AEA Papers and Proceedings 2020, 110: 324–328 https://doi.org/10.1257/pandp.20201072

The Polarization of Reality †

By Alberto Alesina, Armando Miano, and Stefanie Stantcheva*

Evidence is growing that Americans are voters have been illustrated in the political sci- polarized not only in their views on policy issues ence literature. Bartels 2002 shows that party and attitudes toward government and society identification shapes (perception) of economic but also in their perceptions of the same factual indicators that can be seen as the government’s reality. In this paper, we conceptualize how to “performance indicators” e.g., unemployment think about the polarization of reality and review or inflation , with Republicans( being more opti- recent papers that show that Republicans and mistic than) Democrats on economic variables Democrats as well as Trump and non-Trump­ during the Reagan presidency.1 Similar results voters since (2016 view the same reality through about the importance of partisan assessment a different lens. Perhaps) as a result, they hold of the government’s performance in shaping different views about policies and what should perceptions of economic indicators is found in be done to address different economic and social Conover, Feldman, and Knight 1987 . More issues. recently, Jerit and Barabas 2012( shows) that The direction of causality is unclear. On the people perceive the same reality( in a) way consis- one hand, individuals could select into political tent with their political views and that learning affiliations on the basis of their perceptions of is selective: partisans have higher knowledge reality. On the other hand, political affiliation for facts that corroborate their world views and affects the information one receives, the groups lower knowledge of facts that challenge them. with which one interacts, and the media to Heterogeneous updating to identical infor- which one is exposed, all of which can shape mation is also shown for attitudes toward the perceptions of reality. Regardless of the direc- Iraq War in Gaines et al. 2007 . Prior, Sood, tion of causality, though, this is not about hav- and Khanna 2015 and Bullock( ) et al. 2015 ing different attitudes about economic or social demonstrate that( people) give “partisan” answers( ) phenomena or policies that could justifiably be to factual questions in surveys, but the partisan viewed differently from different angles. What gap is reduced if monetary incentives are offered is striking, rather, is to have different percep- for correct answers. tions of realities that can be factually checked. We highlight evidence about differences in I. Conceptual Framework perceptions across the political spectrum on social mobility, inequality, immigration, and A simple conceptual framework can help public policies. We also show that providing organize the empirical results reviewed for the information leads to different reassessments of ­full-fledged model, see Stantcheva 2019( . It is reality and different responses along the policy illustrated in Figure 1. ) support margin, depending on one’s political People can hold many “perceptions,” which leaning. are estimates of true parameters on a variety of topics such as the share of immigrants, the share Earlier Literature.—Some differences in per- of national income going to the top 1 percent, ceptions between Republican and Democratic or the elasticity of top incomes to top tax rates. These perceptions have true empirical counter- * Alesina: Harvard University, CEPR, IGIER, and NBER parts. “Policy views” are formed as functions email: [email protected] ; Miano: Harvard University of these perceptions and can range from the (email: [email protected]) ; Stantcheva: Harvard University,( CEPR, and NBER ) email: sstantcheva@fas. harvard.edu . ( † Go to https://doi.org/10.1257/pandp.20201072) to visit 1 For instance, in 1988, Democrats were more likely to the article page for additional materials and author disclo- report that unemployment and inflation had increased since sure statement s . 1980, when the opposite was true. ( ) 324 VOL. 110 THE POLARIZATION OF REALITY 325

means that the weight and updating for an iden- inal tical signal will depend on all prior perceptions. nforation A signal can move more than one perception at the same time. The weight on the signal is endogenous to perceptions as also indicated by eihte ( ( ) an arrow in the figure because it could depend on its assessed reliability,) which is yet another perception held by people. Peretion Second, imagine information is costly to acquire. Then, in addition to the interaction among existing perceptions and updating just described, the set of signals acquired is also P P P endogenous to perceptions as indicated again 1 ⇄ 2 .… N by an arrow in the figure(. People have to decide which information to) incur costs for, which will depend on their baseline perceptions. Below, we describe how people with differ- Poliy ie ent baseline perceptions indeed have different ­willingness-to-pay for information. Figure 1. From Information to Policy Views What makes learning particularly difficult in this setting is that the actual true values of the variables that people form perceptions about desired top tax rate to the ideal level of govern- either change over time e.g., the share of immi- ment intervention. Perceptions interact with one grants or are difficult( to estimate, even for another: each policy view can be a function of experts) e.g., the elasticity of unemployment to several or all perceptions. For instance, as we unemployment( benefits . will see below, a given perceived level of social Finally, while behavioral) features could also mobility will translate into different support for play a role, there is no need to suppose that if redistribution based on people’s perceptions people had the exact same vector of perceptions of the competence and trustworthiness of the there would be disagreement on policy views government.2 or updating i.e., willful ignorance or partisan How perceptions are determined depends on bias per se in( the shape of the function mapping how learning occurs. People receive “signals,” perceptions to policy views . As long as people which are pieces of information and which are have a whole set of heterogeneous) perceptions, weighted in order to be translated into a change there will be completely different policy views, in perceptions. Signals do not have homoge- and any signal will be rationally acquired and neous impacts on all people’s perceptions, and weighted on the basis (of the full )set of percep- not all people receive the same signals. People tions, which leads to different updating pro- may thus end up with very heterogeneous per- cesses as well. ceptions and misperceptions. First, suppose that information and signals are II. The American Dream costless. Even entirely rational updating rules will depend on the prior level of possibly, all Is the American dream alive? The answer perceptions, as well as on the weight( placed on) people give to this question turns out to be a the signal. The interaction among perceptions key determinant of support for redistribution. If perceived social or intergenerational mobil- ity is high, the resulting inequalities in income 2 In addition, some perceptions may put consistency and wealth are perceived as more fair, since it is constraints on others. For instance, one cannot simultane- thought that everyone had more equal opportu- ously believe that all immigrants are unemployed and that nities with which to start. one’s sector’s jobs are mostly taken by immigrants. This could be called “Schrödinger’s” immigrant, and we do not Using detailed surveys on several thou- need to rely on such mental models to rationalize the results sand respondents, Alesina, Stantcheva, presented. and Teso 2018 investigates perceptions about ( ) 326 AEA PAPERS AND PROCEEDINGS MAY 2020

­intergenerational mobility in the United States Panel A. Mobility perceptions and Europe. The authors show that not only are Percent Americans more optimistic about social mobil- Perceived probability 10 20 30 for child from bottom ity than Europeans, but they are overly optimis- quintile to... 10.5 tic given reality: they overestimate the chances ...move to the top quintile 7.8 12.0 37.4 of making it from the bottom to the top—that is, ...remain in the bottom a child born in the bottom quintile making it to quintile 29.5 33.1 19.3 the top quintile. ...move to the third quintile 18.7 24.1 Perceptions of social mobility are correlated with political orientation. Americans who iden- Panel B. Inequality perceptions tify themselves as conservative i.e., ­right-wing ( ) Has income inequality on economic issues believe that the probability increased in US 77.6 in recent decades? that a child born in the bottom quintile makes 61.3 Were top income 59.5 it to the top is 12 percent; liberals ­left-wing tax rates higher in respondents believe that it is 10.5 percent.( Thus,) 1950s−60s than today? 46.7 Was economic growth 46.5 both groups overestimate the correct answer, faster in 1933−1980 36.6 than in 1980−2010? which is 7.8 percent, implying that belief in the 40% 50% 60% 70% strongest form of the “American dream”—mak- ing it from rags to riches—is still relatively prev- Share answering yes alent. However, perceptions about the chances Democrats liberals / of making it out of poverty and into the mid- Republicans conservatives / dle class are very different across the political Reality spectrum, as illustrated in panel A of Figure 2. The perceived probability that a child born in the bottom quintile remains in that quintile as an Figure 2. Differences in Perceptions across the adult is 37.4 percent for ­left-wing respondents Political Spectrum and 29.5 percent for ­right-wing respondents the ( Note: The lines are 95 percent confidence intervals around correct number is 33.1 percent . The perceived the mean. probability such a child makes) it into the mid- Source: Panel A: Alesina, Stantcheva and Teso 2018 ; panel dle class the third quintile is 19.3 percent for B: Kuziemko et al. 2015 ( ) liberals and( 24.1 percent for) conservatives the ( ) actual probability is 18.7 percent . ( In a qualitative version of )this question, Alesina, Stantcheva, and Teso 2018 shows 51.3 percent of ­left-wing respondents, com- that pessimism about social mobility( )is asso- pared with 31.3 percent of ­right-wing respon- ciated with more favorable views toward dents, believe that chances are very low for redistribution, especially in terms of a more children born in the bottom quintile to make it progressive tax system and of more spending to the top. A total of 72 percent of right-wing­ for ­equal-opportunity policies like education respondents versus 38.6 percent of ­left-wing or health. Different perceptions about the same ones agree with the statement that “in the US reality across the political spectrum are thus cor- everybody has a chance to make it and be eco- related with different policy preferences. nomically successful.” To establish causality between perceptions Perceptions of social mobility are especially and support for policies, the authors use an over optimistic in areas where actual social experimental treatment, whereby a randomly mobility( ) is the lowest, namely, the South and selected group of respondents sees pessimistic Southeast of the United States, which are regions information about mobility, highlighting that where the Republican vote is particularly high.3 the chances of children from poor families of rising up the income ladder are small while the chances of children from rich families remaining rich are relatively high. The control group sees 3 The correlation between the perceived probability to no such information. After seeing this pessimis- make it from the bottom to the top quintile and the actual tic information, both left-wing and right-wing­ ­state-level probability is 0.29. respondents become more pessimistic­ about − VOL. 110 THE POLARIZATION OF REALITY 327 mobility, suggesting that the information is understand that top income tax rates were higher indeed convincing. But only ­left-wing respon- in the 1950s­ to 1960s than today. Strikingly, dif- dents become even more supportive of redis- ferent views also extend to one’s own position tribution. Right-wing­ ( ) respondents do not, in society: Stantcheva 2019 shows that con- possibly because, as the authors argue on the ditional on actual income,( )being Republican basis of their detailed survey questions, they increases one’s perceived social class. view the government as the “problem” and not the “solution.” As explained in the framework, IV. Immigration even when faced with the exact same informa- tion about reality, people may translate it into Another issue on which right-wing and political preferences in different ways according ­left-wing respondents have starkly differ- to their other existing perceptions. ent views is immigration. Alesina, Miano, and Stantcheva 2018 investigates how natives III. Inequality and Tax Policy perceive immigrants( in) their country and how this affects their preferences for immigration Perceptions of reality also differ along the policies and redistributive policies, by using political spectrum when it comes to inequality ­custom-designed surveys in the United States and tax policy. Kuziemko et al. 2015 shows that and five European countries France, Italy, 61 percent of Republicans against( )78 percent Germany, Sweden, and the United( Kingdom . of Democrats correctly believe that income The authors ask detailed questions about a wide) inequality in the( United States) has increased in array of immigrants’ characteristics: their share, recent decades see Figure 2 . Actually showing education, unemployment levels, reliance on respondents information( about) the level of and government transfers, and countries of origin.4 change in inequality in the United States has the All respondents starkly overestimate the share unexpected effect of making them trust the gov- of immigrants in the United States and believe ernment less, perhaps because they believe—as on average that it is 36 percent; the actual share in the aforementioned study on social mobil- of legal immigrants is 10 percent of the US ity—that the government may be responsible population around 13.5 percent if we include for the rise in inequality or ineffective at mit- illegal immigrants,( too, and around 26 percent igating it. In line with our framework, percep- if we include ­second-generation immigrants . tions are jointly determined, and what appears at While there is no heterogeneity in the mis per)- first sight to be a signal about only one type of ceived share of immigrants, perceptions( differ) a perception here, inequality can end up shifting great deal when it comes to the socioeconomic other perceptions,( too here,) the competence of and cultural composition of immigrants. Both the government . ( Republican and Democratic respondents overes- Stantcheva )2019 shows that Republicans timate the share of immigrants that are Muslim, believe that the( top 1) percent of earners receive but Republicans’ misperceptions are 5 percent- 40 percent of national income and the top age points higher than Democrats’ 25.2 per- 1 percent of wealth holders hold 53 percent; cent versus 20.7 percent; the true share( is about for Democrats these numbers are 48 percent 10 percent . and 64 percent, respectively. Furthermore, Republicans) believe fewer immigrants have Stantcheva 2019 highlights that political a college degree than do Democrats; they also polarization exists( )even in views of current fac- overestimate the share of immigrants that have tual features of the tax system and in directions not completed high school to a greater extent. that can be expected on the basis of ideology. They perceive more immigrants to be unem- For instance, Democrats believe that 23 percent ployed. The divide is even more significant of households pay no income tax; Republicans when it comes to perceived reliance on the wel- believe this figure is 28 percent. Republicans fare state. Republicans are almost twice as likely perceive the average top income tax rate to be 31 percent; Democrats believe it is 25 per- cent. When it comes to historical perceptions, 4 Perceptions about immigrants are benchmarked against Kuziemko et al. 2015 shows that 47 percent ( ) perceptions about natives, by asking respondents the same of Republicans and 60 percent of Democrats questions about ­nonimmigrants in their country. 328 AEA PAPERS AND PROCEEDINGS MAY 2020 as Democrats to think that an average immigrant whereby the signal information acquisition receives twice as many transfers or more as a itself is endogenous to( existing perceptions) and native resident, and that on net, (“Mohammad”) can thus prevent learning. receives more from the government than “John” who is identical in all respects to Mohammad, except( that he is not an immigrant . The effect REFERENCES of political affiliation on these perceptions) is robust to controlling for the full array of individ- Alesina, Alberto, Armando Miano, and Stefanie ual characteristics e.g., age, education, income, Stantcheva. 2018. “Immigration and Redistri- occupation, being (a second-generation immi- bution.” NBER Working Paper 24733. grant or local factors at the commuting zone Alesina, Alberto, Stefanie Stantcheva, and Edo- level ) e.g., local unemployment and poverty ardo Teso. 2018. “Intergenerational Mobility rates, racial( segregation, share of immigrants or and Preferences for Redistribution.” American minorities . Economic Review 108 2 : 521–54. This heterogeneity) in perceptions does not Bartels, Larry M. 2002. (“Beyond) the Running appear in the answers about nonimmigrants’ Tally: Partisan Bias in Political Perceptions.” characteristics, suggesting that, while respon- Political Behavior 24 2 : 117–50. dents may in general have inaccurate perceptions Bullock, John G., Alan S.( Gerber,) Seth J. Hill, and on many issues, it is mostly on partisan issues, Gregory A. Huber. 2015. “Partisan Bias in Fac- such as immigration, that perceptions diverge tual Beliefs about Politics.” Quarterly Journal across the political spectrum. If respondents are of Political Science 10 4 : 519–78. split according to whether they voted for Trump, Conover, Pamela Johnston,( Stanley) Feldman, and the differences in perceptions are wider than Kathleen Knight. 1987. “The Personal and between Democrats and Republicans overall. Political Underpinnings of Economic Fore- Why do these misperceptions persist? First, casts.” American Journal of Political Science perhaps because this issue is prone to political 31 3 : 559–83. narratives, providing factual information on the Gaines,( )Brian J., James H. Kuklinski, Paul J. actual shares and origins of immigrants—as these Quirk, Buddy Peyton, and Jay Verkuilen. 2007. authors do experimentally—only weakly moves “Same Facts, Different Interpretations: Parti- respondents’ perception of these statistics. On san Motivation and Opinion on Iraq.” Journal the other hand, simply priming respondents in of Politics 69 4 : 957–74. an experimental way to think about immigrants Jerit, Jennifer, and( ) Jason Barabas. 2012. “Parti- before answering questions about redistribution san Perceptual Bias and the Information Envi- reduces their support for redistribution. Second, ronment.” Journal of Politics 74 3 : 672–84. demand for accurate information on this politi- Kuziemko, Ilyana, Michael I. Norton,( ) Emman- cally charged topic seems to be systematically uel Saez, and Stefanie Stantcheva. 2015. “How correlated with political views and with baseline Elastic Are Preferences for Redistribution? misperceptions. When respondents are given the Evidence from Randomized Survey Exper- option to pay a randomized amount to receive iments.” American Economic Review 105 accurate information about the characteristics 4 : 1478–508. of immigrants, respondents who have the most Prior,( ) Markus, Gaurav Sood, and Kabir Khanna. inaccurate which is equivalent here to the most 2015. “You Cannot Be Serious: The Impact negative views( of immigrants are less willing of Accuracy Incentives on Partisan Bias in to pay for) information. In addition, even condi- Reports of Economic Perceptions.” Quarterly tional on the level of misperception and other Journal of Political Science 10 4 : 489–518. individual characteristics, Republican respon- Stantcheva, Stefanie. 2019. “Understanding( ) Eco- dents are 14% less willing to pay to receive nomic Policies: What Do People Know and correct information about immigrants. This How Can They Learn?” Harvard University echoes the phenomenon described in the model, Working Paper.