Testing Geographical Bias in International News

Whose Lives Count?:
TV Coverage of
Natural Disasters

by William C. Adams

Journal of Communication 36 (Spring 1986):113-122
Reprinted in Television: Critical Concepts in Media and
Cultural Studies, Toby Miller, ed. (Oxford: Routledge, 2003).


The severity of foreign natural disasters explains less than ten percent of the variation in the amount of attention they are given in nightly U.S. television newscasts.
How do the U.S. news media prioritize the rest of the world? Recent research on television news has helped describe the way countries and regions are treated (e.g., 1, 2, 7, 9, 19). Larson (9), for example, calculated detailed country-by-country coverage by ABC, CBS, and NBC each year from 1972 through 1981. But describing patterns in international coverage is far easier than explaining the factors that produced that coverage; describing is also easier than critiquing the content. The academic literature and the New World Information Order debate are filled with arguments over the causes and merits of U.S. television's apportionment of attention to various countries and various issues (e.g., 7, 19). It is popular, for example, to call for more in-depth coverage of the Third World, but at which countries' expense should such coverage come? Ultimately, the television news agenda, in terms of relative attention, is a zero-sum game. If Ghana and Peru are to receive more coverage, then France or the Soviet Union and some other objects of news attention must receive proportionately less.

Much of the analytical quagmire derives from the difficulty in defining news on a uniform standard. (Is an economic crisis in France more "important and consequential" to the world than one in Ghana? Are all coups equally significant?) What can be said about content data apart from assertions of personal preferences and apart from comparisons with other content data?

Rosengren (14; see also 15) has argued that, where possible, research should begin with "extra-media data" in order to "establish a universe of events, and this universe of events, rather than the universe of news reported during a given time period, provides the starting point for the investigation." Rosengren then advocates using extra-media data in regression analyses to determine what proportion of the variation in media coverage can be explained by various factors. Although this approach is not without problems extra-media data and ideal measures of explanatory factors are often difficult to secure Rosengren makes an important contribution to addressing some of the comparative and explanatory issues that pervade studies of media content.

Using Rosengren's conceptual framework, this article begins with extra-mass media data on events that have some intrinsic degree of uniformity natural disasters.
Earthquakes, hurricanes, and floods all strike and kill ordinary people. An appropriate measure of the severity of the disaster is the total number of people who lost their lives. The initial hypothesis to be tested is that the amount of attention U.S. television news devotes to a natural disaster reflects the magnitude of that disaster. Is it true, as Sreberny-Mohammadi (18) wrote, that "coups and catastrophes [are] newsworthy wherever they occur"? Or does the locale of the catastrophe make all the difference in the world? A commonly used formulation raises these questions in a dramatic way: "A hundred Pakistanis going off a mountain in a bus makes less of a story than three Englishmen drowning in the Thames" (13); "One dead fireman in Brooklyn is worth five English bobbies, who are worth 50 Arabs, who are worth 500 Africans" (5); "One thousand wogs, fifty frogs, and one Briton" (17). To what degree are these cynical ratios accurate?

If the magnitudes of these tragedies, as measured in number of dead, do not correlate with their prominence on U.S. television, then it will be useful to go to Rosengren's second step, estimating the explanatory power that can be attributed to other factors in the treatment of these events.

From January 1972 through June 1985, according to World Almanac 1986 (1 1, pp. 688-689), 35 major natural disasters occurred that each took the lives of at least 300 people. These 35 disasters included 17 major earthquakes, the most devastating of which killed 800,000 people in Tangshan, China, in July of 1976. In this same 13.5-year period, there were 11 major tidal waves and floods, the most severe in Morvi, India, where an estimated 10,000 people died. Seven major cyclones, typhoons, hurricanes, and monsoons were also recorded; the most damaging was the cyclone that struck Bangladesh in May of 1985 and killed 10,000 people.

U.S. television's attention to these events was measured with the use of the Vanderbilt TV News Index and Abstracts. A period of one month after each natural disaster was examined to discover exactly how much time each network devoted to news about each locale. All coverage of all 35 major natural disasters was calculated for "ABC World News Tonight," "CBS Evening News," and "NBC Nightly News" the flagship network newscasts with audiences far surpassing those of all other network news efforts. Intercoder reliability among the four coders, measured as percent agreement, was 98 percent.

In measuring the severity of the natural disasters, it was tempting to use the total number of lives lost (as recorded in the World Almanac). However, these official figures are usually announced many days or weeks after the early tentative reports to which broadcast journalists must respond. So the best measures of the perceived magnitude of the tragedies are the preliminary reports from the region. Initial estimates of lives lost in each major disaster were drawn from the wire services and other reports in the New York Times and Washington Post as well as early network stories; after each episode, the highest estimate quoted within the first three days was selected.

Had the starting point of the analysis been television news, a few of these events would have been missed entirely because they were totally ignored on nightly newscasts; others were given so little news time they might well have been overlooked in any search. By beginning instead with extra-media data on disaster deaths, a more thorough examination was possible. Substituting the preliminary death estimates for the subsequent official figures then corrects for the information available at the time. As expected, the amount of newscast time devoted to each disaster correlates more with the early estimates than with the official figures, although both associations turned out to be weak.

Coverage priorities on all three networks were highly similar. Correlations for the volume of coverage each network gave the 35 disasters all exceeded .80: ABC and CBS, .94; ABC and NBC, .84; CBS and NBC, .83. In very few instances did one network focus on a story that the others ignored. This strong comparability made it reasonable to merge all three networks into a "mean TV" variable. The composite TV variable had a correlation of.97 with ABC,.96 with CBS, and.94 with NBC and is used for the rest of the analyses.


Table 1: 1976 coverage of earthquakes in six countries
Early est. deaths Mean network TV news time TV minutes per est. 1,000 deaths Later "official" deaths
Italy     1,000   7.6 Min. 7.60    946
Guatemala     5,200 12.5 Min. 2.40 4,000
Turkey     3,000   7.2 Min. 2.38    946
Philippines     3,130   2.9 Min.   .93 8,000
China 100,000   8.5 Min.   .09 800,000 
Indonesia     9,000   0.3 Min.   .04     943

Overall, attention paid by TV shows no relationship to disaster severity. The r2 correlation for estimated natural disaster deaths and the corresponding volume of television news coverage is .03. One might expect that tragedies causing 10,000 or 25,000 deaths would attract and deserve more attention than those causing 300 or 1,000. Yet the estimated loss of life statistically explains only three percent of the variation in the amount of coverage disasters were given on nightly network newscasts.

Television's treatment of natural disasters can be illustrated with specific examples drawn from earthquake coverage in 1976 and by using the grim statistic of newscast minutes per estimated 1,000 dead.
Six major earthquakes rocked the globe in 1976 Guatemala in February, Italy in May, Indonesia in June and July, China in July, the Philippines in August, and Turkey in November. Table 1 shows the sizable differences in the magnitudes of each tragedy, ranging from about 800,000 deaths in China and 23,000 in Guatemala to less than 1,000 in Italy. The table also shows curious discrepancies in coverage.

Earthquake coverage in 1976 reflects the disparities in television coverage of natural disasters found throughout this period. Contrast, for example, coverage for the earthquakes in Turkey and in the Philippines. Originally, the disaster in the Philippines was thought to be as severe as the one in Turkey, but it received less than half as much coverage. Guatemala experienced one of the worst earthquakes in this century in the Western hemisphere. Yet, proportionate to the number of victims, it received one-third of the coverage given the Italian earthquake.

Asian countries received the shortest shrift, with the Philippines, China, and Indonesia given the least attention. The disaster in the Philippines caused eight times as many deaths as the one in Italy, but it received less than half as much coverage. And the earthquake in Indonesia was initially announced by the networks to have killed 9,000 people nine times as many as the one in Italy but Indonesia's losses received about 20 seconds of airtime (on a single night), compared with those of Italy, which received coverage for more than a week, typically including four nights of major (1.5-to 2.5-minute) stories.

The greatest outlier of all was the Tangshan quake of 1976, which caused the greatest loss of life of any natural disaster in the twentieth century. For example, the number of dead 800,000 was 160 times greater than the 5,000 people killed in the Italian earthquake of 1980. Even using the very early estimate of 100,000 dead, the Tangshan quake was worse than all other recent earthquakes (1972-1985) combined.

The enormous scope of the disaster contrasts sharply with its failure to register on the world's consciousness. China's near-total embargo on news during this period led to average U.S. network evening newscast airtime of less than 9 minutes devoted to the earthquake and subsequent flooding. The disparity between casualties and coverage is so enormous that the Chinese data skew the statistics and are excluded from the balance of the data analysis.

One other natural disaster 1979's Hurricane David is also excluded from subsequent data analysis, but for a different reason. Available resources did not make it feasible to disentangle coverage about the hurricane's approach and landfall in the United States from coverage about its rampage in the Caribbean, where most of its real victims resided.

If all Chinese and Hurricane David cases are excluded, there are still 30 major natural disasters to examine in more detail. Excluding these outliers, the r2 correlation between estimated disaster deaths and TV coverage increases only from .03 to .08.

Massive loss of life is so difficult to fathom that there may be some sort of automatic psychological logarithm at work. An earthquake with 20,000 deaths may not be comprehended as ten times worse or as meriting ten times more coverage than one with 2,000 deaths. If that is so, then the logarithm of estimated disaster deaths may correlate better with TV coverage than does the raw number believed dead. And, in fact, the log of estimated disaster deaths does produce a modest increase, raising the r2 from 08 to .18, but this still explains less than one-eighth of the variation in nightly network news attention.

Table 2 shows the broad pattern in coverage. Western Europe (Italy) is indisputably on top. Eastern Europe (Romania) is a distant second and Latin America (Brazil, Colombia, Guatemala, Honduras, Nicaragua, Peru) a distant third. The least relative news time is given to the Middle East (Turkey, Algeria, Yemen, Iran) and to Asia (Bangladesh, India, Pakistan, Indonesia, the Philippines). Were we to set up an equation of relative coverage using our data, the deaths of 1 Italian would equal those of 3 Romanians, 9 Latin Americans, 11 Middle Easterners, and 12 Asians.


Table 2: Natural disaster coverage by region
TV minutes per est. 1,000 deaths Coefficient of
relative variation
(Std. dev. / mean)
Number of major natural disasters TV minutes per official 1,000 deaths
Western Europe 9.20   .28   2 6.72
Eastern Europe 3.60   1 2.53
Latin America 1.02 1.24   7   .92
Middle East   .87   .74   7   .83
Asia   .76   .96 13   .63

Regional differences are not entirely definitive, however; the standard deviations suggest some lack of uniformity within regions. Especially in the case of Latin America, the coefficient of relative variation (CRV = standard deviation/mean) indicates that the mean masks considerable variation in television's treatment of disasters in that area.

U.S. evening newscasts allocate more newscast attention to human death in some parts of the world than in others. Factors other than the actual magnitude or even rough order of magnitude of the tragedy must influence broadcast journalists' decisions on how to apportion coverage. What other factors help predict these patterns?

Organizational factors and attitudinal/ideological factors have been suggested as possible influences on the news agenda.
Epstein (6) has stressed the importance of factors such as logistics and the organizational imperative of dramatic stories that will appeal to a broad U.S. audience. Lichter and Rothman (10) have emphasized the role of journalists' personal world views. Paletz and Entman (12) and others have weighted heavily the power of the government and capitalist elites to sway media coverage. Walter Cronkite and other journalists of the "that's the way it is" school have maintained that the news simply reflects events. Graber (8), Roscho (16), and others have suggested that television news echoes society's general cultural norms and dominant values.

For the multiple regression equations, these factors were operationalized in a variety of ways. The severity of "actual events" was operationalized with three different variables: the number of estimated disaster deaths, the logarithm of estimated deaths, and the later "official" number of deaths. Measures of the objective significance of the "event's country context" were the population of the country in which the natural disaster occurred, the gross national product of the country, and the gross national product per capita (24). Sympathies of the U.S. government and power elites were calibrated with four indicators: the extent to which countries vote with the United States in the United Nations' General Assembly (3); the amount of foreign aid and credit granted by the United States (23); the amount of U.S. exports (23); and exports per capita (23).

Social and cultural affinities of leading journalists and U.S. citizens at large were gauged in three different ways: the number of U.S. tourists to each country (21); the number of Americans who identify their ancestors (or parents or themselves) as coming from each country (22); and the proportion of leading journalists with ethnicity in common with each country (10).

The networks' penchant for drama was partially measured by creating dummy variables for each type of disaster (e.g., earthquakes vs. floods, vs. cyclones) and by coding each country's share of international news time during the two years preceding the disaster. The latter variable was in effort to capture any effects from "continuing saga" coverage (whereby a country already in the news would have an established priority). The logistical, and perhaps psychological, impact of geographical proximity was estimated by the variable of miles from New York City to the capital of each country. One final important organizational factor the competition of other events for the finite "newshole" was not amenable to a convenient measurement and will be discussed below.

Following Rosengren, the next task is to see how much of the total variation these independent variables explain. Accordingly, all 16 variables were entered in a multiple regression equation to try to predict the amount of newscast time devoted to each major natural disaster. Three of these 16 variables stood out, together explaining 61 percent of all variation in coverage. The other variables could all together explain only in additional nine percent of the variation. (Thirty percent of the variation in coverage remained unexplained.)

These three most potent factors were (a) number of U.S. tourists (i.e., cultural proximity and social interest), (b) logarithm of estimated disaster deaths (i.e., severity of the news event, modified by a logarithmic scale), and (c) distance from New York City (i.e., geographical proximity). As long as these three variables were in an equation, no other variables came close to absorbing more than three percent of the variation. These three independent variables were entered as the sole predictors in a multiple regression equation used to calculate the unique variation explained by each one.

Their independent contributions to the total explained variation can be partitioned using the coefficient of partial determination, which "measures the proportion of variation in the dependent variable that is explained by each independent variable while controlling for, or holding constant, the other independent variable(s)" (4, p. 280). The coefficient of partial determination for U.S. tourists is .331; for the logarithm of estimated deaths, .201; and for distance from New York City, .047. Since these variables are highly orthogonal to one another, most of the explained variation is unique rather than overlapping explained variation (.032).

A third of the variation in network news coverage of disasters can be accounted for by a country's popularity with U.S. tourists.
Rosenblum (13) sees tourism as emblematic of ethnic and cultural affinity. In discussing why an earthquake in Italy gets more coverage than one in Guatemala, he notes:
[It occurs] partially because Italy is easier to cover than Guatemala, and more reporters are immediately available. But it is mainly because Italians are seen as individuals, with physical and cultural characteristics familiar to Americans. Many editors and readers have been to Italy, and they recognize place names in the stories. Guatemalans are seen, on the other hand, only as faceless residents of the underdeveloped world.
Note that tourism had a correlation (r) of .84 with U.S. exports; .91 with the ethnic ancestry of elite journalists; and .66 with the ancestry of the U.S. public. This further suggests that the tourism variable is a reasonable and convenient surrogate for the sociocultural affinity of the United States for other countries.

In the equation, the tourism figure is calculated in a way that at least technically controls for "distance from New York." Geographical proximity explained one-twentieth of the variation in coverage, independent of the magnitude of the disaster and the social proximity (tourism). And, not to overlook "reality," logarithms of the preliminary numbers of deaths uniquely explained another 20 percent of the variation.

What about the remaining variation that is left statistically unexplained? One strong candidate to account for the balance is one for which no suitable measure was found to incorporate into the multiple regression equations the newshole of competing stories. It is axiomatic that a story's rank on the news agenda depends on what else is happening in the world at the same time. Could competing stories account for much of the unexplained variation in disaster coverage?

One way to investigate this matter is to select those instances that are most poorly predicted by the three-variable multiple regression equation. Inspection of the residuals revealed three cases at or beyond two standard deviations. Those cases were the Nicaraguan earthquake in 1972 and the Italian earthquake in 1980 both of which received substantially more coverage than the regression equation predicted and the Italian earthquake of 1976, which received less coverage than predicted by tourism, distance, and lives lost.

The Nicaraguan earthquake's unusually high degree of airtime in December 1972 may be attributed to three factors: (a) Roberto Clemente, a popular baseball player for the Pittsburgh Pirates, was killed in a plane crash as he was taking part in a relief effort to aid his native land; (b) the earthquake occurred December 23, and much of the story emerged during the period between Christmas and New Year's Day, traditionally considered a slow news week; and (c) other top stories were almost all ongoing stories about the continuing anti-war efforts in the United States.

The Italian earthquake occurred on November 23, 1980, after the presidential race was over. It clustered nicely with a "disaster array" that included news about a big Los Vegas hotel fire, a Reno hotel fire, and California brush fires, with the brief counterpoint of a small tremor in western Nevada. These stories would appear to bolster rather than minimize events in Italy. Without dramatic developments in the Iranian hostage crisis, Italy was well situated to compete for air time.

In contrast, another Italian earthquake was given much less coverage than tourism, distance, and deaths predicted was its due after May 6, 1976. At that time, the competition was substantial: the unusual challenge of Ronald Reagan to the renomination of President Gerald Ford as well as various last-ditch efforts to stop Jimmy Carter from locking up the Democratic nomination, amid assorted primaries and caucuses held during this period.

Retrospective searches do run the risk of mistakenly confirming the researcher's presumptions. Nonetheless, these three cases lend plausible support for the notion that newshole competition can account for much of the otherwise unexplained variation in the way these 30 natural disasters were treated.

One widely voiced complaint has been that far too much of the coverage of the Third World consists of natural disaster stories (although Stevenson and Cole [19, p. 59] "could find no evidence that more attention was paid to this category of news in the Third World than in any other part of the globe"). The present research has concluded, however, that earthquakes, typhoons, and floods in the Third World, given their severity, have received proportionately little attention.

Stevenson and Gaddy (20), who argued that coverage of the Third World was somewhat heavy with negative and conflictual stories because there was more actual conflict in the Third World, speculated that a count of "the coups and earthquakes around the world" would show [that] "violence and conflict get reported pretty much the same way wherever they occur." This research suggests that this is not the case, however. Where earthquakes occur makes a great deal of difference. The severity of natural disasters alone explains little (less than one-tenth) of the variation in their coverage on nightly U.S. newscasts. A little more (about one-fifth) of the variation in news attention can be explained using logarithms of estimated disaster deaths. Overall, the globe is prioritized so that the death of one Western European equaled three Eastern Europeans equaled 9 Latin Americans equaled 11 Middle Easterners equaled 12 Asians.


REFERENCES
1. Adams, William C. (Ed.) Television Coverage of the Middle East. Norwood, N.J.: Ablex, 1981.
2. Adams, William C. (Ed.) Television Coverage of International Affairs. Norwood, N.J.: Ablex, 1982.
3. Anti-Defamation League, B'nai B'rith. "Keeping Score at the U.N." New York, 1980-1982.
4. Berenson, Mark, David Levine, and Matthew Goldstein. Intermediate Statistical Methods and Applications. Englewood Cliffs, N.J.; Prentice-Hall, 1983.
5. Boyer, Peter. "Famine in Ethiopia." Washington Journalism Review 7, January 1985, pp. 18-21.
6. Epstein, Edward Jay. News from Nowhere. New York: Vintage, 1974.
7. Gerbner, George and Marsha Siefert (Eds.) World Communication. New York: Longman, 1984.
8. Graber, Doris. Mass Media and American Politics. Washington, D.C.: Congressional Quarterly Press, 1980.
9. Larson, James. Television's Window on the World. Norwood, N.J.: Ablex, 1984.
10. Lichter, S. Robert, Stanley Rothman, and Linda S. Lichter. The Media Elite. Washington, D.C.: Adler & Adler, 1986.
11. Newspaper Enterprise Association. The World Almanac: 1986. New York: NEA, 1985.
12. Paletz, David and Robert Entman. Media Power Politics. New York: Free Fress, 1981.
13. Rosenblum, Mort. Coups & Earthquakes. New York: Harper & Row, 1981.
14. Rosengren, Karl Erik. "Four Types of Tables." Journal of Communication 27(l), Winter 1977, pp. 67-75.
15. Rosengren, Karl Erik. "International News: Methods, Data and Theory." Journal of Peace Research 11, 1974, pp. 145-156.
16. Roscho, Bernard. Newsmaking. Chicago: University of Chicago Pressi 1975.
17. Schlesinger, Philip. Putting "Reality" Together. London: Constable, 1978.
18. Sreberny-Mohammadi, Annabelle. "Results of International Cooperation."Journal of Communication 34(l), Winter 1984, pp. 121-134.
19. Stevenson, Robert and Richard Cole. "Patterns of Foreign News." In Robert Stevenson and Richard Cole (Eds.) Foreign News and the New World Information Order. Ames: Iowa State University Press, 1984, pp. 37-62.
20. Stevenson, Robert and Gary Gaddy. " 'Bad News' and the Third World." In Robert Stevenson and Richard Cole (Eds.) Foreign News and the New World Information Order. Ames: Iowa State University Press, 1984, pp. 88-97.
21. United Nations. 1982 Statistical Yearbook. New York: United Nations, 1985.
22. U.S. Bureau of the Census. Ancestry of the Population, 1980. Washington, D.C.: U.S. Bureau of the Census, 1983.
23. U.S. Bureau of the Census. Statistical Abstracts of the United States: 1986. Washington, D.C.: Bureau of the Census, 1985.
24. World Bank. 1983 World Bank Atlas. Washington, D.C.: World Bank, 1983.

William C. Adams is Professor of Public Administration at The George Washington University. An earlier version of this article was presented at the American Association for Public Opinion Research conference in St. Petersburg, Florida, May 17, 1986.  The author acknowledges and thanks Gregory Giaquinto, Brian Smith, and Joseph Fiore for their assistance.
Return to Research Methods Home Page