Media coverage and charitable giving after the 2004 tsunami.

Brown, Philip H. ; Minty, Jessica H.

1. Introduction

Relief workers have long believed in the causal relationship between media coverage of humanitarian crises and charitable giving to relief agencies. For example, private donations to relief agencies during the early stages of the 1994 Rwandan genocide were sufficient to support approximately one million displaced Rwandans, but after the tribulations of O. J. Simpson and Tanya Harding eclipsed Rwanda in the media, funding for relief activities began to decline. The perception among many aid workers in the camps for Rwandan refugees (including one of the authors) was that the media had turned its back on the crisis, with potentially dire consequences. (1)

While anecdotal evidence abounds, few studies have systematically assessed the relationship between media coverage and the behavior of private donors, and those that do suffer from important analytical shortcomings. For example, an analysis undertaken by the Institute for Philanthropy finds that 14 of 15 surveyed British philanthropists believe that the media has the power to encourage private giving, with 11 of the 15 being inspired to make charitable donations themselves (Breeze 2005). Similarly, Olsen, Carstenson, and Hoyen (2003) find a high correlation between the total number of relevant articles in western newspapers and the total amount of humanitarian assistance allocated to victims of the flooding in Mozambique in 2000 and the cyclone in eastern India in 1999. Unfortunately, neither sample is large enough to make generalizations, nor is there any attempt to quantify the magnitude of the causal relationship. In part, the difficulty in quantifying the relationship stems from disentangling the effect of new developments in stories covered in the media from the pure effect of media coverage. Returning to the Rwanda case, the former is tantamount to asking "What is the combined effect of 10,000 additional deaths and a five-minute news story on donations to U.S. charities?" while the latter is equivalent to asking "Given that 10,000 people were killed, what is the effect of increasing the news story from five minutes to six?"

In this paper we use the December 26, 2004 tsunami as a case study to quantify the causal effect of media coverage on donations to charity. Specifically, we consider the effects of reporting on three nightly network news broadcasts and articles in two prominent newspapers on private donations to seven U.S. relief agencies. To eliminate the time lag between when news reports are seen or read and when donations are received by relief agencies, we focus exclusively on donations made via the Internet. We also control for donor fatigue, tax incentives, and weekends to better isolate the effect of media coverage. Finally, to alleviate simultaneity concerns and omitted variable bias--including new developments in the tsunami story itself--we use announcements of casualties among U.S. military personnel in Iraq to instrument for media coverage of the tsunami. We find that one additional minute of nightly news coverage on network television or one additional 700-word story in major newspapers increases that day's total Internet donations by 16.5-20.8%, controlling for the time that has elapsed since the tsunami struck, tax incentives, and weekends. Using instrumental variables to account for endogeneity concerns reveals that an additional minute of television news coverage raises donations by about 2.5%, an effect that remains large in magnitude and statistically significant. We also find considerable evidence of donor fatigue and evidence that the Tsunami Disaster Aid Tax Relief Act successfully encouraged private giving. These findings are consistent with Andreoni's (1989, 1990) "warm glow" and Sugden's (1984) "commitment" motivations for charitable giving.

The remainder of this paper is organized as follows: Section 2 describes theoretical motivations for charitable giving and provides details about the tsunami disaster and the media coverage that ensued; section 3 describes the data and variables used; section 4 describes the empirical specification and identification; section 5 provides the results of this study; and section 6 concludes.

2. The Tsunami, the Media, and Charitable Giving

Early research into the economics of private giving to charitable organizations classifies the primary motivation for giving as either fostering the provision of public goods or increasing private consumption. The "public goods" model is exemplified by donors who give based on the anticipated private return to some form of public good (Warr 1982; Roberts 1984). By contrast, the "private consumption" model arises when donors derive utility from the act of giving, either because the well-being of others enters their own utility functions or because the public approval of giving benefits the donor (Arrow 1972; Steinberg 1987). For example, conspicuous donations may signal wealth, thereby enabling donors to interact with people in higher socioeconomic strata (Glazer and Konrad 1996). However, donors may also receive a "warm glow" from making charitable contributions (Andreoni 1989, 1990) even when their donations displace those of other donors, when there are no direct social benefits to donors, and when the beneficiaries of charitable giving are far away.

Rose-Ackerman (1982, 1996) posits that people only derive utility from charitable programs if they personally donate. Complementary research in this area suggests that the utility gains from donating depend critically on the behavior of others. For example, Duncan (2004) proposes that some donors may engage in "impact philanthropy" in which individuals make charitable gifts only if their donations represent large proportions of the total received by a given charity. At the other extreme, Sugden (1984) suggests that individuals are averse to free riding and hence donate if others in their peer group have also given. In this "commitment" model, individuals believe that they should donate at least as much as others in their reference group.

Regardless of whether motivated by public goods or private consumption, the majority of American households donate to charities. In 2000, for example, 69% of U.S. households made charitable donations, each donating $1942 on average (Steinberg and Wilhelm 2003). Moreover, both the beneficiaries of charitable giving and the level of donations are influenced by world events, as evidenced by the $2.4 billion in donations made to the victims of the September 11, 2001, terrorist attacks and by the $3.3 billion raised by U.S. charities for disaster relief after Hurricane Katrina. These striking examples of generosity coincided with highly concentrated media coverage in the weeks following the disasters.

The December 26, 2004 Sumatra-Andaman earthquake that triggered a devastating tsunami provides a similar example. The tsunami spread across the Indian Ocean, inundating coastal communities and claiming victims in Indonesia, Sri Lanka, India, Thailand, Somalia, Myanmar, the Maldives, Malaysia, Tanzania, Seychelles, Bangladesh, South Africa, Yemen, and Kenya. According to United Nations statistics, 229,866 people were either killed or listed as missing (UN Office of the Special Envoy for Tsunami Recovery 2006), rendering the tsunami one of the deadliest natural disasters in modern history.

The ensuing media response was unprecedented. For example, CNN deployed over 80 anchors, correspondents, and producers to provide 24-hour coverage of relief efforts. Similarly, the tsunami dominated the front page of the New York Times, garnering over half of the articles on the front page in the week following the disaster, and Time, Newsweek, The Economist, and numerous other news magazines featured the tsunami and recovery efforts in multiple cover stories. Indeed, the tsunami dominated worldwide media attention well into January 2005, much longer than any natural disaster in modern history (Wynter 2005). Possible explanations for the high level of media attention include the facts that the tsunami had little competition in the U.S. news during the last week of 2004, that media outlets allocate disproportionately more attention to unanticipated crises than to ongoing troubles, particularly when such crises are easily explained from a scientific perspective (Wynter 2005), and that a number of westerners were killed in the deluge, potentially raising the demand for news (CARMA International 2005).

Private donations to relief agencies were equally unprecedented. Catholic Relief Services reported that it raised more than $1 million in under three days, a record for the group (Slavin 2004). While Save the Children USA typically receives several hundred thousand dollars in the month following a disaster, the agency received $6 million in donations in the first four days following the tsunami (Strom 2005). Lutheran World Relief raised more money in one week than it typically does in one year (Cooperman and Salmon 2005). An Associated Press poll revealed that 30% of American households had donated to the relief efforts within two weeks of the disaster (El Nasser 2005). All told, U.S. charities received approximately $1.6 billion in private donations for the purposes of tsunami relief (Wallace and Wilhelm 2005).

At least five phenomena may have encouraged this high level of giving. First, the crisis occurred at a time of year when many Americans celebrate holidays that emphasize compassion and giving, perhaps fostering the "warm glow" associated with charitable giving (Falk 2004). Second, Southeast Asia's beaches have become increasingly familiar to Americans in recent decades because of increasing American tourism to the region (Hall and Page 2000) and because of exposure from Hollywood. (2) Third, tax incentives have been shown to motivate charitable giving (e.g., Clotfelter 1985; Kingma 1989; Slemrod 1989; Greene and McClelland 2001; Auten, Sieg, and Clotfelter 2002), including giving to international relief and development organizations (Ribar and Wilhelm 1995). Tax considerations may have been especially important in charitable contributions after the tsunami because the Tsunami Disaster Aid Tax Relief Act extended the deadline for making tax-deductible charitable contributions until January 31, 2005. Fourth, technology facilitated giving, with approximately half of all donations made online (El Nasser 2005). For example, Internet sites such as Google and Amazon added links to charities on their own websites, the latter raising $2.5 million for the Red Cross in 24 hours (Slavin 2004). Finally, extraordinary treatment of the disaster by the media may have prompted extraordinary donations by Americans.

So, news coverage of humanitarian crises may effectively serve as a form of persuasive advertising for relief agencies. Stories that focus on human suffering--such as the coverage of the large number of parents who came forward to claim "Baby 81" in Sri Lanka or the publication of photographs by Canadian vacationers who drowned on a beach in Thailand--may be particularly compelling to donors. Furthermore, media outlets help to ensure that individuals know where their donations will be put to use, either by depicting relief workers (often in uniforms bearing the logos of well-known agencies) in photographs and video footage or by publishing addresses and websites of agencies that participate in the relief efforts.

Media coverage reached a crescendo on December 30, and donations to relief agencies peaked one day later (Figure 1). A second peak in media coverage occurred on January 4 following two minor aftershocks and Secretary of State Colin Powell's visit to the affected region (Figure 2); (3) donations remained high through this period, and both media coverage and donations fell slowly thereafter. A second spike in donations occurred on January 31, when 2004 tax deductions for donations expired. Subsequent peaks in media coverage included February 1 (when the custody battle over "Baby 81" began), February 19 (when former presidents George H. W. Bush and Bill Clinton toured affected countries), February 24 (when photographs of the tsunami taken by Canadian vacationers who subsequently drowned were released), (4) March 8 (when Bush and Clinton briefed President George W. Bush on the disaster), and March 26 (the three-month anniversary of the tsunami, when the media spotlighted recovery and reconstruction). Media coverage peaked a final time on March 28, when a magnitude 8.7 earthquake occurred close to the epicenter of the December 2004 Sumatra-Andaman earthquake.

[FIGURE 1 OMITTED]

3. Data and Variables

In analyzing the relationship between media coverage of the 2004 tsunami and donations to relief agencies, this study employs daily donations made via each agency's website. Focusing on donations made via the Internet minimizes the lag between when a donor observes a news item and when his or her donation is recorded by the charity. Equally importantly, computerized records eliminate any potential for recall error. Finally, online donations accounted for approximately half of all donations for tsunami relief (El Nasser 2005), which suggests that they are a very good indicator of donations to relief agencies more generally.

Three dozen U.S. charities that helped tsunami victims were asked to share their daily receipts from Internet sources for the purpose of this research. Eight agreed to participate in this study, including Catholic Relief Services, CARE USA, Mercy Corps, SurfAid International, and four others who asked to remain anonymous; although, one is excluded from the analysis because it did not accept Internet donations until part way through the study period. Five of these relief agencies are among the nation's largest nonprofit groups by private funding, and three have religious affiliations. All eight organizations receive donations from across the United States.

Summary statistics are presented in Table 1. Between January 1 and April 5 (100 days after the tsunami struck), the mean total donation to the seven relief agencies that accepted Internet donations throughout the study period is $285,202. Total donations for tsunami relief reach as high as $3,101,581 on January 4 (after peaking at $10,221,825 on December 31) and fall as low as $5841 on April 3. The value of donations varies considerably by agency, with the smallest averaging $142 per day in donations while the largest averages $113,491. Four agencies receive over $50,000 on average each day, but only two of the agencies receive donations specifically for tsunami relief every day during this period. Agencies 4 and 7 record donations made on weekends on the following Monday, reducing the number of observations for these agencies to 65. Data for all 95 days of the study period are therefore available for five of the seven relief agencies included in the study.

[FIGURE 2 OMITTED]

Media coverage is first measured by the number of minutes of tsunami reporting on each of the three network evening news broadcasts. According to the Pew Research Center for the People and the Press (2005), 34% of Americans regularly watch the nightly news on one of the three major broadcast networks, making the nightly broadcasts among the most common sources for news. The mean number of minutes of tsunami coverage per broadcast of ABC Worm News Tonight, the CBS Evening News, and NBC Nightly News during the study period are 1.23, 1.10, and 1.63 minutes, respectively, (5) representing a significant proportion of the 18.5 minute broadcasts (which typically also include between eight and nine minutes of commercials). The standard deviations of these broadcasts are 3.15, 2.70, and 3.46 minutes, respectively, indicating that coverage varies considerably over the study period. Tsunami coverage within each broadcast is similar for all three programs, with simple correlations of between 0.66 (ABC and CBS) and 0.83 (CBS and NBC). Both ABC and CBS occasionally preempted nightly news broadcasts for sports events, but NBC aired its program every day.

Media coverage is also measured by tsunami-related articles published in major newspapers, of which 42% of Americans report being regular readers (Pew Research Center for the People and the Press 2005). Specifically, coverage is measured by word counts of articles appearing in the New York Times and the Wall Street Journal, two of the largest U.S. newspapers by both circulation and subscription (Newspaper Association of America 2006). During the study period, the New York Times averaged 1628 words in articles pertaining to the tsunami, (6) approximately 2.3 articles of average length. The Wall Street Journal averaged 1327 words (approximately 1.9 articles) per day; although, this newspaper is not published on weekends or holidays. Word counts of tsunami coverage in the New York Times are highly correlated with those in the Wall Street Journal, with a simple correlation of 0.82, suggesting that the papers were similarly focused on the disaster during this period. Moreover, coverage in newspapers closely mirrors television news coverage, with a simple correlation of 0.78.

4. Empirical Specification and Identification

This paper assesses the causal effect of daily media coverage of the 2004 tsunami on total donations to seven U.S. relief agencies. Because Internet donations are made and recorded on the same date, this measure precisely captures the time of giving. Thus, Internet donations serve as a proxy for total giving to relief agencies on each date.

In addition to media coverage, timing may affect the level of donations: Giving typically peaks immediately following a major disaster and recedes gradually (Wynter 2005). This phenomenon is known as "donor fatigue," a state in which donors have exhausted their resources or in which they grow complacent about appeals from charities, leading to a diminished public response. The gradual decline in donations is seen clearly in Figure 2. Similarly, tax incentives may also affect daily donations independent of media coverage (Clotfelter 1985; Kingma 1989; Slemrod 1989; Auten, Sieg, and Clotfelter 2002; Greene and McClelland 2001) as evidenced by the two spikes in charitable giving on December 31, 2004, and January 31, 2005 (Figures 1 and 2), the two deadlines to make eligible deductions for 2004 tax purposes. Finally, weekends may influence giving because people may be exposed to different levels of media on weekends. The model may thus be written

[DONATIONS.sub.t] = [[beta].sub.0] + [[beta].sub.1] + [MEDIA.sub.t] + [[beta].sub.2] [DATE.sub.t] + [[beta].sub.3] [TAX.sub.t] + [[beta].sub.4] WEEKEND + [[epsilon].sub.t], (1)

where DONATIONS reflects the total donations received via the Internet on day t by the seven agencies included in the study; MEDIA is the media coverage of the tsunami and recovery efforts on day t, measured either as nightly news reporting on the three largest networks or as word counts in the New York Times and the Wall Street Journal; DATE reflects the number of days since the tsunami, in log form, which accounts for the possibility of donor fatigue; TAX is a dummy reflecting whether day t is a deadline for 2004 charitable deductions for tsunami relief; and WEEKEND is a dummy indicating Saturdays and Sundays. The error term is assumed to follow a normal distribution such that [epsilon] ~ N(0, [[sigma].sup.2]).

Unfortunately, there is some variation in accounting practices across the seven relief agencies. As noted in section 3, two of the surveyed agencies did not report donations made on weekends until the following Monday. The WEEKEND dummy in Equation 1 controls for the resulting downward bias in weekend donations, but for robustness, Equation 1 is also estimated for total donations for the five relief agencies that recorded donations on both weekends and weekdays.

Because the Tsunami Disaster Aid Tax Relief Act was not signed into law until January 7, 2005 individuals rationally anticipated that December 31, 2004 was the deadline for making deductible charitable donations for tax year 2004. As a result, charitable donations were likely to swell in the waning days of 2004 even in the absence of a major natural disaster. Thus, the extraordinarily high level of donations to relief agencies recorded in late December (Figure 1) may reflect tax incentives rather than media coverage of the tsunami, and any estimates of the relationship between media coverage and donations would thus be biased upward. To eliminate this concern, the analysis begins on January 1, six days after the disaster; if media coverage truly influences donations to relief agencies, the effect should persist. (7) Hence, the TAX variable in Equation 1 becomes a dummy for January 31, 2005.

Measurement error, simultaneity, and omitted variables each potentially bias the estimated effect of media coverage on donations. However, measurement error in media coverage is unlikely because electronic databases maintain precise figures on the timing of news broadcasts and the word counts of newspaper articles. Moreover, any measurement error that does exist is unlikely to be systematic, leading simply to attenuation bias. Simultaneity bias may be more problematic, however. For example, approximately 2.1% of the tsunami-related articles published in the Wall Street Journal and 1.4% of those published in the New York Times focused on the magnitude and form of the response by private donors, suggesting reverse causality in a small number of cases.

Omitted variable bias is also potentially problematic. First, other media coverage--for example, articles in other newspapers and news broadcasts on other networks, other television and print media, Internet sites, and radio news--is likely to influence giving. Internet news sites may be particularly influential in Internet giving because approximately one-third of American adults read Internet news sites (Horrigan 2006) and because the online content of the New York Times, the Wall Street Journal, CNN, and others provided links directing visitors to the websites of charities working in tsunami relief. Second, several of the relief agencies included in this study solicited donations for tsunami victims via advertising and direct mail campaigns. Third, the efforts of some relief agencies were noted in the media, often including specific instructions for donating to those agencies (see, e.g., Strom 2004). Finally, and more fundamentally, developments in the tsunami story also affect media coverage, making it difficult to discern the effect of news coverage rather than the news event itself on donations to relief agencies: that is, it is difficult to identify the marginal effect of news coverage when the news itself is changing. (8) Each of these may result in increased Internet donations, leading to upward-biased estimates of the effect of television and newspaper coverage.

Fortunately, these endogeneity problems may be addressed using instrumental variables. Specifically, we use daily announcements of the number of casualties among U.S. military personnel in Iraq to instrument for media coverage of the tsunami (9) because dramatic news has been shown to crowd out reporting on disasters (Eldridge 2005). Casualties among U.S. military personnel are likely to influence media coverage of the tsunami as the two stories compete for time on television and space in newspapers. Moreover, this instrument is plausibly exogenous to donations to the seven relief agencies because none of them worked in Iraq during the study period.

Finally, sample selection bias may arise if the relief agencies included in the study are not representative of agencies involved in tsunami relief more generally. If compiling daily records of Internet donations is costly, for example, then the agencies that provided data are likely either to be large enough to dedicate staff time to data collection or to be efficient enough that data collection is not costly. Given that the agencies differ by size, focus, religious affiliation, and other characteristics and that virtually all relief agencies track Internet giving electronically, we find either possibility unlikely. In addition, virtually all of the agencies that worked in the affected areas accepted online donations, and all seven organizations fall under the purview of the United Nations Office for the Coordination of Humanitarian Affairs. Still, we cannot rule out the possibility that the agencies included in the study are not fully representative of all relief agencies.

5. Results

Table 2 describes the effect of nightly news coverage of the 2004 tsunami by the major broadcast networks on total private donations to relief agencies between January 1 and April 5 (100 days after the disaster) using ordinary least squares estimation. To eliminate time lags between television news coverage of the disaster and recording of the gift by the relief agency, donations made via the Internet proxy for total donations. Furthermore, because two of the relief agencies did not report donations made on weekends until the following Monday, Equation 1 is estimated both for total donations to all seven agencies (Table 2, columns 1-4) and for donations to the five agencies with donations recorded on every day of the study period (columns 5-8). Television news coverage is measured as the number of minutes allocated to coverage of the disaster on ABC World News Tonight (columns 2 and 6), the CBS Evening News (columns 3 and 7), and NBC Nightly News (columns 4 and 8), respectively, in minutes. Columns 1 and 5 aggregate the three broadcasts into a total number of minutes. In these and all subsequent estimates, heteroskedasticity-robust standard errors are reported.

One additional minute of total news coverage increases total Internet donations to the seven agencies by $47,092, or 16.5% of average total daily donations, ceteris paribus. An additional minute of relevant coverage on ABC World News Tonight increases donations by $82,175 (28.8%). An additional minute of coverage on the CBS Evening News or NBC Nightly News increases donations by $94,920 (33.3%) and $92,985 (32.6%), respectively. All of these estimated coefficients are larger than those for total nightly news coverage; although, this reflects the very high correlation of time allocated to tsunami coverage among the three networks. Results are similar when restricting the analysis to the five agencies with donation data for weekends as well as weekdays; in this case, an additional minute of total television news coverage increases Internet donations to the five relief agencies by $42,866, or 20.8%, ceteris paribus. All of these effects are significant at the 1% level.

Both media coverage and the response of individual donors to relief agencies decay as days, weeks, and months pass, as the images become less shocking, and as individual givers experience donor fatigue. To account for these nonlinear time trends, the number of days since the tsunami struck is included in log form as an additional regressor. Total donations to the seven agencies on the tenth day after the tsunami are predicted to fall by $21,012 from those received on the ninth day, (10) while donations on the 50th day are predicted to be $4029 lower than those received on the 49th day. In seven of the eight specifications, these point estimates are negative and significant at the 1% level; in the eighth, the point estimates are negative and significant at the 5% level. The extended tax deadline also produced a dramatic--if very short-spike in donations (Figure 2). Results show that total donations to the seven agencies rose by $1,110,515 during the last day of January 2005, an effect that is significant at the 1% level. Weekend donations were some 40% lower than weekday donations, reflecting the fact that two agencies did not record weekend donations until the following Monday. This effect is moderated considerably when restricting the analysis to the five agencies that recorded donations every day.

Table 3 is the newspaper analog of Table 2 in estimating the effect of media coverage on total private donations to relief agencies after the 2004 tsunami. Estimates for total donations to all seven agencies (columns 1-3) and for donations to the five agencies with weekend donation records (columns 4-6) are made separately. Media coverage is measured by the length of stories in words in the New York Times (columns 2 and 5) and the Wall Street Journal (columns 3 and 6), respectively. The total number of words in tsunami-related stories from the two sources is added together in columns 1 and 4. Because the Wall Street Journal is not published on Saturdays and Sundays, the dummy for weekends is included only in assessing the effect of coverage in the New York Times.

One additional word of coverage in the New York Times raises total Internet donations to the seven relief agencies by $109.97, or 0.039%, ceteris paribus. An additional word of coverage in the Wall Street Journal raises donations by $113.84, or 0.040%. As with television news coverage, the estimated effect of total word counts is smaller than that of either the New York Times or the Wall Street Journal alone, reflecting the fact that coverage among print media is highly correlated. One additional word increases total Internet donations to relief agencies by $74.91; an additional story of 700 words (an average among major daily newspapers) therefore raises donations by $52,437, or 18.4% of average total Internet donations, ceteris paribus. Restricting the sample to the five agencies that recorded donations on weekends affords similar results: 700 additional words raise Internet donations by $41,335, or 20.2% of average total Internet donations. In addition, the effects of passing time, the tax extension, and weekends are very similar to the estimated coefficients calculated when assessing the impact of television coverage.

While the above results are robust across specifications, endogeneity concerns remain. As noted in section 4, both simultaneity and omitted variable bias may be problematic. For example, approximately 2% of the articles related to the tsunami published by the newspapers under consideration focused on the response by private donors, suggesting the existence of reverse causality. In terms of omitted variable bias, other media coverage of the disaster may also influence donations; for example, many online news sites (including websites maintained by the New York Times and Wall Street Journal) provided direct links to relief agencies' websites, thereby facilitating online giving. Similarly, some relief agencies purchased advertising and/or undertook direct mail campaigns during this period, while others were identified in the media specifically for their assistance to tsunami victims. Finally, developments in the tsunami story itself may influence donations as, for example, the death toll rises over time. Thus, estimates of the effect of media coverage may be biased upward.

Fortunately, these problems may be addressed using instrumental variables. The number of U.S. military casualties announced each day in Iraq is used as an instrument for media coverage of the tsunami. Announcements of U.S. casualties in the Iraq conflict compete with the tsunami disaster for coverage in U.S. media (11) while remaining plausibly exogenous to Internet donations made to the seven relief agencies. Table 4 gives the first-stage results for total nightly news coverage on the three major television networks (column 1) and for total newspaper coverage in the New York Times and Wall Street Journal (column 2). Given the large number of days without any coverage of the tsunami in the 100 days following the disaster, the first stage is estimated as a Tobit.

The number of U.S. casualties announced each day has a strong negative effect on media coverage of the tsunami. The F-statistic is 9.54, exceeding Stock and Yogo's (2005) critical value for rejecting the hypothesis that the true significance level of hypothesis tests based on two-stage least squares is below 15% when the nominal level is 5% (Murray 2006). Unfortunately, with a first-stage F-statistic of 1.51, announcements of U.S. military casualties in Iraq are not a valid instrument for newspaper coverage of the tsunami, possibly because newspapers are less constrained in the number of stories covered on any given day than is broadcast news. Thus, the instrumental variables estimates will focus only on the effect of television coverage.

Table 5 describes the effect of television coverage of the tsunami on total private donations to relief agencies using linear IV estimation to control for endogeneity. Because the first stage is estimated with a Tobit model, errors are calculated using nonparametric bootstrapping methods with 500 repetitions. According to these estimates, one additional minute of tsunami coverage on the nightly news raises total Internet donations to the seven agencies by $6559, or 2.3% of the average total daily donation, significant at the 0.10 level. Restricting the sample to the five agencies that immediately record weekend donations, the effect of an additional minute of television news coverage is to raise donations by $5821, or 2.8% of the average total. The fact that these point estimates are much smaller than those presented in Table 2 underscores the notion that news coverage of the humanitarian disasters is but one determinant of charitable giving, albeit an economically and statistically important one.

6. Conclusion

This paper investigates the empirical relationship between media coverage of humanitarian crises and total private donations to relief agencies. The 2004 tsunami provides an excellent case study because it prompted unprecedented media coverage, unprecedented charitable giving, and a great deal of speculation about the relationship between the two.

We assess the effects of tsunami-related reporting on ABC Worm News Tonight, the CBS Evening News, and NBC Nightly News and tsunami-related articles in the New York Times and the Wall Street Journal on donations to seven U.S. charities that assisted tsunami victims. Because Internet donations played an important role in facilitating giving after the tsunami, because they are more likely to reflect contemporaneous media coverage of the disaster, and because they are not subject to recall error, we use donations made online to proxy for all donations. We also control for donor fatigue via a variable measuring the number of days after the tsunami that donations were made, for tax incentives via a dummy indicating the last day that donations were eligible for 2004 deductions, and for weekends.

We find that an additional minute of tsunami coverage on the evening news increases that day's donations to five agencies that record donations during each day of the study period by $47,092, or 20.8% of average total daily donations. Similarly, an additional word allocated to tsunami coverage in major newspapers increases daily donations to those five agencies by $59.05, or 0.029% of average total daily donations. To account for endogeneity problems, announcements of U.S. military casualties in Iraq instrument for television news coverage; the resulting point estimates are much smaller; although, the effect remains positive and significant at the 0.10 level.

Finally, we find that donations decay over time, providing evidence of donor fatigue. However, the Tsunami Disaster Aid Tax Relief Act, which extended tax deductions for tax year 2004 until January 31, 2005, for tsunami-related charitable contributions, contributed to record giving. This result provides further evidence that tax policy complements media coverage in inducing donations to charities as public interest begins to wane.

From the perspective of economic theory, U.S. residents donating to victims half a world away provides further evidence for Andreoni's (1989, 1990) "warm glow" motivation for individual altruism. Because most Internet donations are unheralded, it further appears that many donors to tsunami relief were motivated neither by conspicuous giving for socioeconomic recognition (Glazer and Konrad 1996) nor by opportunities to partake in "impact philanthropy" (Duncan 2004). Instead, people may have participated in charitable giving at these record levels precisely because so many others were also giving (Sugden 1984).

From a broader social perspective, our results demonstrate the causal impact of media coverage of humanitarian crises on charitable giving. This conclusion suggests that encouraging media to keep humanitarian crises in the news is in the best interest of relief agencies, as they themselves have already discovered. (12) It is our hope that the television and newspaper media will also recognize this relationship and that the beneficiaries of relief agencies' efforts will enter into their calculus when deciding which news stories to present. In this way, celebrity trials may no longer push disasters out of the media spotlight.

We are grateful to Claudio Agostini, Michael Donihue, Martin Farnham, Rebecca Graninger, Daniel Hamermesh, Melanie Siben, Tom Tietenberg, Andreas Waldkirch, seminar participants at Colby College, and two anonymous referees for numerous helpful comments and suggestions. We would also like to acknowledge Reuters AlertNet and Chuck Lakin for their assistance in identifying data sources and CARE USA, Catholic Relief Services, Mercy Corps, SurfAid International, and four other relief agencies for providing data for the purpose of this research. Finally, Philip H. Brown would like to thank the International Food Policy Research Institute for sabbatical support during the preparation of this manuscript.

Received October 2006; accepted June 2007.

References

Andreoni, James. 1989. Giving with impure altruism: Applications to charity and Ricardian equivalence. Journal of Political Economy 97:1447-58.

Andreoni, James. 1990. Impure altruism and donations to public goods: A theory of warm-glow giving. Economic Journal 100:464-77.

Arrow, Kenneth. 1972. Gifts and exchanges. Philosophy and Public Affairs 1:343-62.

Auten, Gerald R., Holger Sieg, and Charles T. Clotfelter. 2002. Charitable giving, income, and taxes: An analysis of panel data. American Economic Review 92:371-82.

Breeze, Beth. 2005. Research review. Charity Times. 12 July, p. 19.

CARMA International. 2005. The CARMA report on western media coverage of humanitarian disasters. Washington, DC: CARMA International.

Clotfelter, Charles T. 1985. Federal tax policy and charitable giving. Chicago: University of Chicago Press.

Cooperman, Alan, and Jacqueline Salmon. 2005. Charities report record donations: $337 million in private funds raised for relief. Washington Post. 8 January, p. A1.

Duncan, Brian. 2004. A theory of impact philosophy. Journal of Public Economics 88:2159-80.

Eldridge, Christopher. 2005. The tsunami, the Internet and funding for forgotten emergencies. Humanitarian Exchange 30:39-41.

El Nasser, Haya. Poll: U.S. donations still flowing to Asia. USA Today. Accessed 10 January 2005. Available at http:// www.usatoday.com/news/world/2005-01-10-poll-donations-tsunami_x.htm.

Falk, Armin. 2004. Charitable giving as a gift exchange: Evidence from a field experiment. CEPR Discussion Paper No. 4189.

Glazer, Amahai, and Kai A. Konrad. 1996. A signaling explanation for charity. American Economic Review 84:1019-28.

Greene, Pamela, and Robert McClelland. 2001. Taxes and charitable giving. National Tax Journal 54:433-53.

Hall, C. Michael, and Steven Page. 2000. Tourism in south and southeast Asia. Oxford: Butterworth-Heinemann.

Horrigan, John B. 2006. For many home broadband users, the internet is a primary news source. Accessed 26 June 2005. Available at http://www.pewinternet.org/pdfs/PIP_News.and.Broadband.pdf.

Kingma, Bruce Robert. 1989. An accurate measurement of the crowd-out effect, income effect, and price effect for charitable donations. Journal of Political Economy 97:1197-207.

Murray, Michael P. 2006. Avoiding invalid instruments and coping with weak instruments. Journal of Economic Perspectives 20:111-32.

Newspaper Association of America. Top 50+ newspapers reporting to ABC. Accessed 25 June 2006. Available at http:// www.naa.org/nadbase/top50reporting.html.

Olsen, Gorm Rye, Nils Carstensen, and Kristian Hoyen. 2003. Humanitarian crises: What determines the level of emergency assistance? Disasters 27:109-26.

Pew Research Center for the People and the Press. Pew Research Center biennial news consumption survey. Accessed 26 June 2005. Available at http://people-press.org/reports/pdf/215.pdf.

Ribar, David C., and Mark O. Wilhelm. 1995. Charitable contributions to international relief and development. National Tax Journal 48:229-44.

Roberts, Russell D. 1984. A positive model of private charity and public transfers. Journal of Political Economy 92:136-48.

Rose-Ackerman, Susan. 1982. Charitable giving and "excessive" fundraising. Quarterly Journal of Economies 97:193-212.

Rose-Ackerman, Susan. 1996. Altruism, nonprofits, and economic theory. Journal of Economic Literature 34:701-28.

Slavin, Barbara. Rush of donations from USA is immediate and immense. USA Today. Accessed 30 December 2004. Available at http://www.usatoday.com/news/world/2004-12-29-tsunami-aid-cover_x.htm.

Slemrod, Joel. 1989. Are estimated tax elasticities really just tax evasion elasticities? The case of charitable contributions. Review of Economics and Statistics 71:517-22.

Steinberg, Richard. 1987. Voluntary donations and public expenditures in a federalist system. American Economic Review 77:24-36.

Steinberg, Richard, and Mark Wilhelm. 2003. Tracking giving across generations. New Directions for Philanthropic Fundraising 42:71-82.

Stock, James H., and Motohiro Yogo. 2005. Testing for weak instruments in linear IV regression. In Identification and inference for econometric models: A Festschrift in honor of Thomas Rothenberg, edited by James H. Stock and Donald W. K. Andrews. Cambridge, UK: Cambridge University Press, pp. 80-108.

Strom, Stephanie. Tsunami followed by another kind of flood: U.S. citizens' dollars. New York Times. Accessed 30 December 2004. Available at http://www.nytimes.com/2004/12/30/ international/worldspecial4/30aid.html.

Strom, Stephanie. Relief groups hail level of donations by individuals. New York Times. Accessed 1 January 2005. Available at http://www.nytimes.com/2005/01/01/international/ worldspecial4/01charity.html.

Sugden, Robert. 1984. Reciprocity: The supply of public goods through voluntary contributions. Economic Journal 94:772-87.

UN Office of the Special Envoy for Tsunami Recovery. Country fact sheets: The human toll. Accessed 25 June 2006. Available at http://www.tsunamispecialenvoy.org/country/humantoll.asp.

Wallace, Nicole, and Ian Wilhelm. A year of progress, a year of setbacks. Chronicle of Philanthropy. Accessed 8 December 2005. Available at http://www.philanthropy.com/premium/ articles/vl8/i05/05000601.htm.

Warr, Peter G. 1982. Pareto optimal redistribution and private charity. Journal of Public Economics 19:131-8.

Wynter, Alex. 2005. Humanitarian media coverage in the digital age. In Red Cross Red Crescent worm disasters report 2005. Broomfield, CT: Kumarian Press, pp. 126-49.

(1) Oxfam Canada incorporated this perception in a highly successful advertisement in nationally circulated newspapers. A large headline shouted "O. J. SIMPSON LIMITED TO TEN VISITORS A DAY." Below, in much smaller print, the text read, "In other news: '500,000 Rwandan slaughtered. Another half million fighting for lives in refugee camps.'"

(2) Recent films shot in Phuket, Thailand, an area particularly hard hit by the tsunami, include Tomorrow Never Dies (1997), The Beach (2000), and Bridget Jones: The Edge of Reason (2004).

(3) Note that the scales in the two figures are different to better highlight small changes after 27 days.

(4) CARMA International (2005) notes that the comparatively small number of western victims of the tsunami received a disproportionate share of media attention.

(5) Television news data were provided by the Television News Archive at Vanderbilt University.

(6) Article word counts were provided by Proquest Newspapers.

(7) An alternative specification entails including a 2004 dummy in the empirical specification; results are quite similar to those obtained from beginning the analysis on January 1.

(8) We are grateful to an anonymous referee for this insightful observation.

(9) Daily announcements, including the number of casualties among U.S. military personnel, are achieved by the U.S. Department of Defense.

(10) -199,433 x (1n 10 - 1n 9) = 20,107.

(11) The simple correlation between announcements of U.S. military casualties in Iraq and total nightly news coverage of the Iraq conflict is 0.433.

(12) For example, relief workers from Oxfam provided commentary about tsunami cleanup on all three network evening news broadcasts in late March 2005. Some humanitarian agencies even create news, as did the International Rescue Committee in awarding its Freedom Award to Presidents Bush and Clinton for their efforts on behalf of tsunami victims, which pushed the tsunami back into the news in mid-November 2005.

Philip H. Brown * and Jessica H. Minty ([dagger])

* Department of Economics, Colby College, 5246 Mayflower Hill, Waterville, ME 04901, USA; E-mail phbrown@colby.edu; corresponding author.

([dagger]) Analysis Group, Inc., 111 Huntington Avenue, 10th Floor, Boston, MA 02199, USA; E-mail jminty@analysisgroup. com.

Table 1. Summary Statistics

Variable Unit Obs. Mean

Total donations to all
 seven agencies Dollars 95 $285,202
Total donations to five
 agencies that recorded
 charitable giving on
 weekends Dollars 95 $205,031
Donations to agency 1 (a) Dollars 95 $142
Donations to agency 2 (a) Dollars 95 $54,341
Donations to agency 3 (a) Dollars 95 $62,053
Donations to agency 4 (a) Dollars 65 $3,682
Donations to agency 5 (a) Dollars 95 $74,740
Donations to agency 6 (a) Dollars 95 $13,755
Donations to agency 7 (a) Dollars 65 $113,491
Total television coverage Minutes 95 3.96
ABC television coverage (b) Minutes 95 1.23
CBS television coverage (b) Minutes 95 1.10
NBC television coverage (b) Minutes 95 1.63
Total newspaper coverage Words 95 2956
NYT coverage (c) Words 95 1628
WSJ coverage (c) Words 95 1327

Variable Std. Dev. Min. Max.

Total donations to all
 seven agencies $547,046 $5,841 $3,010,581
Total donations to five
 agencies that recorded
 charitable giving on
 weekends $450,343 $4,450 $2,715,758
Donations to agency 1 (a) $541 $0 $5,000
Donations to agency 2 (a) $108,453 $1,143 $605,000
Donations to agency 3 (a) $147,427 $0 $886,702
Donations to agency 4 (a) $13,060 $0 $98,346
Donations to agency 5 (a) $171,716 $0 $1,083,890
Donations to agency 6 (a) $27,914 $0 $140,006
Donations to agency 7 (a) $146,566 $1,670 $662,732
Total television coverage 8.24 0 35.67
ABC television coverage (b) 3.15 0 18.00
CBS television coverage (b) 2.70 0 12.67
NBC television coverage (b) 3.46 0 14.83
Total newspaper coverage 5062 0 30,020
NYT coverage (c) 2640 0 15,132
WSJ coverage (c) 2798 0 14,888

(a) Receipts measuring daily donations made via the Internet provided
by seven U.S.-based relief agencies. Two agencies recorded donations
made during weekends on the subsequent Monday. Donations records for
every day of the study period are therefore available for five
agencies.

(b) The duration of tsunami coverage on each of the nightly news
broadcasts was provided by the Television News Archive at Vanderbilt
University.

(c) The length of tsunami-related newspaper stories was provided by
Proquest Newspapers. The Wall Street Journal is not published on
weekends or holidays, a point reflected in the lower number of
observations.

Table 2. Effect of Nightly Television News Coverage on Donations to
Relief Agencies, January 1-April 5 (Ordinary Least Squares)

 Donations to Seven Agencies

Variable Units (1) (2)

Total Minutes 47,092 ***
 coverage (8647)
ABC Minutes 82,175 ***
 (25,282)
CBS Minutes

NBC Minutes

Days after Log days -199,433 *** -389,270 ***
 (60,734) (43,859)
January 31 Dummy 1,110,515 *** 989,326 ***
 (26,914) (39,199)
Weekend Dummy -151,168 *** -141,711 **
 (42,800) (59,374)
Constant 885,146 *** 1,686,618 ***
 (239,957) (188,145)
Obs. 95 95
R-squared 0.88 0.79

 Donations to Seven Agencies

Variable (3) (4)

Total
 coverage
ABC
CBS 94,920 ***
 (31,605)
NBC 92,985 ***
 -25,205
Days after -384,001 *** -268,843 ***
 (68,757) (55,390)
January 31 1,000,316 *** 1,041,974 ***
 (47,681) (37,168)
Weekend -120,657 * -218,002 ***
 (64,040) (61,274)
Constant 1,656,748 *** 1,202,419 ***
 (285,280) (229,722)
Obs. 95 95
R-squared 0.78 0.80

 Donations to Five Agencies

Variable (5) (6)

Total 42,735 ***
 coverage (7794)
ABC 73,164 ***
 (20,050)
CBS
NBC
Days after -124,937 ** -300,964 ***
 (51,408) (43,915)
January 31 547,660 *** 435,035 ***
 (23,031) (31,497)
Weekend -39,625 -31,600
 (30,176) (46,333)
Constant 513,797 ** 1,257,220 ***
 (204,379) (182,668)
Obs. 95 95
R-squared 0.88 0.77

 Donations to Five Agencies

Variable (7) (8)

Total
 coverage
ABC
CBS 82,501 ***
 (27,639)
NBC 89,121 ***
 (21,448)
Days after -301,028 *** -170,609 ***
 (65,574) (43,328)
January 31 441,310 *** 496,033 ***
 (41,396) (28,201)
Weekend -13,990 -102,504 **
 (50,159) (48,741)
Constant 1,251,175 *** 729,093 ***
 (270,293) (178,807)
Obs. 95 95
R-squared 0.75 0.81

Robust standard errors in parentheses.

* Significant at 10%.

** Significant at 5%.

*** Significant at 1%.

Table 3. Effect of Daily Newspaper Coverage on Donations to Relief
Agencies, January 1-April 5 (Ordinary Least Squares)

 Donations to
 Seven Agencies

Variable Units (1)

Total Words 74.91 ***
 coverage (12.34)
NYT Words
WSJ Words
Days after Log days -203,427 ***
 (59,708)
January 31 Dummy 1,131,825 ***
 (35,531)
Weekend Dummy
Constant 820,469 ***
 (253,154)
Observations 95
R-squared 0.85

 Donations to Seven Agencies

Variable (2) (3)

Total
 coverage
NYT 109.97 ***
 (31.11)
WSJ 113.84 ***
 (22.07)
Days after -292,248 *** -323,157 ***
 (82,704) (57,135)
January 31 995,859 *** 1,092,502 ***
 (48,444) (36,875)
Weekend -172,035 ***
 (59,502)
Constant 1,250,590 *** 1,343,565 ***
 (356,089) (238,313)
Observations 95 95
R-squared 0.77 0.84

 Donations to Five Agencies

Variable (4) (5) (6)

Total 59.05 ***
 coverage (14.40)
NYT 100.08 ***
 (30.08)
WSJ 81.75 ***
 (24.99)
Days after -179,290 ** -208,340 *** -293,332 ***
 (72,415) (75,962) (65,118)
January 31 508,226 *** 444,034 *** 562,833 ***
 (43,455) (42,603) (42,536)
Weekend -58,555
 (49,486)
Constant 702,537 *** 841,844 ** 1,199,984 ***
 (314,335) (326,978) (273,243)
Observations 95 95 95
R-squared 0.80 0.74 0.77

Robust standard errors in parentheses.

* Significant at 10%.

** Significant at 5%.

*** Significant at 1%.

Table 4. Effect of U.S. Military Casualties in Iraq on Tsunami
Coverage, January 1-April 5 (Tobit)

 Total
 Total TV Newspaper
 Coverage Coverage
Variable Unit (1) (2)

U.S. military casualties Number -1.08 *** -121.24
 in Iraq by date of (0.58) (98.60)
 announcement (a)
Days after Log days -14.64 *** -6376.52 ***
 (1.56) (608.04)
January 31 Dummy -44.75 -2254.26
 (77.72) (3913.25)
Weekend Dummy -2.14
 (2.46)
Constant 57.14 *** 26,480.86 ***
 (5.86) (2320.23)
Observations 95 95
F(1,92) instrument equals 0 9.54 1.51
Prob. > F 0.003 0.2219

Robust standard errors in parentheses.

(a) Announcements of U.S. military conflicts in the Iraq conflict are
archived by the U.S. Department of Defense.

* Significant at 10%.

** Significant at 5%.

*** Significant at 1%.

Table 5. Effect of Nightly Television News Coverage on Donations to
Relief Agencies, January 1-April 5 (Linear IV)

 Donations to Donations to
Variable Unit Seven Agencies Five Agencies

Total TV news Minutes 6559 * 5821
 coverage (a) (3926) (3121)
Days after Log days -519,107 *** -416,815 ***
 (78,568) (63,133)
January 31 Dummy 1,160,350 *** 586,388 ***
 (175,861) (130,246)
Weekend Dummy - 158,280 ** -46,397
 (64,256) (52,765)
Constant 2,302,789 *** 1,806,735 ***
 (332,018) (272,672)
Observations 95 95
R-squared 0.66 0.61

Robust standard errors in parentheses.

(a) Fitted values using instrumental variables.

* Significant at 10%.

** Significant at 5%.

*** Significant at 1%.