constraint on elites, guiding or channeling government policy. As an alternative source of information about the issues that matter in people’s daily lives—often involving people and social groups who are absent from the historical record—personal names supplement previous methodological approaches to key questions concerning public opinion.²¹