Reality versus propaganda in the formation of beliefs about privatization

Abstract

Argentina privatized most public utilities during the 1990s but re-nationalized the main water company in 2006. We study beliefs about the benefits of the privatization of water services measured immediately after the 2006 nationalization. Negative opinions about privatization prevail. We find that “reality” can change beliefs: people who had first-hand experience observing the investments made by the privatized company have a better opinion of water privatization (relative to other privatizations) than people who did not gain access to water. The effect, while statistically significant, seems small adding only 0.8 points on a 1–10 scale. Moreover, the effect of priming subjects with government propaganda against privatization has an effect that almost offsets the effect of gaining water. However, our evidence suggests that the presence of firm investments makes beliefs about the benefits of privatization less susceptible to be affected by propaganda.

Introduction

A growing literature in political science, sociology and economics has emphasized the connection between beliefs and economic organization. For example, the amount of redistribution observed in the US and Europe, or the amount of market reform that we can expect in developing countries, appears to be connected to voter's beliefs about actors or elements of the economic system, such as weather “effort pays” or if the rich are corrupt. Two dimensions of these beliefs are particularly important: their variability and their accuracy. Indeed, if these beliefs were fixed, perhaps because they were historically determined, then the possibility of changing economic systems or of implementing long lasting market reforms might be limited. And if these beliefs can diverge from reality there is of course the possibility of large welfare losses. Indeed, one question that has confronted this research is the extent to which beliefs can be maintained in the face of available evidence to the contrary.¹ A natural question, dealing with both dimensions, asks the extent to which an agent (perhaps an “ideological entrepreneur”) can persuade others of a particular point of view using old or fabricated data. And if this were possible, how do such effects of “propaganda” compare with the effects of objective changes in “reality” on the formation of beliefs (assuming these exist).

To attempt an answer to these questions, we study the formation of beliefs about the benefits of privatizing the main water company in Argentina during a period where the government made several attempts to persuade the public of its negative views on the private company, an effort that was viewed as a propaganda campaign. Specifically, in June 2006, three months after the government re-nationalized the company, we implemented a survey to elicit views about the 1990's market reforms in general and the water privatization in particular. It covered households living in middle and low income neighborhoods in the outskirts of Buenos Aires, Argentina. Two “treatments” were studied: the presence or not of firm investment and the priming or not of propaganda. Using detailed historical maps indicating which households had access to water services, we ensured that about half the addresses in our sample had gained water during the privatization period while the other half remained without access at the time of the survey. We then reminded a (random) sub-group of our sample of some negative statements made by the country's president (Mr. Kirchner) concerning the lack of investment of the water company around the time of the nationalization (exactly as it was reported in the press). The statement made by the President referred to a fact that was untrue, and this was particularly evident to the sub-sample that had gained water as a result of the company's investment after privatization. As controls, a statement that was made by the water company defending their record on investment was read to another sub-sample, while a third sub-group was not read any statement.

Our approach exploits a key feature of this episode: it produced a group that benefited substantially from privatization. Indeed, the privatized water company's investment during the 1990s brought water services to a large group of people (see Galiani et al., 2005). These households benefited greatly from the enormous convenience of having direct connection to the water service. Since the ensuing water charges were lower than what these households were paying for substitute services, we have a group that is unambiguously better off in material terms after privatization. Moreover, four years into the contract this group experienced a large drop (of 74%) in their connection charges, so these material gains were particularly salient to newly connected users. This group can be compared to the group that remained unconnected to the water network.

Several other features of this episode are helpful in our empirical design. First, the policy in question (privatization) is salient to voters. Privatization of most state owned companies was a key component of the market reforms of the 1990s in Argentina. They have been widely discussed in the media and political debates. The incumbent Kirchner government, appointed in 2003, had repeatedly discussed the problems of the privatizations in their appearances in the media and the March 2006 nationalization of the water company made the specific issue we focus on particularly salient to the public. Second, the firm that was the target of the attack was foreign owned, which increased the receptiveness of the public to the President's attacks. Third, during this episode the President gathered support for the nationalization by personally attacking the water company on repeated opportunities in the media and in political rallies for lack of investment. Thus, we have one concrete example of a political agent trying to affect people's beliefs about the privatized water service. This is helpful because, rather than designing a piece of information that we think might work as propaganda, and devising a setting in which there is a presumption that propaganda might be useful, we obtained the content and setting of our piece of propaganda from the real political “market”. The repeated nature of the president's public statements against the water company matches the episode with one theoretical dimension of propaganda campaigns.² It should be noted, however, that it is conceivable that priming in fact overstates the effect of propaganda in real markets, both because of the proximity between the stimulus and the elicitation of views in our survey and because of the lack of contextual cues. In brief, our empirical exercise studies propaganda using a priming setting, with a message that originated in an actual situation where propaganda was deemed useful (privatization) by an agent who has been successful in the political market, and employs one set of statements that were actually used as propaganda.

We find that the 1990's market reforms are unpopular, receiving relatively low scores. The average score for the privatizations is 3.07, out of a maximum possible 10. As a mild anchor to these numbers, we note that 7 or more is a passing grade in the Argentine schooling system (while 4 is a passing grade in make-up exams). The average score given to the water privatization is 3.59, very similar to the score for privatizations in general. Interestingly, the score for water privatization is significantly higher for those that received connection to the water service and for those that were not reminded of the government's propaganda against the water company.

A potential problem with relying on a simple cross-sectional approach is the presence of unobservables that might confound the estimation of the effects of interest. Indeed, it is possible that favorable opinions about privatization are correlated with income or the ideological inclination of the respondents (see, for example, Shirley, 2004). Since the group that gained water is closer in income to the middle-class (which historically has tended to favor State intervention in Argentina), than those that remained unconnected, the estimate of the difference in scores without a good control for income or ideology would provide a biased estimator of the causal effect of receiving access to water on the beliefs that individuals hold about the privatization. Even if data were available to pursue that strategy, we find that the group that gained water is different from the group that remained unconnected along other dimensions for which we do have data, such as age or skill. In order to deal with this statistical nuisance we define Water Score Gap as the score for the water privatization minus the score for all the privatizations given by the same respondent and then use it as the dependent variable in order to control for differences in unobservables in our sample.

We find that Water Score Gap is positively correlated with having gained water. The effect is statistically significant, but small in absolute value: 0.8 points on a 1–10 scale. Even those that have gained water are rating the water privatization in the low fours out of a possible 10. The effect is however large relative to the average score given to the water privatization: a gain of almost 23%. The effect of priming subjects with propaganda is statistically significant and seems quite large when compared to the effects of firm investment: those that were reminded of the statements made by the President against the water company score it approximately 0.5 points lower, or a drop of almost 14%. There is no discernible effect of reading the statement made by the company. The effects of “reality” and “propaganda” appear similar in this context (equality of the main effects cannot be rejected). Comparing them is informative because they refer to the same thing: “propaganda” focuses on a particular activity (firm investment) and our measure of “reality” concerns changes in that particular activity. Our approach, however, does not allow us to discern the relative duration of these effects.³

Our paper is connected to prior work on the formation of beliefs.⁴ Within psychology, individuals can be viewed as “lay empiricists” or “lay scientists”. Lay empiricists use empirical observations to update their view of the world, much as in Bayesian inference. In contrast, lay scientists have a “scientific” theory of the world that determines the beliefs (much as in classical inference).⁵ The possibility of persuasion was the focus of earlier work in political science, although the effects found were often described as “minimal” (see, for example, Klapper, 1960).⁶ As described in an influential paper by Iyengar et al. (1982), “Four decades ago, spurred by the cancer of fascism abroad and the wide reach of radio at home, American social scientists inaugurated the study of what was expected to be the sinister workings of propaganda in a free society. What they found surprised them. Instead of a people easily led astray, they discovered a people that seemed quite immune to political persuasion. … later research on persuasion drove home the point repeatedly: propaganda reinforces the public's preferences; seldom does it alter them (e.g., Katz and Feldman, 1962, Patterson and McClure, 1976, Sears and Chaffee, 1979).” Accordingly, research moved away from persuasion and towards the possibility of other effects of the media.⁷ Our paper is also related to a large body of work in psychology using priming, for example to investigate the possibility that messages received at one point affect beliefs at a later stage (see Schacter, 1996, for an overview of work on memory). One class of experiments finds that subjects are more likely to believe statements when they heard them before, even when they were explicitly told they were false. This is sometimes called the “illusion of truth effect” (see, for example, Hasher et al., 1977, and Begg et al., 1992) and is seen as an expression of implicit memory (where previous experiences affect later tasks, even with amnesiac subjects who claimed to be unaware of the first experience).

Work in economics on the subject focuses on the possibility of using information, perhaps strategically, to affect people's beliefs.⁸ Recent work takes a broader perspective. For example, Glaeser (2005) provides a model where citizens are persuaded to hold a negative point of view about particular groups. Citizens' willingness to be persuaded by hate-creating stories depends on the costs and benefits of acquiring information and on the existence of an out-group that is perceived to be influential politically but socially segregated. There is also previous work on the possibility that persuasion is easier to attain using categorical thinking and metaphors (as in Mullainathan et al., 2008, Lakoff, 1996) or when social networks are important (see, for example, De Marzo et al., 2003, Murphy and Shleifer, 2004). Note that if we detect persuasion when using a simple untrue fact, it is likely that less extreme forms of persuasion (for example involving fewer untrue statements) can be employed to affect people's beliefs.

Two recent papers present convincing evidence of how certain types of coverage affect voting behavior. Such a connection can arise because a particular coverage convinces viewers that some problems are more important than others, favoring candidates that emphasize those issues (agenda setting). Or it can affect voting because coverage convinces a viewer to change his or her mind (persuasion). For example, Della Vigna and Kaplan (2007) show that areas in the US where the cable operator offered Fox News Channel observed voter turnout increases relative to other areas where Fox did not enter, as well as increases in the Republican party vote share in Presidential elections. Gerber et al. (2009) designed a field experiment to measure the effect of exposure to newspapers on political behavior in Washington DC. They randomly assigned households (that were not receiving newspapers up to then) to receive copies of either a left or a right leaning newspaper and later surveyed them. They found that those treated with the left-leaning newspaper were up to 8 percentage points more likely to report voting Democrat than the control group (although no significant difference was found with the group receiving the right leaning newspaper). The paper by Gerber et al. is particularly interesting because it is able to provide evidence of persuasion versus agenda setting by looking at a battery of questions on specific issues.⁹ Our paper is also an attempt to isolate the effect of persuasion as it links misinformation on a specific issue with an opinion about that issue.

Section 2 provides a brief historical description of the privatization and subsequent nationalization of the main water company in Argentina. It also describes our data and empirical strategy. Section 3 presents our main results, while Section 4 concludes.