Only 15 Minutes? The Social Stratification of Fame in Printed Media

Sociology of Fame

In colloquial speech, fame is typically used only in reference to celebrities who are talked about by millions of people across vast geographical areas (e.g., Queen Elizabeth, Tom Cruise, or Karl Marx). As a sociological concept, fame is defined as indicating a position on a continuum anywhere between being known only by family, friends, and colleagues and being world-renowned. According to Currid-Halkett (2010:29, 66), fame is “pure renown—literally the sum of all people who have heard a person’s name. . . . [It is] fundamentally about sheer numbers of people who know one’s name . . . measured by quantity of recognition.” This definition is compatible with Braudy’s (1986:608) account of fame: “In its root sense, fame means to be talked about.” Conceptualization of fame as the volume of public discourse about a person is also consistent with the Oxford English Dictionary (2nd edition, 2011) definition of fame as “the condition of being much talked about.” Sociologists explicitly do not reserve the term just for individuals who merit public attention because of some extraordinary accomplishment. Fame is used indiscriminately for people who are loved and adored and for those we collectively fear or hate (i.e., notorious or infamous persons). Finally, as a continuous concept, fame must also include the many intermediate cases from the middle and lower echelons of the public attention hierarchy, such as “local newscasters, minor league athletes, or local politicians” (Ferris 2010:393).

Three elements make up the predominant perspective in the contemporary scholarship of fame: a mechanism of self-reinforcement and distributional patterns of extreme inequality and high mobility.

Self-Reinforcement

Perhaps the best-known statement on the self-reinforcing nature of fame is Boorstin’s (1961:57) definition of “celebrity” as a person who is “known for well-knownness.” As an example of this circular causality, he notes how “endorsement advertisement not only uses celebrities; it helps make them” (p. 58). In Boorstin’s view, celebrities’ “chief claim to fame is their fame itself. They are notorious for their notoriety” (p. 60).

In more recent work, Cowen (2000:14) speaks of “snowball effects,” whereby a “small initial burst of support for fame can lead to cumulative and self-reinforcing support over time.” Cowen identifies several mechanisms through which public attention may expand. When a performer shows signs of ascent, fans will jump on the bandwagon. In Rosen’s (2001:151) words, “the more others know about someone, the more it pays you to know about that person too.” In addition, “fame-producing institutions serve to strengthen reputational snowballs” (Cowen 2000:15). Fan clubs, distributors, and bestseller lists are examples of such fame-producing institutions.

Kurzman and colleagues (2007) argue that these cumulative effects continue to occur once celebrity status has been reached. This is partly because the public becomes interested in missteps and scandals involving famous individuals: “Publicizing the peccadillos of celebrities seems to help reinforce their celebrity” (p. 353). Moreover, special treatment further increases the distance between the known and unknown: “The privileges that we grant them in person help to reinforce their superior status” (pp. 355–56).

In summary, the sociology of fame suggests that fame is subject to self-reinforcement whereby every increase leads to a greater chance for recognition in the future. This idea embodies the more general sociological concept of cumulative advantage (DiPrete and Eirich 2007; Merton 1968), which may occur through a variety of mechanisms: one journalist’s or editor’s coverage decision may spur another’s coverage decision, as suggested by models of collective action (Granovetter 1978; Marwell and Oliver 1993) and theories of institutional isomorphism (DiMaggio and Powell 1983); knowledge of persons may spread to friends-of-friends or to adjacent areas (Bass 1969; Hedström 1994; Rogers 2003; Strang and Soule 1998); success may breed further success in various career paths (Allison, Long, and Krauze 1982; Merton 1968; Restivo and van de Rijt 2012; Salganik, Watts, and Dodds 2006); and news coverage may interact with careers, whereby more visible individuals obtain more resources, which in turn spurs subsequent media attention.

Extreme Inequality

Another common observation in scholarship of fame and celebrity regards high inequality. In any public arena, most attention is directed at a very small number of individuals. Inequality in fame in some domains of public life—such as politics, religion, and the military—relates to the preexistence of structural hierarchies of importance. Occupational tenure in these positions may be of limited duration, but at any one moment a few individuals occupy ranks of extreme importance while others naturally draw much less public attention. Even in the absence of clear, generally agreed upon differences in merit and newsworthiness, self-reinforcing spirals nonetheless generate overemphasis on a small number of characters (Boorstin 1961; Cowen 2000; Gabler 1999; Gamson 1994; Gitlin 1998; Kurzman et al. 2007; Milner 2005, 2010; Rojek 2001). Fame thus follows a distribution similar to that of prestige, which is thought to be a convex function of performance (Erickson and Nosanchuk 1984; Frank and Cook 1995; Goode 1978; Rosen 1981; Schneider 1935). It also bears resemblance to the way public attention to social problems is differentially allocated: “A very small number of social problems are extremely successful and become the dominant topics of public discourse” (Hilgartner and Bosk 1988:70–71).

Most scholars of fame see inequality as deriving from a disconnect between fame and merit. They suggest that status may become decoupled from quality through self-reinforcement (Ferris 2007; Kurzman et al. 2007; Milner 2005, 2010), as in formal models of status hierarchies (Bothner et al. 2010; Gould 2002; Lynn, Podolny, and Tao 2009). Much in the same vein, media hypes, herd dynamics, and other self-reinforcing processes (see Vasterman 2005) generate and exacerbate inequalities in public discourse, not necessarily resulting in the most accomplished being the most celebrated.

High Mobility

It is the third and final feature of fame upon which scholars’ views converge that parts with a traditional understanding of stratification systems, namely that fame should exhibit high mobility. The presumed hypes and spirals of attention driven by various reinforcement processes that generate extreme inequality should also make the fame system open.

Conceptualization of fame as an open system with ready entry and exit is grounded in the notion of ephemeral fame. The same waves that generate sudden fame can lead journalists and audiences to just as abruptly lose interest in and abandon an idol: “Fame bubbles can burst as quickly as they formed” (Cowen 2000:15). “The ephemeral nature of fame” means “it can be attached to and detached from individuals relatively easily” (Marshall 2004). As a result, “fame moves readily and easily between the domains of the public and the private for public consumption” (Marshall 2004). In a similar vein, Ferris (2007:373) concludes that “celebrity does not usually last very long.” Kurzman and colleagues (2007:347) speak of fame as “status on speed.” Fame should have become particularly fleeting in recent days: “Increasingly, the time span between the rise and evaporation of celebrity is getting shorter” (Currid-Halkett 2010:219). Some attribute ephemeral fame to the “elevation of the ordinary” (Gamson 2011:1061) or the “demotic turn” (Turner 2004, 2006) in modern celebrity culture, whereby the public becomes interested in the lives of otherwise regular persons.

The idea of ephemeral fame is echoed in today’s public perception of celebrity and its volatile nature. The presumed short attention span of today’s media and public is a common notion in art and entertainment (e.g., Andy Warhol’s idea of “15 minutes of fame” or the movie Chicago). This perception is further strengthened by tabloids’ in-and-out lists, informing us of who is now worthy of public attention, and who is no longer interesting. Newspaper columnists often lament the shallowness of today’s culture, citing reality TV stars such as Kim Kardashian and celebrities such as Paris Hilton, and question if they deserve their fame. Both scholarly work and public discourse suggest that fame is short-lived in today’s world, due to rapidly changing fashions and an unending search for the next big thing.

It is this scholarly and popular notion of ephemeral fame that has led theorists to conceptualize fame as an open system exhibiting qualitatively less stability than traditional stratification systems. In Milner’s (2005, 2010:383) theory of fame as a status system, increased mobility is a defining feature: “celebrity status is likely to be less stable than more traditional forms of status.” According to Rojek (2001:94), fame undergoes high mobility in “the manifold ranks of celebrity relating to sport, music, art, literature, humanitarianism, politics and the other institutions of modern culture. Within these ranks, upward and downward mobility is a continuous characteristic of the status hierarchy to which celebrity watchers, and the general public, are perpetually attuned.” Fame is “much more openly linked to economic and political power” (Milner 2010:382) and, like charisma (Weber 1966), has a “flexible association with wealth” (Ferris 2007:373), allowing it to escape the strong persistence and immobility that the stratification of wealth exhibits; it is “unstable” (Ferris 2007:373). Kurzman and colleagues (2007:347) describe fame’s relative mobility as a characteristic feature that makes it distinctly different from other forms of power, resources, and status: “It demands a constant supply of new recruits, rather than erecting barriers to entry.” Based on this, we derive the following hypothesis:

Low Mobility

Scholars of fame concur that fame should exhibit high mobility, but from the perspective of the sociology of stratification, the combination of self-reinforcement and high mobility is surprising. In the literature on cumulative advantage, self-reinforcement is seen as a “mechanism for inequality” (DiPrete and Eirich 2007:271). ¹ If new resources are disproportionately allocated to individuals who received prior resources, then the pecking order is preserved. ² For example, Merton’s (1968) classic analysis of Zuckerman’s (1967) interviews with Nobel laureates suggests a cumulative advantage dynamic in the allocation of fame to scientists (a Matthew Effect). This cumulative advantage dynamic solidifies the gap between the haves and have-nots, thus inhibiting mobility. Scholars have observed similar dynamics for the consecration of baseball players (Allen and Parsons 2006), song popularity (Salganik et al. 2006), author success (Sorenson 2007), and social status on the Internet (Restivo and van de Rijt 2012). The sociological literature on cumulative advantage and inequality suggests that in a reward system in which reinforcement is strong, mobility will tend to be low.

A second force that generates stability in stratification systems is the existence of formal and informal structures of power in a society. Higher level structural positions—whether in politics, science, or sports—tend to come with higher income, greater power, and greater media attention. The tenure associated with their occupancy guarantees a fixed rank for some period, and self-reinforcing processes may, at the end of tenure, convert the experience gained and power accumulated into a follow-up position of importance. Furthermore, network connections built with peers at the same stratum of power may provide indirect means of maintaining prominence. In the context of art, Lang and Lang (1988:95, 97) speak of a “satellite effect” and “elite connections” to other famous artists that enhance legitimacy (Schmutz and Faupel 2010) and ultimately help preserve an artist’s oeuvre. At the highest fame strata, endorsement of political candidates and commercial products, as well as charities and social movements that bring in famous individuals, provide means for perpetuating celebrity status (Meyer and Gamson 1995). Furthermore, theories of status suggest that status is borrowed from one’s social contacts (Bothner et al. 2010; Graffin et al. 2008; Podolny 1993, 2001). Members of status groups may thus reinforce one another’s status through isolation from those who lack status. Scholars of fame have indeed noted such strong homophilous interaction patterns (Lazarsfeld and Merton 1954; McPherson, Smith-Lovin, and Cook 2001) among the famous (Alberoni 1972; Collins 1998; Currid-Halkett 2010; Kurzman et al. 2007; Milner 2005, 2010).

The idea of high mobility in fame is not only at odds with a classic understanding of stratification systems, but it also conflicts with scholarly insights into how individuals in fame-producing positions act. In the domain of artistic reputation and renown, a key determinant of posthumous acclaim is the survival of artistic materials (Lang and Lang 1988; Taylor 1996). Consequently, curators and collectors play a central role because they “vie with one another for the contents of a newly deceased artist’s studio, if he is already famous, while things left by another who is equally talented will disappear because they have no value” (Lang and Lang 1998:105). Moreover, “ambiguities with regards to authorship tend to be resolved in favor of the more renowned” due to the common but “false assumption that any unidentified masterpiece must be the work of a renowned master” (p. 105). The unintended consequence is again a self-reinforcing dynamic whereby “small and non-measurable differences in lifetime recognition can result in significant disparities in posthumous acclaim” (p. 106). Critics and scholars, functioning as “reputational entrepreneurs” (Fine 1996:1159; see also Shrum 1991), may selectively disseminate (Lamont 1987) or retrospectively consecrate work by a few key players in a field, through halls of fame or identification of “the most exemplary achievements by cultural producers” (Allen and Lincoln 2004:873) that stood the “test of time” (Becker 1982:365, cited in Allen and Lincoln 2004). This reinforces a “magical division” (Bourdieu 1984:6) between the official canon and everything else (Braden 2009; DiMaggio 1982). Established artists are often given roles of differential status at art gatherings—such as well-established poets serving as gate-keepers who advise on, edit, curate, and publish the works of their less-known counterparts (Craig and Dubois 2010)—again reproducing preexisting status differences. Finally, orchestras also serve as fame producers, perpetuating distinctions through conservative choices of works by often-played composers (Dowd et al. 2002).

A different category of fame producer is formed by individuals working in the news industry who often act in ways that should reduce mobility in the fame system. Although the literature on mass communication does not focus on fame per se, beginning in the 1970s media sociologists recognized the unequal distribution of media attention and the active role of both news promoters (individuals who seek media attention and their assistants) and news assemblers (journalists, editors, and producers) in creating this inequality (Andrews and Caren 2010; Fishman 1980; Gans 1972, 1980, 1983; Oliver and Maney 2000; Tuchman 1973). News promoters situated in positions of corporate or political power enjoy habitual access to the media and can drum up attention through press conferences and media releases, while others must resort to disruptive behavior to attract journalists’ attention (Molotch and Lester 1974). Experienced organizations know how to write a good press release with vivid quotations, time major events to match news media deadlines, and notify the press in advance about upcoming events (Gamson and Wolfsfeld 1993; Oliver and Myers 1999; Salzman 1998). News assemblers, on the other end, for fear of failure, recycle past stars in new configurations and media formats, producing further stickiness in media dynamics (Bielby and Bielby 1994; Gitlin 1983). Moreover, news assemblers’ limited ability to obtain information from novel sources given pressing deadlines leads them to report on news beats (Gans 1980; Oliver and Myers 1999); in covering a particular beat, reporters routinely attend similar events and repeatedly encounter the same individuals.

In summary, insights from the sociology of stratification, commemoration, and media, when applied to fame, suggest a dynamic contrary to the reigning consensus in the sociology of fame. Individuals who manage to gain publicity and achieve media attention are likely to sustain their high levels of coverage, while those previously judged to be of little public interest will probably be so judged again the next time around. We can thus derive the following competing hypothesis:

Operationalization of Fame

In accordance with how the sociology of fame defines fame, we conceptualize fame as a position on a continuum between being known only by friends and family to being world-renowned. We measure a person’s fame as the number of appearances of that person’s name in newspaper records. More precisely, for every day, we count the number of distinct newspaper articles in which a name appeared. ³ In our analysis, we sum these daily counts over the course of a year, standardizing by annual volume, to obtain a quantity robust to the burstiness of news reporting and annually recurrent patterns of seasonality (we show that substantive results are similar when intervals shorter than a year are chosen). Annual measures are also the standard in studies of income mobility (e.g., DiPrete 2002; Shorrocks 1978). Clearly, there will be a close correspondence between how famous an individual is and how often she is regularly referenced in newspapers. Becoming renowned today is, to a great extent, determined by receiving media attention, as recognized by the long tradition of research on agenda-setting in the field of mass communication (McCombs 2004; McCombs and Shaw 1972) and in historical studies of fame (e.g., Braudy 1986). Conversely, individuals with a large public following will likely accumulate more references in newspapers. Thus, the extent to which people appear in the news likely corresponds closely with the extent to which they are on our minds and who we talk about in our day-to-day conversations, beyond our own circle of family, friends, and co-workers.

Although we believe that newspaper mentions are an appropriate measure of fame, we also recognize its limitations. Some individuals will be widely known among audiences that are less targeted by newspapers. In addition, scholars have long argued that media coverage correlates with sociodemographic variables such as ethnicity and race (Braham 1982; Entine 2000; Entman 1990; Hartman and Husband 1974; Shor and Yonay 2011), gender (Larson 2005; Russo 1981; van Dijk 1993; van Zoonen 1994), religion (Hussain 2000), and sexuality (Alwood 1996; Kielwasser and Wolf 1992). Such differences do not always correspond with differences in underlying fame. Furthermore, due to growing competition from other media—radio, television, and more recently the Internet and social media—newspapers have lost some of their historical dominance as the primary producers and reinforcers of knownness.

A final objection to our operationalization may be that while coverage of persons dynamically follows the notable events in which they partake, their presence in public memory remains more constant. Maria Sharapova will draw more attention when she wins a Grand Slam tournament, Katie Holmes when starring in a newly released movie, and David Sedaris right after his latest book comes out. However, the degree to which people would affirmatively answer the question “Do you know who this person is?” may change little in between tournaments, movies, and books. If so, it means fame will appear more ephemeral in the newspaper data at hand than it actually is in society’s collective mind. Hence, if we find fame to be stable in newspapers, we may expect it to be even more stable in public consciousness.

Datasets

Our data come from English-language newspapers that were scanned for person names by the Lydia text analysis system (for details, see Bautin et al. 2010; Lloyd et al. 2005). ⁴ Lydia performs named entity recognition, classification, and analysis of text corpora. A named entity can generally be thought of as a proper noun, most commonly a person, place, or organization. Lydia employs natural language processing (NLP) to reduce text streams to time-series data on the news volume associated with each news entity and their juxtapositions in sentences, articles, and newspapers with other news entities.

Here, we analyze data collected by Lydia from approximately 2,200 U.S. daily newspapers, weekly and more infrequent periodicals, and a significant number of foreign English-language newspapers. Some examples are the New York Times, the Guardian, and local papers such as the Toledo Blade and the Sacramento Bee. A select list of major U.S. newspapers is presented in Part B of the online supplement and a complete list of all English-language newspapers in the database in Part C (http://asr.sagepub.com/supplemental). The list includes a great heterogeneity of online written news sources, which we refer to as newspapers, ranging from reputable journals with nationwide circulation to college newspapers to fashion magazines to TV stations’ websites. ⁵ Daily data collection from these sources has been ongoing since November 2004, and a very large corpus of text for analysis has accumulated (henceforth, the Dailies text corpus). For 13 newspapers (henceforth, the Archival text corpus), Lydia data go back beyond 2004 because these papers have made scanned articles available. ⁶

We focus most of our analysis on samples of roughly 10⁵ person names randomly drawn from the Dailies and Archival corpora. We use seven distinct datasets in our analyses: (1) the Dailies dataset, containing timestamps of occurrences of random names in the Dailies corpus between November 2004 and January 2009; (2) the Archival dataset, consisting of names randomly drawn from the Archival corpus; (3) the New York Times dataset, which contains annual frequencies of the complete population of names mentioned in available pages from the New York Times in 1988, 1993, 1998, 2003, and 2008; (4) the Entertainment dataset, containing all occurrences of random names in entertainment sections of the Dailies corpus; (5) the Celebrities dataset, containing all occurrences of names that have at least 50 percent of their mentions in newspaper entertainment sections in the Dailies corpus; (6) the Blogs dataset containing all article references to names mentioned on the blog site LiveJournal; and (7) the Television dataset containing mentions of names on the websites of broadcasting corporations, such as CNN and FOX, in the Dailies corpus. We indicate the name of the respective dataset when used. To give readers a sense of what names are included in our samples, the three most highly mentioned names in the Dailies dataset are Brett Favre (prominent U.S. football quarterback), Alex Salmond (first minister of Scotland), and John Roberts (U.S. Chief Justice). Note that the reason a name like George W. Bush is not in this list is that we did not draw a random sample of all mentions, but rather a random sample of all names. The former U.S. president was not sampled. For each name sampled, the dataset includes every mention in any newspaper.

Data Challenges to Testing the Hypotheses

Distributional Analysis

Figure 1 is a double-logarithmic plot of the frequency distribution of annual coverage in our New York Times sample in five distinct years spanning two decades of coverage. Use of a single newspaper in this distributional analysis permits historical comparison without the confounding effect of the changing composition of our newspaper database. The horizontal axis measures the number of newspaper references to a name. The vertical axis measures the proportion of names among the names with the respective reference volume. Most names are mentioned only a handful of times and can be found in the top left of the chart. Famous names are located in the bottom right.

OPEN IN VIEWER

Figure 1.

Distribution of Fame in the New York Times across Two Decades (New York Times Dataset)

All five distributions approximate a straight line with a slope of −1.1, suggesting a power-law distribution with 1.1 as scaling exponent. Classic models of feedback produce distributions that approximate a power-law in their tail (Barabási and Albert 1999; Price 1976; Simon 1955; Yule 1925), consistent with theoretical notions of self-reinforcement and cumulative advantage in fame discussed earlier, although other processes can also produce similar heavy-tailed population-level distributions (Clauset, Shalizi, and Newman 2009; Jones and Handcock 2003; Newman 2005).

Apart from being consistent with the notion of reinforcement in fame, the frequency distribution indicates that fame indeed exhibits extreme inequality, whereby nearly all public attention is allocated to a highly select group of individuals from among those who receive any coverage. To give an indication of the severity of inequality in the sample, in each of the five years approximately 60 percent of all newspaper coverage goes to only 1 percent of the names.

Interestingly, this degree of inequality remained largely constant over the course of this 20-year period, with a stable fraction of newspaper space allocated to each fame stratum. The top left of Figure 1 does not show an increase in the fraction of minor names, as theories about the increased celebrification of ordinary people would suggest (Gamson 2011; Turner 2004, 2006). Nor does the bottom right of Figure 1 show a noticeable increase in the proportion of superstars (Frank and Cook 1995; Rosen 1981). Although the distribution of fame remained mostly unchanged, the composition and relative ranking of the individuals who made up this stable distribution may have shifted dramatically from year to year, as indeed the sociology of fame predicts.

Mobility Analysis

The dash-dotted lines in Figure 2 show Spearman’s rank correlation in fame across pairs of years in the Dailies dataset. The one-year rank correlation pertains to consecutive years, the two-year rank correlation to a two-year difference, and the three-year rank correlation to a three-year difference. The vertical axis measures the correlation for all names that, during the observation period 2005 to 2008, had on average at least the threshold degree of fame indicated along the horizontal axis. Figure 2 shows only a moderate correlation between fame in consecutive years. These correlations further weaken in subsequent years; the two- and three-year curves indicate significant change in annual fame ranks. Annual correlations are higher among more famous names, suggesting that fame is more stable at higher levels.

OPEN IN VIEWER

Figure 2.

Biannual Correlation and Proportion Coverage of Old Names above Varying Threshold Levels of Fame (Dailies Dataset)

The three solid lines in Figure 2 represent the coverage of old names: people who were also in the news one, two, and three years earlier. The figure shows the fraction of coverage of names that were mentioned at least once one, two, and three years earlier. We calculated this fraction for the coverage of all names with at least the threshold number of mentions in the focal year indicated along the horizontal axis. For example, the value .96 for three-year-old names at fame threshold 100 indicates that out of all newspaper coverage of names with at least 100 references in 2008, 96 percent went to names that were already in the news at least three years earlier. Figure 2 thus shows that the vast majority of coverage goes to names that have already been in the news for several years, and that new names rarely penetrate the higher strata of fame. These results suggest that although the fame hierarchy undergoes significant re-ranking from year to year, this is largely a reshuffling of already familiar names and not rapid replacement of an outgoing cohort by an incoming cohort.

The positive relationship between magnitude of fame and the age of names displayed in Figure 2 suggests that names that experience significant coverage over the course of a year maintain much of this coverage from one year to the next. Only names that did not receive much coverage disappear and are replaced. This inference can be tested by calculating the typical fame level in a year for levels of fame in the previous year. However, because of regression toward the mean, above-average fame will drop even if change is entirely random. To reduce this bias, we examined change in fame between subsequent years after this expected initial drop. ⁸ Figure 3 displays the median reduction in fame from year 1 to year 2 and from year 1 to year 3 for different levels of fame in year 0. A value of 1 indicates that fame in the destination year (year 2 or year 3) is equal to fame in year 1. As predicted, the sustenance of fame increases as we move along the horizontal axis. Small fame quickly shrinks from one year to the next, but big fame remains mostly steady from year to year.

OPEN IN VIEWER

Figure 3.

Annual Change at Varying Levels of Fame (Dailies Dataset)

The Role of Randomness

The strong persistence of fame at high levels in Figure 3 suggests the weak correlations in Figure 2 are mostly due to random, transitory fluctuation around a relatively stable level of fame (Bielby, Hauser, and Featherman 1997; Black and Devereux 2010; Rytina 1989). To explore this possibility, we calculated annual variation for each name. A natural measure of variation for count data is the modified coefficient of variation (MCOV), corresponding to the overdispersion parameter in the negative binomial distribution (Allison 1980). We computed annual variation of a name’s fame as the MCOV of its annual article total. Figure 4 shows the median variation for different degrees of fame. The pattern in Figure 4 indicates that annual variation decreases with average fame, reconfirming the pattern found earlier of stratum-dependent mobility. The MCOV scores permit qualitative examination of typical variation at different levels of average fame. Whereas an annual average of a single-digit number of mentions is characterized by high variation (a four-year pattern such as 7 – 5 – 0 – 0 with MCOV = 1.1), at high levels the mobility of fame typically exhibits only moderate annual change (e.g., 1050 – 1250 – 500 – 1200 with MCOV = .12). The degree of annual variation at high levels of fame is particularly low if one takes into account event-driven news coverage spikes even when underlying fame is robust; for example, a famous author publishing a new book or a famous actor starring in a new movie. Variation of big fame is also low given differences across newspapers in people’s latent fame. Namely, such heterogeneity will produce positive MCOV values even if newspaper coverage follows a Poisson process with a fixed rate (Allison 1980), suggesting that structural change in fame is even lower than indicated by the MCOV. Results thus suggest that at significant levels of fame, annual change is close to the theoretical minimum of pure stationarity.

OPEN IN VIEWER

Figure 4.

Annual Variation (Dailies Dataset)

One may object that names that enjoy significant coverage for just a brief period should nonetheless be considered moderately famous even when their fame on an annual basis is minor. To explore to what degree annual variation depends on the length of the episode over which fame is measured, the dashed lines in Figure 4 display the degree of annual variation as a function of the number of newspaper articles in a name’s peak year (thick), peak quarter (medium), and peak month (thin). Figure 4 shows that annual variation among the most famous names varies little with episode length.

Based on the evidence presented, we may conclude that Hypothesis 1 (high mobility) from the sociology of fame characterizes fame dynamics only at low levels. Hypothesis 2 is confirmed in the dynamics for higher strata, where annual variation and mobility in fame were found low. Among individuals with a large body of mentions (those we would typically characterize as famous or renowned) mobility is low. Boundaries in the stratification system become less permeable as we move up to higher levels of newspaper coverage. Famous individuals maintain most of their fame from one year to the next, leaving little room for new arrivals.

Concern I: Entertainment Fame Is Ephemeral; Fame in Sports, Business, and Politics Is Durable

One possibility is that the theory of an open fame system proposed by sociologists of fame is limited in scope to the domain of entertainment, whereas fame in other domains is mostly stratified. Indeed, examples in the fame literature often come from the world of celebrity and entertainment. In news, business, and sports, formal hierarchies of importance may command unequal media attention and mobility may be low; in entertainment and arts, however, fads and fashions prevail, rendering mobility potentially higher. Our analysis has so far pooled all names irrespective of their domain, but one could argue that the ephemeral nature of fame implicitly excludes politicians, public officials, CEOs of major companies, newscasters, and major league athletes. Instead, scholars may implicitly reserve the phrase for actors, performers, and presenters. This raises the possibility that the pattern of stratum-dependence observed earlier may represent a combination of low-mobility names in politics, business, and sports mostly at high levels of annual fame, with high-mobility names in entertainment mostly at lower levels of annual fame. Authors, actors, and artists must regain popularity after each book, movie, or CD, whereas employment in professional sports or public office guarantees consistent coverage throughout a season.

To examine such potential domain-specific mobility, we isolated entertainment-oriented news categories, in which we would expect to find higher mobility. First we performed a separate mobility analysis for names that occurred in newspapers’ entertainment sections (Entertainment dataset). To verify that these sections indeed had a greater representation from the arts and show business, we inspected the top-10 names in our sample (recall that because this is a sample, these are not the top-10 names in the population of all names from which the sample was drawn). Table 1 demonstrates that indeed all but one (Howard Hughes) are unambiguously entertainers.

OPEN IN VIEWER

Table 1.

Top-10 Names in the Entertainment Sample with Corresponding Profession

Name	Profession
Jamie Foxx	musician / actor / comedian / talk radio host
Bill Murray	actor / comedian
Natalie Portman	actress
Tommy Lee Jones	actor / film director
Naomi Watts	actress
Howard Hughes	film producer / director / entrepreneur / aviator / engineer
Phil Spector	record producer / song writer
John Malkovich	actor / producer / director / fashion designer
Adrien Brody	actor / film producer
Steve Buscemi	actor / film director

Figure 5 shows the reduction of fame from year 1 to year 2 at different levels of year 0 fame, for the Dailies dataset (solid, thick) from Figure 3 and the Entertainment dataset (dashed, thick). The lines mostly overlap. Findings for mentions in entertainment sections are similar to findings for newspaper fame in general: low levels of coverage peter out quickly but significant coverage is sustained. We also considered the Celebrities dataset, containing names for which over 50 percent of their mentions occurred in entertainment sections of newspapers (dash-dot, thin). At intermediate levels of fame—around 10² articles per year—we find names such as Alan LeQuire (sculptor), Anita Yavich (costume designer), and Ann Peebles (singer). Again, the pattern resembles the general pattern from Figure 3. Apparently, celebrities’ fame—as thus defined—exhibits the same dynamics as that of famous people in other domains: high mobility at entry levels and low turnover at higher levels.

OPEN IN VIEWER

Figure 5.

Annual Change at Varying Levels of Fame, by News Category and Medium (Dailies, Entertainment, Celebrities, Blogs, and Television Datasets)

Although we were able to capture entertainment-oriented fame only through broad, high-level categorization, our results do suggest that in entertainment—where the media spotlight is affected more by tastes, fashion, and consumer demand than by formal, durable terms of public service—fame follows a pattern of mobility similar to that in business, sports, and politics. Just like fame in more formal domains, fame in entertainment appears topped by a stable elite of celebrities who do not easily go away.

Concern II: Dynamics Are an Idiosyncratic Feature of Newspapers as a Medium

A second possibility is that low mobility characterizes dynamics of newspaper coverage more than it characterizes general fame dynamics. Newspapers may bring persons to public attention in a way that is qualitatively different from other media. Newspapers may disproportionately report on fame tied to a particular public function (e.g., major league sports, politics, or artistic excellence) and on extremely minor figures such as those in obituaries, advertisements, and involved in local crime. By doing so, newspapers may underrepresent exactly the kinds of celebrities whose fame is not institutionally supported and who lack a good reason for being in the news. Perhaps other media that are not primarily concerned with news provision exhibit different dynamics.

To explore this possibility we analyzed data from two other media: blogs and television. Blogs’ content is created in a distributed fashion, by unconstrained individuals and organizations that can make a name for themselves or draw attention to a person or topic of their choosing. By and large, blogs lack a formal institutionalized practice or code for granting public attention, although items and discussions that originated in other media such as newspapers, television, and radio are often re-consumed in the blogosphere (Leskovec et al. 2009). Television is perhaps different with regard to the categories of people it allocates attention to, sharing with newspapers the top-down tradition of news reporting on individuals in public office, but adding bottom-up elements such as reality TV shows featuring previously unseen characters.

Figure 5 shows fame in year 2 by fame in year 1 for names of varying fame in year 0 in the Blogs (dotted) and Television (dashed, thin) datasets. The figure shows patterns similar to those observed in newspapers: names with low frequency more readily disappear from the public eye than names that are blogged and broadcast regularly. Fame at significant levels is mostly preserved from year to year. The degree of stratum-dependent mobility appears larger for television than for newspapers and blogs; however, these fine differences may be due to variation in size and source heterogeneity across the corpora from which the datasets were taken. A conservative interpretation is that the limited data from these other media provide little evidence for the supposition that the stickiness of famous names is unique to newspaper coverage. Rather, patterns in Figure 5 suggest it is a more general feature of fame.

Concern III: Annual Mobility Analysis Overemphasizes Stability

The pattern of low mobility at high levels of fame thus appears to robustly describe change across domains of fame (addressing Concern I) as well as across media through which fame is broadcast (addressing Concern II). The final objection we consider is that our annual mobility analysis overemphasizes stability. While the analytic approach of measuring mobility in fame as the annual change in the quantity of references accumulated is analogous to the classic measurement of income mobility as the annual change in the number of dollars accumulated, one could argue that this approach fails to capture short intense periods of fame lasting only a quarter or even just a month. Furthermore, we have shown that high levels of fame in one year tend to be maintained the year after, but it is in principle possible that most cases of fame last only briefly while a few last very long, because longer episodes by definition are more likely to have occurred during any observation window.

An empirical assessment of this objection requires a life course analysis. Just like standard income mobility analysis ignores members’ individual career trajectories when calculating population-level changes, in the same way can our analysis so far say little about the average life course of individual fame. To examine a longer period of time we turned to the Archival dataset, which has several decades of scanned newspaper content. Accordingly, we calculated the typical trajectories of fame growth and decline for different levels of fame. This time, however, we measured the level of fame as the number of newspapers a name has ever appeared in, allowing even very brief fame to reach the highest level. We first defined new names as those that never occurred during the first 15 years of the observation period (1977 to 1991) and only kept names “born” before 1999 so that 11 full years of coverage were available in the dataset.

For a new name’s birth year and for each of the years during the decade following that birth year, we calculated the name’s fraction of its total decennial coverage. Then we stratified names by the number of newspapers they appeared in and calculated annual means within strata. Figure 6 shows how the shape of the trajectory radically changes with the degree of fame. Names mentioned in 1 to 4 newspapers received virtually all their coverage in the first year. Qualitative inspection of these patterns confirms that many involve a burst in a single week in which all attention was granted. By contrast, names that reached the readerships of eight or more newspapers show a pattern of growth suggestive of a long-lasting, perhaps even career-transcending, trajectory of fame. ⁹ The intermediate case shows an aggregate trend that suggests a mixed population of names; close examination of these data confirms that the constituent names typically follow one of the two extreme categories.

OPEN IN VIEWER

Figure 6.

Long-Term Dynamics of Fame by Peak Year Fame (Archival Dataset)

Figure 6 demonstrates that while small names quickly disappear, big names experience a career-type pattern of growth and sustenance. Contrary to what sociologists of fame argue, it is atypical to see large-scale attention to an individual during some period and hear little of this person ever after.