Government Press Releases and Citizen Perceptions of Government Performance: Evidence from Google Trends Data

ABSTRACT:

The “media-malaise” thesis argues that media coverage of government activities may have only a minimal or even negative effect on citizens’ perceptions of government performance. This contention is challenged by showing a positive association between government press releases and citizen perceptions. Two streams of empirical evidence are presented. First, a positive association is shown between the number of times the New York City mayor’s name appears in official city press releases and the volume of Internet searches for his name. Then, a positive correlation is shown between Internet search volume for the mayor’s name and the mayoral approval rating.

KEYWORDS::

• citizen perception
• mayoral approval
• media-malaise
• press releases
• public communication

Acknowledgments

The author would like to thank the three anonymous reviewers and Dr. Kaifeng Yang for their helpful comments.

Notes

Public communication is also important because, by influencing citizens’ perceptions of government, it may affect the way citizens interact with the government and their willingness to coproduce public service (e.g., Hong, 2015).

Previous studies (e.g., Hood, 2006; Norris, 2000) generally focus on the relationship between government transparency and trust in government. Although the main interest of the present study lies in citizens’ perceptions of government performance rather than trust, we do not distinguish strictly between these concepts in our literature review, as prior research (Holzer & Zhang, 2004; Kampen, Van de Walle, & Bouckaert, 2006; Yang & Holzer, 2006) finds strong associations between them.

For example, the news media tend to give inadequate coverage of candidate statements (Hallin, 1992), provide journalistic interpretations of political information rather than simply describing and delivering it to the public (Patterson, 1993), and overemphasize the content of negative campaign ads (Johnston, Hagen, & Jamieson, 2004).

See also Grosso and Van Ryzin (2011) and Halachmi and Greiling (2013).

Further evidence suggests that such efforts are often compensated. For instance, over an 8-week period, Turk (1986) tracked the content of both newspapers and press releases provided by public officials with Louisiana state government agencies, finding that slightly over half of the press releases provided to journalists were published with minimal content modification.

Previous research suggests that even a simple count of press releases issued on a given policy issue is often positively correlated with the amount of resources allocated to that issue—for example, the budget allocated to implement a related policy (Hong & Sohn, 2014).

Because the Google Trends data are normalized in this way, they must be modified in order to generate a comparable 5-year daily time series. Google Trends makes daily search-volume data available only for 3-month windows, while weekly search volume is available for 5-year windows. A problem arises because the daily data are normalized for every 3-month window, so that generating a year-long daily time series requires also downloading weekly time series data for the relevant 5-year period and adjusting each 3-month block of daily data based on the weekly data. Specifically, at the beginning of each 3-month window, one obtains the corresponding normalized value from the weekly data. This initial value is multiplied by the percentage change between days, yielding the new daily value. Repeating this for each 3-month window, one obtains daily relative search volumes for the 5-year window. This modification process is widely used by researchers. For instance, see http://erikjohansson.blogspot.kr/2013/04/how-to-get-daily-google-trends-data-for.html.

The nominal value of total expenditures is deflated using the inflation rate.

ARIMA(1,0,1) is not widely used, as its approach is often better represented by simpler models (McDowall, McCleary, Meidinger, & Hay, 1980, p. 47). Nevertheless, iterative comparisons of model residuals indicate that ARIMA(1,0,1) outperforms simpler models such as ARIMA(1,0,0) or ARIMA(0,0,1).

The media likely play a critical mediating role between the two variables in Model 1. Thus, a positive association should be observed between the mayor’s salience in press releases and media coverage of the mayor. In addition to the results presented in Table 2, we also conducted this “reality check” and found that increased salience of the mayor in government press releases is significantly associated with the number of times he is mentioned in The New York Times on a given day (not reported).

Public communication about policy and priorities may have a circular relationship with the quality of performance measurement as well as the actual performance (e.g., see Melkers, 2006).

Additional information

Notes on contributors

Sounman Hong

Sounman Hong is an Assistant Professor at the College of Social Sciences, Yonsei University, Seoul, Korea.