NBER WORKING PAPER SERIES

THE POLARIZATION OF REALITY

Alberto F. Alesina Armando Miano Stefanie Stantcheva

Working Paper 26675 http://www.nber.org/papers/w26675

NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 January 2020

The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

© 2020 by Alberto F. Alesina, Armando Miano, and Stefanie Stantcheva. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. The Polarization of Reality Alberto F. Alesina, Armando Miano, and Stefanie Stantcheva NBER Working Paper No. 26675 January 2020 JEL No. D71,D72,H41,J15,P16,P35

ABSTRACT

Americans are polarized not only in their views on policy issues and attitudes towards government and society, but also in their perceptions of the same factual reality. We conceptualize how to think about the “polarization of reality” and review recent papers that show that Republicans and Democrats view the same reality through a different lens. Perhaps, as a result, they hold different views about policies and what should be done to address economic and social issues. We also show that providing information leads to different reassessments of reality and different responses along the policy support margin, depending on one's political leaning.

Alberto F. Alesina Stefanie Stantcheva Department of Department of Economics Littauer Center 232 Littauer Center 210 Harvard University Cambridge, MA 02138 Cambridge, MA 02138 and IGIER and NBER and also NBER [email protected] [email protected]

Armando Miano Department of Economics Harvard University Littauer Center Cambridge, MA 02138 [email protected] The Polarization of Reality

By , Armando Miano, and Stefanie Stantcheva∗

Evidence is growing that Americans are Earlier Literature: Some differences polarized not only in their views on policy in the perceptions between Republican issues and attitudes towards government and Democrat voters have been illustrated and society, but also in their perceptions of in the political science literature. Bar- the same factual reality. In this paper, we tels (2002) shows that party identifica- conceptualize how to think about the polar- tion shapes perception of economic indica- ization of reality and review recent papers tors that can be seen as the government’s that show that Republicans and Democrats “performance indicators” (e.g., unemploy- (as well as Trump and non-Trump voters ment or inflation), with Republicans being since 2016) view the same reality through more optimistic than Democrats on eco- a different lens. Perhaps as a result, they nomic variables during the Reagan presi- hold different views about policies and what dency.1 Similar results about the impor- should be done to address different eco- tance of partisan assessment of the govern- nomic and social issues. ment’s performance in shaping perceptions The direction of causality is unclear: On of economic indicators is found in Conover, the one hand, individuals could select into Feldman and Knight (1986) and Conover, political affiliation based on their percep- Feldman and Knight (1987). More recently, tions of reality. On the other hand, polit- Jerit and Barabas (2012) show that people ical affiliation affects the information one perceive the same reality in a way consistent receives, the groups with which one inter- with their political views and that learning acts, and the media to which one is ex- is selective: Partisans have higher knowl- posed, all of which can shape perceptions edge for facts that corroborate their world of reality. Regardless of the direction of views and lower for facts that challenge causality though, this is not about having them. Heterogeneous updating to identical different attitudes about economic or social information is also shown for attitudes to- phenomena or policies that could justifiably wards the Iraq war in Gaines et al. (2007). be viewed differently from different angles. Prior, Sood and Khanna (2015) and Bul- What is striking is rather to have different lock et al. (2015) demonstrate that people perceptions of realities that can be factually give “partisan” answers to factual questions checked. in surveys, but the partisan gap is reduced if monetary incentives are offered for cor- We highlight evidence about differences rect answers. in perceptions across the political spec- trum on social mobility, inequality, immi- I. Conceptual Framework gration, and public policies. We also show that providing information leads to differ- A simple conceptual framework can help ent reassessments of reality and different organize the empirical results reviewed responses along the policy support margin, (see Stantcheva (2019) for the full-fledged depending on one’s political leaning. model). It is illustrated in Figure 1. People can hold many “perceptions,” ∗ Alesina: Harvard, CEPR, IGIER and NBER, which are estimates of true parameters on a 1805 Cambridge Street, Cambridge, MA 02138 (email: variety of topics such as the share of immi- [email protected]); Miano: Harvard, 1805 Cam- bridge Street, Cambridge, MA 02138 (e-mail: ami- [email protected]); Stantcheva: Harvard, CEPR, and 1For instance, in 1988, Democrats were more likely NBER, 1805 Cambridge Street, Cambridge, MA 02138 to report that unemployment and inflation had in- (e-mail: [email protected]). creased since 1980, when the opposite was true. 1 2 PAPERS AND PROCEEDINGS MONTH YEAR grants, the share of national income going indicated again by an arrow in the figure). to the top 1%, or the elasticity of top in- People have to decide which information to comes to top rates. These perceptions incur costs for, which will depend on their have true empirical counterparts. “Policy baseline perceptions. Below, we describe views” are formed as functions of these per- how people with different baseline percep- ceptions and can range from the desired top tions indeed have different willingness to tax rate to the ideal level of government in- pay for information. tervention. Perceptions interact with each What makes learning particularly diffi- other: each policy view can be a function of cult in this setting is that the actual true several or all perceptions. For instance, as values of the variables that people form we will see below, a given perceived level of perceptions about either change over time social mobility will translate into different (e.g., the share of immigrants) or are dif- support for redistribution based on people’s ficult to estimate, even for experts (e.g., perceptions of the competence and trust- the elasticity of unemployment to unem- worthiness of the government.2 ployment benefits). How perceptions are determined depends Finally, note that while behavioral fea- on how learning occurs. People receive “sig- tures could play a role as well, there is no nals” which are pieces of information and need to suppose that if people had the ex- which are weighted in order to be translated act same vector of perceptions, there would into a change in perceptions. Signals do not be disagreement on policy views or updat- have homogeneous impacts on all people’s ing (i.e., willful ignorance or partisan bias perceptions and not all people receive the per se in the shape of the function map- same signals. People may thus end up with ping perceptions to policy views). As long very heterogeneous perceptions and misper- as people have a whole set of heterogeneous ceptions. perceptions, there will be completely differ- First, suppose that information and sig- ent policy views and any signal will be (ra- nals are costless. Even entirely rational up- tionally) acquired and weighted based on dating rules will depend on the prior level the full set of perceptions, thus leading to of (possibly, all) perceptions, as well as on different updating processes too. the weight placed on the signal. The inter- action between perceptions means that the Signals Information weight and updating for an identical signal will depend on all prior perceptions. A sig- (Weighted) nal can move more than one perception at the same time. The weight on the signal Perceptions is endogenous to perceptions (as also indi- cated by an arrow in the figure) because it could depend on its assessed reliability P1  P2 . ..PN of which is yet another perception held by people. Second, imagine information is costly to Policy views acquire. Then, in addition to the interac- Figure 1. : From Information to Policy tion between existing perceptions and up- Views dating just described, the set of signals ac- quired is also endogenous to perceptions (as II. The American Dream 2In addition, some perceptions may put consistency constraints on others. For instance, one cannot simulta- Is the American dream alive? The an- neously believe that all immigrants are unemployed, yet swer people give to this question turns out your sector’s jobs are mostly taken by immigrants. This could be called “Schroedinger’s” immigrant and we do to be a key determinant of support for re- not need to rely on such mental models to rationalize distribution. If perceived social or inter- the results presented. generational mobility is high, the resulting VOL. VOL NO. ISSUE THE POLARIZATION OF REALITY 3 inequalities in income and wealth are per- the middle class are very different across the ceived as more fair, since it is thought that political spectrum, as illustrated in the top everyone had more equal opportunities with panel of Figure 2. The perceived probabil- which to start. ity that a child born in the bottom quintile remains in that quintile as an adult is 37.4% for left-wing respondents and 29.5% for right-wing respondents (the correct num- ber is 33.1%). The perceived probability such a child makes it into the middle class (the third quintile) is 19.3% for liberals and 24.1% for conservatives (the actual proba- bility is 18.7%). In a qualitative version of this question, 51.3% of left-wing respondents, as com- pared to 31.3% of right-wing respondents, believe that chances are very low for chil- dren born in the bottom quintile to make it to the top. 72% percent of right-wing respondents versus 38.6% of left-wing ones Figure 2. : Differences in Perceptions across agree with the statement that “In the U.S. the Political Spectrum everybody has a chance to make it and be economically successful.” Notes: The lines are 95% confidence intervals around Perceptions of social mobility are espe- the mean. cially (over)optimistic in areas where actual Source: Top panel: Alesina, Stantcheva and Teso social mobility is the lowest, namely the (2018). Bottom panel: Kuziemko et al. (2015). South and Southeast of the U.S., which are Using detailed surveys on several thou- regions where the Republican vote is par- sands respondents, Alesina, Stantcheva and ticularly high.3 Teso (2018) investigate the perceptions Alesina, Stantcheva and Teso (2018) about intergenerational mobility in the U.S. show that pessimism about social mobility and Europe. They show that Americans are is associated with more favorable views to- not only more optimistic about social mo- wards redistribution, especially in terms of bility than Europeans, but that they are more progressive tax system and of more overly optimistic given reality: They over- spending for equal opportunity policies like estimate the chances of making it from the education or health. Different perceptions bottom to the top, i.e., for a child born in about the same reality across the political the bottom quintile to make it to the top spectrum are thus correlated with different quintile. policy preferences. Perceptions of social mobility are cor- To establish causality between percep- related with political orientation. Ameri- tions and support for policies, the authors cans who identify themselves as conserva- use an experimental treatment, whereby tive (i.e., right-wing) on economic issues be- a randomly-selected group of respondents lieve that the probability that a child born sees pessimistic information about mobil- in the bottom quintile makes it to the top ity, highlighting that the chances of chil- is 12%; liberals (left-wing) respondents be- dren from poor families of rising up the in- lieve that it is 10.5%. Thus, both groups come ladder are small, while the chances of overestimate the correct answer, which is children from rich families remaining rich 7.8%, implying that belief in the strongest are relatively high. The control group sees form of the “American dream” – making it from rags to riches – is still relatively 3The correlation between the perceived probability prevalent. However, perceptions about the to make it from the bottom to the top quintile and the chances of making it out of poverty and into actual state-level probability is -0.29. 4 PAPERS AND PROCEEDINGS MONTH YEAR no such information. After seeing this pes- is 28%. Republicans perceive the average simistic information, both left and right- top income tax rate to be 31%, Democrats wing respondents become more pessimistic believe it is 25%. When it comes to his- about mobility, suggesting that the infor- torical perceptions, Kuziemko et al. (2015) mation is indeed convincing. But only show that 47% of Republicans and 60% of left-wing respondents become (even) more Democrats understand that top income tax supportive of redistribution. Right-wing rates were higher in the 1950s-60s than to- respondents do not, possibly because, as day. Strikingly, different views also extend the authors argue based on their detailed to one’s own position in society: Stantcheva survey questions, they view the govern- (2019) shows that conditional on actual in- ment as the “problem” and not the “solu- come, being Republican increases one’s per- tion.” As explained in the framework, even ceived social class. when faced with the exact same informa- tion about reality, people may translate it IV. into political preferences in different ways based on their other existing perceptions. Another issue on which right and left- wing respondents have starkly different III. Inequality and views is immigration. Alesina, Miano and Stantcheva (2018) investigate how natives Perceptions of reality also differ along perceive immigrants in their country and the political spectrum when it comes to how this affects their preferences for immi- inequality and tax policy. Kuziemko gration policies and redistributive policies, et al. (2015) show that 61% of Republicans using custom-designed surveys in the U.S. against 78% of Democrats (correctly) be- and five European countries (, Italy, lieve that income inequality in the U.S. has Germany, Sweden, and the U.K.). They increased in recent decades (see Figure 2). ask detailed questions about a wide array Actually showing respondents information of immigrants’ characteristics: their share, about the level and change in inequality in their education, unemployment levels, re- liance on government transfers, and coun- the U.S. has the unexpected effect of mak- 4 ing them trust the government less, per- tries of origin. haps because they believe – as in the afore- All respondents starkly overestimate the mentioned study on social mobility – that share of immigrants in the U.S. and be- the government may be responsible for the lieve on average that it is 36%; the ac- rise in inequality or ineffective at mitigating tual share of legal immigrants is 10% it. In line with our framework, perceptions of the U.S. population (around 13.5% if are jointly determined and what appears at we included illegal immigrants too and first sight to be a signal about one type of about 26% including second-generation im- perception only (here, inequality), can end migrants). While there is no heterogene- up shifting other perceptions too (here, the ity in the (mis)perceived share of immi- competence of the government). grants, perceptions differ a great deal when Stantcheva (2019) shows that Republi- it comes to the socio-economic and cultural cans believe that the top 1% of earners re- composition of immigrants. Both Repub- ceive 40% of national income and the top lican and Democrat respondents overesti- 1% wealth holders hold 53%; for Democrats mate the share of immigrants that are Mus- these numbers are 48% and 64% respec- lim, but Republicans’ misperceptions are 5 tively. Furthermore, Stantcheva (2019) percentage points higher than Democrats’ highlights that political polarization exists ones (25.2% vs. 20.7%, the true share is even in views of current factual features of about 10%). the tax system and in directions that can be 4Perceptions about immigrants are benchmarked expected based on ideology. For instance, against perceptions about natives, by asking respon- Democrats believe that 23% of households dents the same questions about non-immigrants in their pay no income tax; Republicans believe it country. VOL. VOL NO. ISSUE THE POLARIZATION OF REALITY 5

Republicans believe fewer immigrants perceptions. When respondents are given have a college degree than do Democrats; the option to pay a randomized amount in they also overestimate the share of immi- order to receive the accurate information grants that have not completed high-school about the characteristics of immigrants, re- to a greater extent. They perceive more spondents who have the most inaccurate immigrants to be unemployed. The di- (which is equivalent here to the most nega- vide is even more significant when it comes tive) views of immigrants are less willing to to perceived reliance on the welfare state. pay for information. In addition, even con- Republicans are almost twice as likely as ditional on the level of misperception and Democrats to think that an average immi- other individual characteristics, Republican grant receives twice as many transfers (or respondents are 14% less willing to pay more) as a native resident, and that on net, to receive correct information about immi- “Mohammad” receives more from the gov- grants. This echoes the phenomenon de- ernment than “John” (who is identical in scribed in the model, whereby the signal all respects to Mohammad, except that he (information) acquisition itself is endoge- is not an immigrant). The effect of the po- nous to existing perceptions and can thus litical affiliation on these perceptions is ro- prevent learning. bust to controlling for the full array of indi- vidual characteristics (e.g., age, education, REFERENCES income, occupation, being a second gener- ation immigrant, etc.) or local factors at Alesina, Alberto, Armando Miano, the commuting zone level (such as the lo- and Stefanie Stantcheva. 2018. “Im- cal unemployment and poverty rates, racial migration and Redistribution.” NBER segregation, share of immigrants or minori- Working Paper No. 24733. ties, etc.). Alesina, Alberto, Stefanie This heterogeneity in perceptions does Stantcheva, and Edoardo Teso. not appear in the answers about non- 2018. “Intergenerational Mobility and immigrants’ characteristics, suggesting Preferences for Redistribution.” Ameri- that, while respondents may in general can Economic Review, 108(2): 521–554. have inaccurate perceptions on many issues, it is mostly on partisan issues Bartels, Larry M. 2002. “Beyond the such as immigration, that perceptions Running Tally: Partisan Bias in Po- diverge across the political spectrum. If litical Perceptions.” Political behavior, respondents are split according to whether 24(2): 117–150. they voted for Trump, the differences in perceptions are wider than between Bullock, John G., Alan S. Gerber, Democrats and Republicans overall. Seth J. Hill, and Gregory A. Hu- Why do these misperceptions persist? ber. 2015. “Partisan Bias in Factual Be- First, perhaps because this issue is prone liefs about Politics.” Quarterly Journal of to political narratives, providing factual in- Political Science, 10(4): 519–578. formation on the actual shares and origins Conover, Pamela Johnston, Stan- of immigrants – as these authors do exper- ley Feldman, and Kathleen Knight. imentally – only weakly moves their per- 1986. “Judging Inflation and Unemploy- ception of these statistics. On the other ment: The Origins of Retrospective hand, simply priming respondents in an ex- Evaluations.” The Journal of Politics, perimental way to think about immigrants 48(3): 565–588. before answering questions about redistri- bution reduces their support for redistri- Conover, Pamela Johnston, Stan- bution. Second, demand for accurate in- ley Feldman, and Kathleen Knight. formation on this politically-charged topic 1987. “The Personal and Political Under- seems to be systematically correlated with pinnings of Economic Forecasts.” Ameri- political views and with the baseline mis- can Journal of Political Science, 559–583. 6 PAPERS AND PROCEEDINGS MONTH YEAR

Gaines, Brian J., James H. Kuklin- ski, Paul J. Quirk, Buddy Peyton, and Jay Verkuilen. 2007. “Same Facts, Different Interpretations: Partisan Moti- vation and Opinion on Iraq.” Journal of Politics, 69(4): 957–974. Jerit, Jennifer, and Jason Barabas. 2012. “Partisan Perceptual Bias and the Information Environment.” The Journal of Politics, 74(3): 672–684. Kuziemko, Ilyana, Michael I Nor- ton, , and Ste- fanie Stantcheva. 2015. “How Elas- tic are Preferences for Redistribution? Evidence from Randomized Survey Ex- periments.” American Economic Review, 105(4): 1478–1508. Prior, Markus, Gaurav Sood, and Kabir Khanna. 2015. “You Cannot be Serious: The Impact of Accuracy Incen- tives on Partisan Bias in Reports of Eco- nomic Perceptions.” Quarterly Journal of Political Science, 10(4): 489–518. Stantcheva, Stefanie. 2019. “Under- standing Economic Policies: What do People Know and How Can they Learn?” Harvard University Working Paper.