“Never awake me when you have good news to announce, because with good news nothing presses; but when you have bad news, arouse me immediately, for then there is not an instant to be lost.”
— Napoleon Bonaparte
Is ‘bad’ news more valuable than ‘good’ news, not just for Napoleon but also for the typical consumer? Does bad news garner more attention when it is uncommon? Or does ‘news’ mirror objective reality? We examine these questions by studying how the media reported infectious disease trends in the US at the end of the 19th and the beginning of the 20th centuries.
Death from infectious disease was common at the beginning of the last century. Cities such as Philadelphia drew their drinking water from rivers contaminated with sewage. Some areas of New York City had access to clean mountain water whereas others were dependent on well water or imported water supplies. Typhoid fever, a waterborne illness for which there was no cure, affected all age groups, killing 10-20% of its victims. Filtration, chlorination, closure of raw sewage outfalls, and the introduction of clean water supplies rapidly reduced both the level and variance of typhoid fever case and death rates (Alsan and Goldin 2015, Cutler and Miller 2005, Troesken 1999).
We focus on typhoid for three reasons: because weekly data are readily available; because there were sharp declines in typhoid case and death rates after clean water interventions, allowing us to examine news reports under very different mortality regimes; and because individuals had a clear self-interest in knowing what the trends were so that they could protect themselves against typhoid outbreaks (by the 1890s people knew how to do so). Individuals would thus respond to news coverage of outbreaks
Did media coverage of typhoid fever reflect objective typhoid disease trends?
The media may both provide readers with the information they want (Gentzkow and Shapiro 2008) as well as carry out public health campaigns in editorial and news pages. In neither case will news reports be unbiased, i.e. determined purely by death or case rates.
- If readers want sensationalist stories, newspapers will focus on the unusual and thus over-emphasise low-risk causes of death (as found in Frost, Frank and Maibach 1997).
- If consumers find bad news more useful than good news then increases rather than decreases in mortality will be emphasised.
- Editors' desire to nudge politicians on public health expenditures and readers on private measures of self-protection may lead to ‘over-reporting.'
Newspaper campaigns in 1894 and 1895 contributed to the public acceptance of diphtheria antitoxin and to public funding for antitoxin (Hammonds 1999:93-117). Case studies suggest that recent media campaigns have reduced smoking, cocaine use by teenagers, HIV infection rates, and deaths from Reye's syndrome (Hornik 2008) but that the media also spread sensationalist misinformation about vaccines (Freed, Katz and Clark 1996). Several different research designs have yielded clear evidence of the impact of media on behaviour. Outcome variables have included voter turnout (Gentzkow 2006, Gentzkow et al 2011), voting outcomes (DellaVigna and Kaplan 2007, Gentzkow et al 2011) and disaster relief contributions (Eisensee and Stromberg 2007).
We assume that consumers demanded typhoid information. In recent times the number of newspaper articles on topics of concern to consumers such as crime, inflation, and disease closely track self-reports of concern in polls and ameliorative actions by consumers (Lowenstein and Mather 1990). Why did consumers demand typhoid information?
Typhoid spread primarily through drinking water contaminated with the wastes of infected individuals. Other modes of transmission were direct contact with a contaminated privy, with the wastes of a typhoid patient, with food prepared by a typhoid carrier, or indirect contact with a contaminated privy through flies. Precautions that individuals could take included using individual water filters, bringing water to a rolling boil, pasteurising milk, thoroughly cooking all vegetables, peeling fruit, disinfecting privies and homes, and sealing privies and homes from flies.
We hypothesise that:
- The number of news reports is linked to the number typhoid death or case rates positively and negatively.
- The number of news reports is more closely tied to actual typhoid death or case rates after clean water interventions.
This could arise either because of diminishing rates of return to self-protection before the intervention, increased stigma from a disease transmitted through the oral-faecal route after the intervention, or a clearer signal from change in death and case rates after the intervention when both the mean and variance of mortality had fallen.
An unexpected change in typhoid death or case rates will have a bigger impact on news reports when the change is an unexpected increase.
This phenomenon could arise either from endowment or reference point effects (Kahneman and Tversky 1979) or from bad news being more valuable in a world where there are high gains in the probability of survival to making the correct self-protection choices.
We created panel data from newspaper articles and from weekly typhoid death and case rates for New York City, Baltimore, Boston, Chicago, Washington DC, and Philadelphia. We obtained weekly counts of the total number of newspaper articles and the number of newspaper articles mentioning typhoid and also typhoid and the city using mechanised searches of The New York Times, The Baltimore Sun, The Boston Globe, The Chicago Tribune, Washington Post and The Philadelphia Inquirer.1 Reports include all types of news, including reports from local public health officials, stories of outbreaks, society news, obituaries of well-known individuals, editorials, and appeals to charity.
Cities had to have both digitised and indexed newspapers and good weekly typhoid death data to be included in our sample. Our final data set includes data for New York City for all weeks for 1890-1938; and, with some weeks missing, for Chicago for 1896-1932; Baltimore for 1900-1932; Boston for 1890-1932; Philadelphia for 1901-1922; and Washington DC for 1890-1932. We also created dummy indicators for a holiday during that specific week and for a major news event that week.
Typhoid death and case rates
The major interventions in each of our cities take the form of either cleaning up the water supply obtained from the nearby river through chlorination or filtration, or obtaining new, clean sources of water.
Figure 1, which also shows missing data, and Table 1 suggest that the interventions were effective in lowering typhoid mortality and case rates. In the sample as a whole, typhoid death rates per 100,000 were 0.8 prior to any intervention, 0.4 after the first intervention but before the second, and 0.1 after the second intervention. Prior to the first intervention, death rates per 100,000 varied widely across cities with highs of 1.0 and 1.5 in Philadelphia and Washington DC, respectively and a low of 0.4 in New York City. After both interventions, death rates per 100,000 varied from a high of 0.2 in Baltimore to a low of 0.03 in New York City. Case rates also fell after an intervention and converged across cities. These interventions were statistically significant negative predictors of death rates, controlling for a year trend.
Figure 1. Weekly typhoid death rates by city
Note: See Costa and Kahn (2015) for sources.
Table 1. Mean death and case rates, before, between, and after interventions
Note: Standard deviations in parentheses. Case rates are not available for Boston and Washington DC prior to the first intervention.
Figure 2 plots the deviation from the expected death rates adjusted for the standard deviation of death rates in each city. In all cities deviations were high prior to any interventions and then narrow after both interventions.
Figure 2. Deviations from expected typhoid death rates adjusted for the standard deviation of typhoid death rates
Note: See Costa and Kahn (2015) for estimation details.
Results: Death and the media
Figure 3 shows smoothed plots of typhoid death rates and of the percentage of typhoid articles. It suggests that reports of typhoid followed mortality patterns, but an increase in city death rates led to more news reports in low mortality regimes. For example, in New York City, the up-tick in typhoid mortality rates in the 1920s is associated with an increase in news reports that is greater than the increase in the early 1890s when typhoid mortality rates spiked higher. Increases in reporting that are not related to city death rates were often associated with world events such as concern over typhoid epidemics during the Spanish-American War and WWI.
Figure 3. Weekly typhoid death rates and percentage of typhoid articles by city
Note: Death rates and the percentage of articles were smoothed using a lowess estimator.
We find that although news reports were positively associated with mortality and case rates, coverage was biased and not just because of public health campaigns. When we use the full panel of cities and control for holidays, major news events, year, and city fixed effects, we find the media were more likely to report changes in typhoid after the clean water interventions than before. Our results hold both for all and for local news.
We also find that the media respond more to bad (unexpected increases in death rates) than to good (unexpected decreases in death rates) news with our full set of controls. The losses to individuals of not knowing bad news are likely to outweigh the losses of not knowing good news. Not knowing good news may lead to too much time and money spent on self-protection. Not knowing bad news may lead to death if not enough time and money is spent on self-protection.
Several hypotheses could explain our results. Although we cannot distinguish between them, all of the hypotheses emphasise that what mattered was how useful the information was to consumers. After the clean water interventions, individuals probably cut back on costly self-protection actions. Knowledge of disease outbreaks may thus have been more valuable after the interventions because the returns to self-protective actions were greater. Knowledge of disease outbreaks may also have been more valuable after the intervention if the stigma of a disease transmitted through contaminated faecal matter was greater after the intervention, thus making costly self-protection measures even more valuable. After the interventions, both the mean and variance of death and case rates fell, thus making any information more informative in a classical signal extraction model. Our findings on asymmetries in reporting are consistent either with high gains in survival probabilities to making the correct self-protection choices, or with prospect and psychological theories of reference points in which information on bad events is more useful than information on good events.
Our findings have implications not only for how the media report about progress, but also for the economic incidence of urban public health improvements. If improvements are common knowledge both to incumbent city residents and to non-residents, then standard no-arbitrage compensating differentials logic implies that landowners in the cities and neighbourhoods that experienced the largest reduction in death would enjoy the windfall of higher prices. But, if outsiders are unaware of the localised quality of life improvements then incumbent renters could gain the windfall. In standard compensating differential models it is assumed that households have full information about the attributes of each choice at each point in time. However, in the case of endogenous dynamic attributes such as typhoid death rates this assumption appears to be extreme. With dynamic and stochastic disamenities, the media may play a key role in spreading information about the shifting location of specific attributes. With asymmetries in media coverage, outsiders may be unaware of progress and information frictions could generate rents for incumbents. [RB3]
Alsan, M and C Goldin (2015) Watersheds in infant mortality: The role of effective water and sewerage infrastructure, 1880-1915, Unpublished MS, Harvard University.
Costa, D L and M E Kahn (2015) “Death and the media: Asymmetries in infectious disease reporting during the health transition”, NBER Working Paper No. 21073.
Cutler, D and G Miller (2005) “The role of public health improvements in health advances: The Twentieth-century United States”, Demography, 42(1): 1-22.
DellaVigna, S and E Kaplan (2007) “The Fox News effect: Media bias and voting”, Quarterly Journal of Economics, 122(3): 1187-234.
Eisensee, T and D Strömberg (2007) “News droughts, news floods, and US disaster relief”, The Quarterly Journal of Economics, 122(2): 693-728.
Freed, G L, S L Katz and S J Clark (1996) “Safety of vacciantions: Miss America, the media, and public health”, Journal of the American Medical Association, 276(23): 1869-1872.
Frost, K, E Frank, and E Maibach (1997)”Relative risk in the news media: a quantification of misrepresentation”, American Journal of Public Health, 87(5): 842-45.
Gentzkow, M (2006) “Television and voter turnout”, Quarterly Journal of Economics, 121 (3): 931-972.
Gentzkow, M and J M Shapiro (2008) “Competition and truth in the market for news”, Journal of Economic Perspectives, 22(2):133-54.
Gentzkow, M, J M Shapiro and M Sinkinson (2011) “The effect of newspaper entry and exit on electoral politics,” The American Economic Review, 101(7): 2980-3018.
Hammonds, E M (1999) Childhood's deadly scourge: The campaign to control diphtheria in New York City, 1880-1930, Baltimore and London: Johns Hopkins University Press.
Hornick, R C (Ed.) (2002), Public Health Communication: Evidence for Behavior Change, Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Kahneman, D and A Tversky (1979), “Prospect theory: An analysis of decision under risk,” Econometrica, 47(2): 263-92.
Troesken, W (1999), “Typhoid rates and the public acquisition of private waterworks, 1880-1920,” Journal of Economic History, 59(4): 927-948.
1 See Costa and Kahn (2015) for more details