Citizen expectations and government performance: Findings from a survey experiment

In the public management literature over the last few years, a number of empirical studies have examined what is called the expectancy-disconfirmation (E-D) model of citizen satisfaction with public services. Briefly stated, the E-D model posits that citizens judge public services based—not simply on the level of experienced (or perceived) service quality—but on an implicit comparison of service quality with prior expectations. The gap between expectations and performance, in other words, is thought to be a better predictor of satisfaction than just performance alone. Van Ryzin (2004) found evidence in support of the E-D model using survey data from New York City, and later replicated his findings using a national online panel (Van Ryzin 2006). Roch and Poister (2006), using data from a statewide survey in Georgia, also report evidence that the disconfirmation of expectations has a strong influence on satisfaction. And using a survey of citizens in England, James (2009) similarly finds that satisfaction depends substantially on the gap between performance and expectations. The E-D model has significance beyond academic curiosity, as it can help explain possible bias in citizen surveys when used as performance measures, and it has implications for how governments communicate with citizens in an effort to manage expectations—especially in times of scarcity and service cutbacks. But these prior studies have been based on survey (observational) data and on subjective measures of expectations and performance, which are potentially endogenous. To overcome these problems, the present study employs an experimental design in which expectations and performance are experimentally manipulated. Using a national online panel, respondents were randomly assigned to receive either low- or high-expectations statements from a hypothetical government official. Next, respondents were randomized to view either lower- or higher-performance street cleanliness, based on photographs used by New York City’s street cleanliness scorecard. The result is a 2 x 2 factorial design, with a total of about 1000 respondents (or about 250 in each of the four arms of the experiment). A manipulation check confirmed that the presentation of the high-expectations statement, in contrast to the ow-expectations statement, did indeed raise self-reported expectations for public service quality. But exposure to high- versus low-expectations statements did not significantly influence satisfaction ratings. Performance (the photos of street cleanliness), however, did influence satisfaction ratings. The interaction of expectations and performance was not significant. Thus, results of this experiment cast doubt on the role of expectations found in previous, observational studies. One explanation is that self-reported expectations and perceived performance are highly correlated, as indeed they are in this study, but only because they both reflect an underlying common orientation toward government services. When using randomly assigned expectations statements to instrument for self-reported expectations, the correlation disappears. This suggests that self-reported expectations, and related measures of disconfirmation, are endogenous in most observational studies and thus result in over-estimates of the magnitude of the expectations effect on citizen satisfaction. The practical implication is that citizens may form their satisfaction judgments in ways that more directly reflect objective government performance than previously thought.