An open access publication of the American Academy of Arts & Sciences
Fall 2022

Fifty Years of Declining Confidence & Increasing Polarization in Trust in American Institutions

Henry E. Brady and Thomas B. Kent
View PDF

Except for the military and science, confidence in most American political and nonpolitical institutions has fallen precipitously over the past fifty years. Declines in trust are partly the result of dissatisfaction with governmental and institutional accountability and concomitant skepticism about the competency and responsiveness of institutions. Declines are also the result of a polarization in trust in institutions, as Republicans trust business, the police, religion, and the military much more than Democrats, whose confidence in these institutions, except the military, has fallen. In turn, Democrats trust labor, the press, science, higher education, and public schools much more than Republicans, whose confidence in these institutions has fallen. Declines and polarization in confidence may be traceable to political polarization stemming from increasing income inequality and segregation in America. With polarization and decreasing trust in institutions, it becomes more difficult to fight epidemics, maintain faith in policing, and deal with problems such as climate change.

Henry E. Brady, a Fellow of the American Academy since 2003, is the Class of 1941 Monroe Deutsch Professor of Political Science and Public Policy at the University of California, Berkeley. He served as Dean of the Goldman School of Public Policy from 2009–2021. He is the author of Letting the People Decide: Dynamics of a Canadian Election (with Richard Johnston, André Blais, and Jean Crête, 1992), Voice and Equality: Civic Voluntarism in American Politics (with Sidney Verba and Kay Lehman Schlozman, 1995), and, most recently, Unequal and Unrepresented: Political Inequality and the New Gilded Age (with Kay Lehman Schlozman and Sidney Verba, 2018).

Thomas B. Kent is a doctoral candidate in the Department of Political Science at the University of California, Berkeley.

Everyday life depends upon confidence in institutions. We trust the military and police to protect us, businesses to deliver safe products at reasonable prices, educational institutions to instruct our children, the media to transmit truthful and useful information, doctors and lawyers to cure and defend us, and government to act in our best interest. But confidence in these institutions has declined and become politically polarized in the past fifty years. How and why has this happened? What does it mean for America when trust declines and some people trust institutions more than others, especially when the split in trust is across party lines?

Every year or two since the early 1970s, the Gallup Poll, NORC’s General Social Survey (GSS), and the Harris Poll have asked a series of questions about the public’s confidence in the “people in charge of running institutions” or in these “institutions” themselves. Each survey typically asks about fifteen or so institutions, although not all the same ones. Over the entire period we have continuous data on four political institutions and sixteen nonpolitical ones.1

The four political institutions are the presidency, the executive branch, Congress, and the Supreme Court of the United States.2  Trust in these governing institutions has declined dramatically in the past fifty years (see Figure 1). On a scale with answers of:

(3) “a great deal of confidence,”
(2) “quite a lot of confidence,”
(1) “some confidence,”
(0) “hardly any confidence at all,”

confidence in the presidency has gone from usually nearer “quite a lot” in the 1970s and 1980s (except for just after the Watergate scandal around 1972 to 1974) to just “some” in the last decade.3 Between the period of 1972 to 1979 and the period of 2010 to 2021, confidence in Congress declined by 45 percent. Confidence in the presidency and in the executive branch declined by about 20 percent, and confidence in the Supreme Court declined by 12 percent.4 These changes mirror the drop in trust in “government” of about 40 percent found over the same period on another set of surveys, the American National Election Studies.5

The sixteen nonpolitical institutions cover all major segments of society.6 Four are associated with the economy: business, banks, Wall Street, and organized labor (see Figure 2). Seven are related to knowledge and information production: the press, TV news, television, public schools, education, higher education, and science (see Figure 3). Three enforce norms and standards: the police, the military, and religion. Two deliver professional services: medicine and law (see Figure 4 for these last five).7 Historically, these have all been considered neutral, non­political institutions for the provision of goods and services

The decline in confidence in these nonpolitical institutions is less well known than the drop in trust in government. In Figures 2–4, confidence falls relatively steadily for almost all institutions, with some ups and downs superimposed. If we compare average confidence in each of these nonpolitical institutions from 1972 to 1979 with average confidence from 2010 to 2021, confidence has declined in fourteen of the institutions, stayed the same for one (science), and increased only for the military. Most of these declines occurred relatively steadily over time, with the largest ones—comparable to the drops for Congress and government as a whole—occurring for Wall Street, TV news, banks, and the press. More moderate declines—comparable to those for the presidency and executive branch—occurred for public schools, medicine, television, business, and religion. Smaller reductions—roughly comparable to those for the Supreme Court—occurred for law, education, and the police. Still smaller decreases occurred for higher education and labor. Figure 5 summarizes these changes in the Gallup-GSS-Harris data, with the four political institutions in the darker bars and the sixteen nonpolitical institutions in the lighter bars.

These declines have been significant, and nonpolitical institutions have gone from being trusted quite a lot to being trusted only somewhat. The American people expressed “quite a lot” of confidence in 1972–1979 in thirteen of the sixteen nonpolitical institutions, and they expressed merely “some” confidence in only three of them (labor, law, and television) back then. By 20102021, only six institutions—education, higher education, medicine, the military, science, and police—still enjoyed “quite a lot” of confidence, and ten institutions warranted just “some” confidence. Recent data suggest that Americans probably have only “some” confidence in higher education as well because the time-series for higher education from the Harris Poll used in Figures 3 and 5 ends at 2012, and data from the Pew Research Center show that trust in higher education has fallen significantly since 2012.8 So Americans have gone from believing that thirteen of sixteen nonpolitical institutions deserved quite a lot of confidence to believing that only five of sixteen merit quite a lot of confidence, and that eleven deserve just some confidence.

Confidence in science is about the same in the 2010–2021 period as it was in the 1972–1979 period, and confidence in the military has increased by about 21 percent, according to the Gallup-GSS-Harris data. Confidence in labor has only gone down by 4 percent, but labor was the least trusted institution (political or nonpolitical) in the 1972–1979 period with only “some” confidence expressed in it, so it did not have much room to fall. Its decline by 4 percent is significant given these circumstances. Medicine and science were the number one and number two most trusted of all institutions in 1972–1979 with “quite a lot” of confidence expressed in them, so the drop in confidence in medicine (becoming the fifth-most trusted institution) is not surprising, but the stability of confidence in science is surprising. As we shall see, this steadiness masks some interesting partisan dynamics. The increase in confidence in the military is real, but feeling thermometers from the American National Election Studies that ask about “warmth toward an institution” starting in 1964 suggest that confidence in the military declined through the rest of the 1960s as a result of the Vietnam War and that some of the gain from the 1970s is a return to 1964 levels of confidence, although the rest of the gain reaches still higher levels of confidence than even the early 1960s.

How do we know that confidence questions are capturing something real? The evidence is clearest for the military: confidence in the military closely tracks the ups and downs of national security events. Figure 6 plots confidence in the military from 1972 onward for three different survey houses: Gallup, GSS, and Harris. The similarities in the ups and downs across the three survey organizations indicate that they are measuring comparable attitudes, and the pattern over time signals that confidence in the military responds to actual events. The low values for confidence on the left correspond to the end of the Vietnam War period when confidence in the military was low. The success of the 1990–1991 Gulf War led to increased confidence in the military. The peak right after September 11, 2001, suggests a rally-around-the-flag effect. High casualties, insurgency, and civil war in Iraq then led to declines in confidence in the military, while the January 2007 “surge” quickly inspired more confidence in the military, as American casualties fell and ethnosectarian violence in Iraq decreased.9 Despite these effects, it is surprising that the military retained the confidence of Americans through the failed war in Afghanistan, finally ended by Joseph Biden in the summer of 2021, although the most recent survey evidence suggests a decline in confidence as indicated by the drop in the 2020–2021 average for confidence in the military in Figure 6.

Just as confidence in the military is affected by national security events, confidence in Wall Street is affected by the major ups and downs of the stock market, such as the peaks from the dot-com bubble of 2000 and the housing mortgage securities–fueled 2007 run-up, and subsequent downturns in the recession of 2001 and the Great Recession of 2008–2009. Confidence in banks was greatly affected by the savings and loans failures of 1988 to 1992 and the bank failures of the Great Recession of 2008–2009. There are local high points in confidence in 1988 and 2006 followed by local valleys in 1991 and 2009–2011, with half-point or more drops on the zero-to-three scale.10 There is also a local peak in confidence in banks in 1977 followed by a precipitous drop until 1981 that cannot be explained by bank failures. Instead, mortgage rates began to increase from about 8 percent in 1977, where they had been for a while, to the peak of the last fifty years of over 16 percent in 1981, as the Federal Reserve tightened the money supply to fight inflation. It is very likely that this drop in confidence reflects concerns about high mortgage rates charged by banks during this period.

Although confidence in other institutions is not so clearly affected by episodic major events (see the relatively smooth declines in Figures 24 for most institutions), other factors, including the accumulations of events and experiences, have substantial impacts on trust. Those who attend religious services nearly every week or more have “quite a lot” of confidence in religion whereas those who never attend trust religion only “some.” African Americans are about 25 percent less confident in the police than White Americans, presumably because of their ongoing negative experiences with the police. Members of union households are 25 percent more confident in labor. Those in the highest quintile of education are 20 percent more trusting of science than those in the lowest quintile. Those in the highest income quintile have 15 percent more confidence in business and Wall Street than those in the lowest income quintile. Young people are about 10 percent more likely to have confidence in higher education than older people. In all these cases, individual life experiences, indexed by socioeconomic and demographic characteristics, affect confidence in these institutions. In addition, the GDP growth rate is positively associated with about 30 percent of the variance over time in confidence in Wall Street, 11 percent in business confidence, and 11 percent in confidence in banks, but it is not, just as we would expect, associated with confidence in any other nonpolitical institution. The unemployment rate is negatively associated with about 24 percent of the variance in trust in labor, but there is no significant correlation of confidence in labor with growth rate.11 These relationships provide evidence that confidence is measuring something meaningful to people.

In addition to significant declines in confidence, there have been substantial increases in partisan polarization in confidence in which the partisans of one party have more confidence in an institution than the partisans of the other party, merely because of perceptions about which party controls the institution. Among the four political institutions, polarization depends upon which party has control of the institution at a given moment. This moment is easiest to define for the presidency and the executive branch, for which periodic elections determine their partisanship. Figure 7 displays trust in the president by partisan group from 1973 to the present.12 There are several interesting features. First, the most trusting are always the partisans of the president’s party. When there is a Republican president, then Republicans are the most confident in the president, and when there is a Democratic president, then Democrats are most confident. And the confidence of the partisans of the incumbent president’s party depends most of all on their partisanship: it remains more or less constant over time at “quite a lot,” with some variation, such as the peak confidence in 2001 for Republicans and even for Democrats—for a president of the other party—due to Americans rallying around the flag after 9/11. Second, the lowest level of trust is for the out-party partisans, those of the opposite party from the president, and it has fallen dramatically over time from midway between “some” and “quite a lot” to closer to “hardly any at all.” Republicans did not have even some confidence in Obama and Democrats had hardly any confidence in Trump. Third, trust among independents has gone down over time. This decline in confidence among out-party partisans and independents has caused trust in the presidency to decline overall.

Confidence in the executive branch behaves similarly. The stories for Congress and the Supreme Court are more complicated given the difficulties of identifying their partisanship, but they reveal some of the same forces at work.

Among nonpolitical institutions, only business and labor showed significant polarization in the 1970s, with Republicans trusting business about 21 percent more than Democrats and Democrats trusting labor about 28 percent more than Republicans.13 By the 2010s, assessments of every institution except banks were more polarized than in the 1970s. These changes in polarization are summarized in Figure 8 for the 1972–1979 period and in Figure 9 for the 2010–2021 period. The graphs plot the average level of confidence in sixteen institutions during the time period for Democrats against the average level of confidence for Republicans.

For the 1972–1979 period depicted in Figure 8, some institutions are trusted more than others, appearing farther to the right and toward the top of the graph. The diagonal “neutral” solid line indicates where the institutions would be located if Democrats and Republicans had the same average confidence in them. To get a sense of how much difference there is in confidence between the parties, we add two other dashed lines: one 0.25 units above the neutral line and another 0.25 units below that line. We chose this number (somewhat arbitrarily) because it is about the same as the decline in confidence in the presidency from the 1970s to the present, somewhat less than half the decline in confidence in the U.S. Congress, and one-twelfth of the entire zero-to-three confidence scale.

As indicated above, only business and labor fall significantly from the diagonal line of neutrality in the 1972–1979 period. Republicans trusted business much more than Democrats and Democrats trusted labor much more than Republicans. In addition, several other institutions are on or near one of the dashed lines: TV news and the press were slightly more trusted by Democrats than by Republicans. On the other side, the police and banks were slightly more trusted by Republicans than by Democrats. While there was some polarization, what is most remarkable about this picture is that there was very little partisanship with regard to trust in most major institutions.

For the 2010–2021 period depicted in Figure 9 using the same scale, many institutions have moved to the bottom-left of the graph, indicating a loss of confidence, and almost all of them have moved outward from the solid line and beyond the dashed lines, indicating polarization. In the current period, there is now polarization in trust across almost all institutions that is comparable to or more than the polarization in partisan trust of business and labor in the 1970s. Democrats trust the knowledge- and information-producing institutions and organized labor more than Republicans: 54 percent more for TV news, 46 percent more for the press, 44 percent more for labor, 37 percent more for television, 28 percent more for public schools, 19 percent more for higher education, 16 percent more for education, and 14 percent more for science.14 Republicans trust the norm-enforcing institutions and business more than Democrats: 38 percent more for police, 30 percent more for religion, 27 percent more for business, 25 percent more for Wall Street, and 13 percent more for the military. Only banks and medicine are clearly within the dashed lines. A comparison of Figures 8 and 9 also reveals that law, public schools, science, and medicine have moved from being more trusted by Republicans than Democrats in the earlier period to being now more trusted by Democrats than Republicans. Finally, an analysis that compares the 2010–2015 period to the 2016–2021 period reveals that polarization is continuing with more polarization in confidence for twelve of the fifteen nonpolitical institutions for which we have data, especially for the police, the press, higher education, television, science, and TV news, whose differences in partisan evaluations of confidence nearly or more than doubled over that short period of time.15

For twelve of the thirteen nonpolitical institutions (excepting banks) in which trust has fallen, the decline in overall confidence is partly explained by this polarization, with confidence among partisans of the currently less-trusting political party dropping especially precipitously, while the confidence of the other, more trusting, political party either declining only a bit or even increasing somewhat.

For the two cases in which confidence has changed slightly or not at all, labor and science, there are two different stories. After a dip in confidence through the Reagan years (perhaps precipitated by President Reagan’s antilabor policies, including his breaking of the Air Traffic Controllers’ Strike of 1981), the confidence of Democrats and independents in labor has increased somewhat over the past thirty-five years while Republicans’ confidence has stayed the same, except for a drop and then rebound during the Obama years. Republicans were more favorable toward science until the mid-2000s, when a switch occurred and Democrats became much more favorable than Republicans.

For banks, partisan differences have remained about the same, and the major cause of the overall decline in confidence appears to be, as noted earlier, episodes of bank failures, whose effects on confidence then persist. And for the one case in which trust has increased—the military—the biggest factor has been the substantial increase in confidence among the partisans of the more trusting party, in this case, the Republicans, although trust in the military among Democrats has gone up as well.

Confidence among political independents is always either lower than that of both Democrats and Republicans or between the levels of those partisans, and trust has changed among independents in almost the same way as in the entire population (see Figure 10). This graph plots independents’ confidence in each nonpolitical institution in 20102021 against their confidence in 19721979. For institutions above the solid diagonal line, trust has increased over time; for those below, it has decreased. Trust in the military increased among independents between 19721979 and 20102021, but declined for every other institution, with especially large drops for Wall Street, TV news, banks, the press, public schools, and medicine. These institutions are all below the dashed line in Figure 10 with changes of 0.40 or more. These six nonpolitical institutions also had the largest overall declines in Figure 5.

These data reveal distinct and complementary patterns of change for non­political institutions. In some cases, changing confidence in a particular institution can be linked to a large-scale event with society-wide consequences; for example, across individuals and groups, a war can affect confidence in the military (see Figure 6), or a financial crisis can diminish confidence in banks and Wall Street. In other cases, individual life experiences might have implications for confidence in a particular institution: for example, being the victim of police harassment or the victim of a crime might influence trust in the police. We have already cited evidence for these kinds of events and life experiences affecting confidence for various nonpolitical institutions.

In a quite different pattern, a set of general nonpartisan forces—affecting independents especially strongly—produces an overall decline in trust in almost all nonpolitical institutions (see Figure 10). Although different groups, including different party groups, vary in their initial levels of confidence in various nonpolitical institutions, such forces operate more or less uniformly across groups to diminish confidence in all institutions, including political ones.

In still another distinct pattern, there is a partisan interaction. Some factors lead to a decline in trust among members of one party or the other, depending upon the type of institution, resulting in polarization in confidence (see Figures 8 and 9). The effects are related to the kind of institution, with trust falling for the knowledge- and information-producing institutions for Republicans and for the norm-enforcing institutions for Democrats.

The forces at work probably interact in complicated ways. To identify what is going on, we must consider the events and life experiences that affect trust, and the multiple forces that have led to a secular decline in trust and those that have led to partisan polarization in trust.

Getting at these explanatory factors requires understanding what institutions need to do to elicit trust. Being viewed as legitimate both by their stakeholders and by the public at large provides the foundation for trust. Legitimacy underlies confidence. As noted in the introduction to this issue of Dædalus, legitimacy can come from four basic sources, and different institutions rely on different mixes of sources of legitimation.16 Legitimacy can stem from the political system sharing its regulatory authority with an institution—such as the military, police, or a corporation—based upon government’s power of coercion to defend the nation, keep the peace, and enforce contracts. As long as the institution conforms to the rules established by the government, it draws legitimacy from its relationship to the government in the form of laws or charters. Legitimacy may also come from adherence to culturally approved and accepted meanings and logics, as with the practice of medicine, religion, education, and science that are shaped by what is culturally appropriate for each institution. It may reside in moral and normative beliefs about how those in institutions behave, as with professional codes of ethics for law, medicine, higher education, and journalism. Finally, it may come from pragmatic authority based on efficiency and high performance, as with corporations, science, or banks.

To be successful, an institution must be seen as legitimate in at least one and often in all four ways. If an institution is legitimate, then it is usually seen as trustworthy as well. It will be trusted by individuals, and people will accept the institution’s advice, services, and decisions. They will have confidence in it.

The four sources of legitimacy are places to look to understand the various patterns in the decline and polarization of trust. It is obvious that events and experiences can affect legitimacy and confidence: legitimacy declines when institutions defy regulatory authority, fail to adhere to culturally approved logics, violate moral and normative beliefs, or simply do not perform. For example, bank failures indicate insufficient performance and an inability to meet regulatory requirements. Corruption in institutions such as religious organizations violates moral and normative beliefs and reduces people’s confidence in those institutions. Rising tuition for higher education suggests a lack of performance and erodes confidence. Legitimacy and confidence are enhanced when events and experiences conform with and reinforce regulatory, cultural, normative, or pragmatic legitimacy. Science gains legitimacy when it uses culturally accepted logics, such as peer review; business corporations gain legitimacy when economic growth is high. People are more confident in labor when unemployment is low. Identifying these events and experiences for each institution can help to explain movements in confidence, but they do not seem to be enough to explain the secular decline in trust and the increasing polarization of trust.

What broad nonpartisan forces could lead to an overall loss of confidence and how could partisan forces lead to polarization in confidence in institutions? These forces might be related to one another, but it is useful to start by looking for nonpartisan forces that undermine legitimacy and hence social cohesion, the sense of trust among people and between people and institutions.17 Other essays in this volume suggest that we live in a skeptical age replete with journalists and pundits constantly questioning authority and brimming with more and more college-educated people trained to doubt and to question authority. Rising skepticism might account for the general trend toward declining confidence in institutions, but we are far from knowing whether this explains the large declines in trust, especially since it suggests that people were overly credulous of institutions in the past.

Declining confidence in institutions is also associated with a diminution in political efficacy, an increase in political alienation, and declining trust in other people when asked: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in life?” Social trust gauged by this question has decreased over the past fifty years, and it is generally correlated with confidence in institutions.18 Taken together, these results reveal additional symptoms of the overall syndrome of splintering social cohesion, but they do not explain its roots unless we can explain falling efficacy, increasing alienation, and diminished social trust.

The overall erosion of trust across multiple institutions suggests that we should look more widely for major social trends that might undermine trust among all groups and increase alienation. Two possible causes that might affect the social fabric are fifty years of mounting inequality and increasing diversity in the United States through immigration and differential birth rates.19

Economic inequality has increased in the United States over the past fifty years, with the top 1 percent’s wealth going from 25 percent to almost 40 percent of all wealth and their income jumping from 10 percent to 20 percent of all earnings. This widening inequality followed an earlier period of growing income for everyone:

From 1946 to 1980, growth [in income] was evenly distributed with all income groups growing at the average 2 percent annual rate (except the top 1 percent which grew slower). From 1980 to 2018, growth has been unevenly distributed with low growth for bottom income groups, mediocre growth for the middle class, and explosive growth at the top.20

Diversity increased dramatically from 1970 as the United States shifted from being over 83 percent non-Hispanic White, with only 11 percent African American, 5 percent Hispanic, and less than 1 percent Asian American, to being 58 percent non-Hispanic White, with 19 percent Hispanic, 12 percent African American, 6 percent Asian American, and 5 percent other in 2020.21

There is no research on how either of these trends affects confidence in institutions, but the increase in economic inequality in America has been implicated in the decline of social trust between people, which, in turn, is related to other forms of trust.22 There has also been a great deal of research on how increases in ethnic diversity where people live, coupled with inequality and segregation of neighborhoods, reduce social trust, notably in the United States.23 Ethnic, economic, and residential divides produce a sense of anomie and isolation that decreases social cohesion and social trust. Perhaps the same forces are undermining confidence in almost all political and nonpolitical institutions. The nonpartisan nature of these forces is suggested by the fact that while social trust has declined over the past fifty years for all groups, it has declined faster for political independents, who do not have the baggage of partisan affiliation, than for partisans.24

Partisan polarization in trust must have additional causes that have politicized formerly nonpolitical institutions. Polarization in presidential confidence is based upon the long history of partisan battles between Democrats and Republicans and the identification of presidents with political parties. Polarization in confidence for business and labor already existed in the 1970s when the New Deal economic cleavage still dominated American politics, and it was based upon a long history of conflict between business and labor in America culminating in the National Labor Relations Act (Wagner Act) of 1935 and the Taft-Hartley Act of 1947. In addition, business had a long-standing identification with the Republican Party that was cemented with the elections in 1920, 1924, and 1928 of three probusiness Republican presidents. Labor became strongly associated with the Democratic Party because of the National Labor Relations Act and the New Deal.

The lack of confidence shown by Democrats in business or by Republicans in labor was at least partly rooted in the calculation that business was not trustworthy for Democrats and labor was not trustworthy for Republicans because of the different bases of legitimation for labor and business. That suggests that we should look for ways since the 1970s that the nonbusiness nonpolitical institutions have appeared to be less trustworthy for partisans of one side or another because they did not meet that partisan group’s standards for regulatory, cultural, normative, or pragmatic legitimacy.

The growth of partisan polarization in trust in nonbusiness nonpolitical institutions tracks the partisan emergence since the 1970s of hot-button social and cultural issues—including, for example, civil rights, abortion, immigration, prayer in school, gay rights, and gun rights—that had long been divisive yet not aligned with partisanship.25 It is tempting to believe that the emergence of this new cultural and social dimension of politics underlies this new form of polarization in institutions.26

Before exploring this possibility, we should ask whether these nonpolitical institutions have actually become associated with particular sides of political debates. In a national survey that we completed in September 2019, a representative group of Americans was asked about their perceptions of the partisan and ideological complexion of a subset of institutions. We found that highly religious people, police, bankers, and military generals are seen as typically conservatives and Republicans, and college professors, journalists, labor union members, public school teachers, and scientists are seen as liberals and Democrats. Only doctors and lawyers were seen, on average, as neither Republicans nor Democrats. Most institutions appear to be politicized.

Unfortunately, these questions have not been asked in the past, so it is hard to know whether these perceptions are new or long-standing. Based upon scattered results from similar survey questions in the past, however, we believe that they are new. In addition, Kent has found evidence that at least some of the perceptions may be right. Since 1980, some professions and semiprofessions have become more partisan in their political contributions in just the ways found on the 2019 survey.27

That still leaves open the question about whether these identifications explain the polarization in trust. Perhaps people make these identifications, but they do not affect their confidence judgments. Perhaps other factors related to legitimacy, such as the importance, cost, or competence of an institution, matter more to or are assessed differently by Republicans and Democrats, and these differences account for polarization of trust. Democrats and Republicans may just have different opinions about the legitimacy of these institutions. To test this possibility, we also asked our national sample whether various institutions do important work that matters, whether they cost too much, and whether they do good work that is competent.

Americans differ in their beliefs about whether institutions do important work. Republicans believe that the police, the military, and religion do important work, but Democrats are less sure. Democrats believe that labor, the press, and higher education do important work, but Republicans are less sure. We found that Republicans think that labor, the press, science, and higher education cost too much and Democrats are less inclined to believe that. Democrats think that the military, the police, and religion cost too much and Republicans do not. Republicans think that the military, the police, and religion do good, competent work, but Democrats are less sure. Democrats think that labor, the press, science, and higher education do good work and Republicans are less convinced. Despite these significant partisan differences regarding the importance, cost, and competency of these institutions, one of the biggest predictors of a respondent’s confidence in an institution is their perception of its partisan makeup.

Our survey results show that nonpolitical institutions have become identified with the parties and that these identifications are associated with the polarization of confidence, but they do not provide any insight into the exact way that previously nonpolitical institutions became politicized. The rise of a social/cultural dimension of American politics in addition to the preexisting New Deal economic cleavage suggest how polarization could have gone beyond business and labor to other institutions by implicating many of these institutions in fundamental political debates, often exacerbated by concerns about increasing inequality and diversity. Issues such as abortion, prayer in school, gay marriage, racial equality, gun rights, and language and immigration policy often involve the knowledge- and ­information-producing institutions (the press, TV news, science, higher education, public schools, and education) and the norm-enforcing institutions (the police, religion, and the military) on different sides of debates about these issues. Recently, with the politicization of COVID policy, they have involved medicine as well.

The story for each institution is probably somewhat different: for religion it is the rise of the Christian right, for public schools the role of teachers’ unions, for police the politics of criminal justice, for higher education the role of liberal professors, for the press its role in Watergate, and so forth. Part of the story may be that people have selected into these institutions based upon values and perspectives that put them on one side or the other of the cultural divide. The rise of talk radio, then cable television, and more recently the internet and the twenty-four-hour news cycle have amplified these political identifications and debates. The threat to a political candidate of being outflanked in a primary on the left for Democrats and on the right for Republicans further reinforces ideological polarization. The result is that ideological debate has gone beyond the business-labor divide of the New Deal to almost all American institutions.

Putting the factors together that appear to have affected trust in institutions—events and experiences specific to institutions, nonpartisan factors affecting all institutions, and partisan factors affecting institutions according to their presumed partisanship—we can speculate about the sources of declines in trust. About one-third of the overall decline in trust might be due to specific events and experiences with institutions. Another one-third might come from nonpartisan factors such as increasing inequality and diversity, leading to anomie that undermines trust in nearly all institutions among all groups. These nonpartisan factors have probably been exacerbated by an increasingly skeptical and cacophonous media environment. And a final one-third might come from partisan factors related to the emergence of cultural, social, and identity issues in American politics that have implicated nonpolitical institutions. Of course, the mix of these factors will differ for each institution, and the nonpartisan and partisan factors do not seem to apply at all for the military and not much for labor or science (until perhaps recently), but they all seem to be parts of the larger story. Because it is likely that they have interacted in complicated ways to reinforce one another, untangling them will require more research and much ingenuity.

Does all of this matter? In our 2019 survey, we asked respondents how they would feel about someone close to them (kin or friends) choosing a career in or marrying someone involved with various institutions. We found that Republicans do not want their kin or friends to have close connections with those in journalism or higher education. Democrats do not want their kin or friends to have close connections with those in the police, the military, or religious institutions. And these effects are sizeable. Republicans, for example, are more worried about their kin or friends becoming involved with a journalist than with a gay person or an atheist.28 These results suggest that the impact of institutional distrust runs deep.

Movements to defund the police, to end mask mandates, to refuse vaccination, and to overturn the 2020 election have revealed the costs of institutional distrust. Arguably, the future of democracy depends upon confidence in our institutions and the ability to bridge partisan divides. The picture painted in this essay is troubling. Confidence in institutions is declining and polarization is increasing. We do not really know enough about why this is so, and we know even less about how to fix it. To take just one example, we need to know a great deal more about when declines in confidence lead to insurrection. Those participating in and sympathetic to the January 6, 2021, insurrection have voiced their lack of trust in election systems, the federal government and its bureaucracy, the Congress of the United States, and many other institutions. How can we restore their trust?

One of the obstacles to success is that many of these controversial matters arising in the past fifty years are rooted in identities related to such characteristics as religion, race, ethnicity, nationality, gender, and sexual orientation.29 When it comes to disagreements over economic interests, it is often possible to split the difference through bargaining over money. That is very much the story of the 1940s to 1960s in America as business and labor bargained over economic matters, made even easier by the fact that economic growth meant that the results did not have to be zero-sum since both parties could benefit. It is also possible to find compromises on cultural and social issues. Roe v. Wade did this by splitting pregnancy into three trimesters, with different rules for each; past gun control legislation found compromise by focusing on outlawing only assault weapons; and immigration policy accepted facts-on-the ground in the Simpson-Mazzoli Act of 1986, which provided legalization for undocumented immigrants who had been in America for five years or more. Such compromises now seem less attainable because powerful interest groups on the left and right are especially entrenched in their positions, often because of moral concerns, deep-seated fears about compromises as “slippery slopes,” and perhaps the “zero-sum” nature of many of these issues. Consequently, it seems to be more difficult to forge compromise by meeting in the middle when it comes to conflicts over social and cultural issues, accounting for the political battles that now beset us.

The resulting oscillations and variations in laws regarding fundamental rights such as voting, immigration, and abortion are mind-boggling. In states controlled by Democrats, voting rights and voting accessibility are expanded, while in states controlled by Republicans, they are reduced and circumscribed. During the Obama administration, the Deferred Action for Childhood Arrivals (DACA) program protected immigrants who came to the United States when they were children from deportation. During the Trump administration, DACA was terminated and new applicants were rejected. In the Biden administration, DACA was restored. For almost fifty years the Supreme Court ruled that the right to abortion is fundamental, making abortion widely available and legally permissible through at least the first two trimesters of pregnancy, and even under some conditions in the third. Now abortion is banned in many states, and pregnant women face constraints that they have not experienced for fifty years. These scenarios have repeated in many other policy areas: gun laws, the environment, criminal justice, education, and even public health. There does not appear to be any middle ground. Instead, presidential administrations, sessions of the Supreme Court, and state governments are going in opposite, usually extreme directions depending upon their partisanship. Social-cultural politics lead to conflicting sets of norms and mores, as well as different cultural logics and meanings that stymie trust across partisan boundaries. It is hard to bridge these divides, especially when almost every ostensibly authoritative institution is identified with one side or the other on most issues.

In 1919, in the aftermath of World War I and the 19181919 flu pandemic, the poet William Butler Yeats used imagery of the apocalypse to describe a topsy-turvy world.30 Today, with declining and polarized trust, the sinews of society seem stretched to the point of snapping. Perhaps it is not overwrought to invoke Yeats:

Things fall apart; the centre cannot hold;
Mere anarchy is loosed upon the world


  • 1The data set we have constructed is described in Henry E. Brady and Thomas B. Kent, “Increasing Partisan Polarization since 1970 in Trust for American Non-Political Institutions,” paper presented at the September 2020 meetings of the American Political Science Association, September 8–13, 2020. The data set includes 128 surveys (43 Gallup, 31 GSS, and 54 Harris) and 165,478 respondents, with at least one survey every year from 1972 to 2021.
  • 2By “political” institutions we mean those that make or adjudicate laws and that have elected members (presidency and Congress) or many presidentially nominated and congressionally confirmed members (Supreme Court and executive branch).
  • 3The mid-point on the scale between (2) “quite a lot” and (1) “some” is 1.5. We interpreted values between 1.5 and 2.5 as indicating “quite a lot” of confidence and those between 0.5 and 1.5 as indicting just “some” confidence.
  • 4GSS and Harris use a three-point scale (a great deal of confidence, only some confidence, or hardly any confidence at all) and Gallup uses a four-point scale (a great deal, quite a lot, some, or very little). To make the responses comparable, we use a scaling technique—described in Brady and Kent, “Increasing Partisan Polarization since 1970 in Trust for American Non-Political Institutions”—that goes from zero to three points. To get these percentages, we take the difference between average trust over all the surveys in the 2010–2021 period and average trust over all surveys in the 1972–1979 period and divide it by the average trust in the earlier period to get a fraction that can be converted to a percentage. Although these percentages are somewhat arbitrary, depending as they do upon the scoring of the four-point scale, they provide a useful comparison across institutions and some idea of the magnitude of the changes.
  • 5The Pew Research Center compiles data from numerous sources on trust in government using the single question: “How much of the time do you think you can trust the government in Washington to do what is right—just about always, most of the time, or only some of the time?” Pew Research Center, “Public Trust in Government: 1958–2022,” June 6, 2022. From 1958–1964 to the present, trust in government has declined by an even larger 73 percent.
  • 6By “nonpolitical” we mean institutions that are private sector (profit-making such as business or nonprofit such as religion) or government bureaucracies that do not make or adjudicate laws and that strive to be nonpolitical, such as the military, public schools, or the police, even though they might have some elected officials (public school boards) or political appointees (military and police leaders) running them. Whatever the proper division between political and nonpolitical, it seems obvious that the U.S. Supreme Court is more politicized than the military, most public schools, or most police departments. Our surveys ask about the “executive branch of the federal government,” which is led by the president, contains many political appointees, and proposes laws, so we classify it as political. Confidence in the executive branch tracks that of confidence in the president, especially among partisan groups.
  • 7Most major nonpolitical institutions are covered. Among the few that are missing are the arts, food systems, tech companies, public utilities, philanthropy, nonprofits, and agriculture. There are occasional questions about them.
  • 8The time-series for higher education from Harris ends in 2012. Data from Pew show that confidence since then has declined. See Kim Parker, “The Growing Partisan Divide in Views of Higher Education,” Pew Research Center, August 19, 2019. Using a 2019 survey fielded as part of the Cooperative Congressional Election Study that employs the same questions as the Harris surveys, we find that confidence in higher education has decreased by 29 percent.
  • 9When we are dealing with just a few peaks and valleys on a graph like this—and there are only a few events relevant to the institution that could explain them and their dates coincide with a particular peak or valley—it seems acceptable to make causal claims, especially if one has thought a bit about possible alternative explanations as we have.
  • 10These ups and down can be seen in Figure 2, but they are even more robustly evident when the data are plotted year by year instead of averaged over two years. Bank failures hit very high peaks in 1989 and 2010. See the chart on Banking Strategist, “Bank Failures over the Decades since 1980.”
  • 11These are the squares of correlations between the confidence measure and the economic measure.
  • 12A respondent’s party identification is their response to a question such as “Generally speaking, do you usually think of yourself as a Republican, Democrat, independent, or what?” We classify those who are not Democrats or Republicans as independents.
  • 13These percentages (and the ones reported later in this section) are calculated as the difference between the average confidence of partisans of each party (Democrats and Republicans) during each time period divided by the mean value for all respondents.
  • 14If we use data from the 2019 Cooperative Congressional Election Study survey that we fielded, then higher education is now polarized by the same amount as TV news. Pew Research Center data produce a similar result.
  • 15This analysis adds the data we collected in 2019, and it shows that for the six listed institutions, partisan evaluations of confidence moved between another 0.30 to 0.50 units apart on our zero-to-three scale. Evaluations of law (which was not very polarized to begin with) and labor became somewhat less polarized by 0.15 and 0.25 units respectively, and business polarization remained the same. Confidence in both business and labor, however, remains highly polarized.
  • 16David L. Deephouse, Jonathan Bundy, Leigh Plunkett Tost, and Mark C. Suchman, “Organizational Legitimacy: Six Key Questions,” in The SAGE Handbook of Organizational Institutionalism, ed. Royston Greenwood, Christine Oliver, Thomas B. Lawrence, and Renate E. Meyer, 2nd ed. (Thousand Oaks, Calif.: SAGE Publishing, 2017), 27–54.
  • 17Social cohesion is a broader concept that includes trust in other people and trust in institutions. “Social cohesion is a state of affairs concerning both the vertical and the horizontal interactions among members of society as characterized by a set of attitudes and norms that includes trust, a sense of belonging and the willingness to participate and help, as well as their behavioural manifestations.” See Joseph Chan, Ho-Pong To, and Elaine Chan, “Reconsidering Social Cohesion: Developing a Definition and Analytical Framework for Empirical Research,” Social Indicators Research 75 (2006): 290. Chan and colleagues argue that social trust is a “horizontal” measure of social cohesion among other people, while confidence in institutions is a “vertical” measure of people’s trust in authority.
  • 18According to the Pew Research Center, “The State of Personal Trust,” July 22, 2019, “People’s views on personal trust are strongly associated with their views on issues related to institutional trust.” The General Social Survey has questions about both social trust and confidence in thirteen institutions (three political) and social trust, and our own analysis of these data shows that social trust is positively correlated with confidence for ten institutions, negatively correlated for two (organized labor and television), and not correlated for one (the military). Two of these last three are institutions for which confidence either declined only very slightly or for which it increased. In addition, just as confidence in most institutions has trended downward over the past fifty years, social trust has trended substantially downward as well. Although the correlations of institutional trust with social trust are relatively small at the individual level, they are substantial at the societal level across countries. See Kenneth Newton and Pippa Norris, “Confidence in Public Institutions: Faith, Culture, or Performance?” in Disaffected Democracies: What’s Troubling the Trilateral Countries? ed. Susan J. Pharr and Robert Putnam (Princeton, N.J.: Princeton University Press, 2000), 52–73.
  • 19These have been proposed as the causes of polarization in the U.S. Congress. See Nolan McCarty, Keith Poole, and Howard Rosenthal, Polarized America: The Dance of Ideology and Unequal Riches (Cambridge, Mass.: MIT Press, 2007).
  • 20Emmanuel Saez and Gabriel Zucman, “The Rise of Income and Wealth Inequality in America: Evidence from Distributional Macroeconomic Accounts,” Journal of Economic Perspectives 34 (4) (2020): 3–26, Figures 1, 3, and 4. Also see Thomas Piketty, Emanuel Saez, and Gabriel Zucman, “Distributional National Accounts: Methods and Estimates for the United States,” Quarterly Journal of Economics 133 (2) (2018): 553–555, 573–589.
  • 21Pew Research Center, “Most Americans Say the Declining Share of White People in the U.S. Is neither Good nor Bad for Society,” August 2021. For the 1970 data, see Campbell Gibson and Kay Jung, “Historical Census Statistics on Population Totals by Race, 1790 to 1990, and by Hispanic Origin, 1970 to 1990 for the United States, Regions, Divisions, and States,” Working Paper No. 56 (Washington, D.C.: U.S. Census Bureau, 2002), Table 1, 15% Sample for 1970.
  • 22Eric Uslaner, The Moral Foundations of Trust (New York: Cambridge University Press, 2002); and Kenneth Newton, Dietlind Stolle, and Sonja Zmerli, “Social and Political Trust,” in The Oxford Handbook of Social and Political Trust, ed. Eric Uslaner (New York: Oxford University Press, 2018), 37–56.
  • 23Tom Van Der Meer and Jochem Tolsma, “Ethnic Diversity and Its Effects on Social Cohesion,” Annual Review of Sociology 40 (2014): 459–478; Peter Thisted Dinesen and Kim Mannemar Sønderskov, “Ethnic Diversity and Social Trust: A Critical Review of the Literature and Suggestions for a Research Agenda,” in Uslaner, ed., The Oxford Handbook of Social and Political Trust, 175–204; and Merlin Schaeffer, Ethnic Diversity and Social Cohesion: Immigration, Ethnic Fractionalization and Potentials for Civic Action (New York: Routledge, 2013).
  • 24Based upon our analysis of GSS data, social trust among independents fell from 1972–1979 to 2010–2021 by more than twice as much as among partisans: 17.8 percent for independents and 8.0 percent among partisans (5.1 percent for Democrats and 13.6 percent for Republicans). The percentage of people calling themselves independents also increased from 34.3 percent to 44.3 percent, with all of the decline in the Democratic percentage.
  • 25Edward Carmines and James Stimson, Issue Evolution: Race and the Transformation of American Politics (Princeton, N.J.: Princeton University Press, 1989); Greg D. Adams, “Abortion: Evidence of an Issue Evolution,” American Journal of Political Science 41 (3) (1997): 718–737; Morris P. Fiorina and Samuel J. Abrams, “Political Polarization in the American Public,” Annual Review of Political Science 11 (2008): 563–588; Alan Abramowitz, The Disappearing Center: Engaged Citizens, Polarization, and American Democracy (New Haven, Conn.: Yale University Press, 2010); and Ezra Klein, Why We’re Polarized (New York: Avid Reader Press, 2020).
  • 26Henry E. Brady, “Trust the People: Political Party Coalitions and the 2000 Election,” in The Unfinished Election of 2000, ed. Jack Rakove (New York: Basic Books, 2001); and Henry E. Brady, “APSA Presidential Address—The Art of Political Science: Spatial Diagrams as Iconic and Revelatory,” Perspectives on Politics 9 (2) (2011): 311–331.
  • 27Thomas B. Kent, “Institutions and Constituents: Ideological Trends Over Time,” paper prepared for the Midwest Political Science Association Annual Meeting, Chicago, Illinois, April 7–10, 2022.
  • 28This research extends the notion of “affective polarization” with respect to political party identification described in Shanto Iyengar, Yptach Lelkes, Matthew Levendusky, et al., “The Origins and Consequences of Affective Polarization in the United States,” Annual Review of Political Science 22 (1) (2019): 129–146. We also find that our respondents who are partisans are less likely to want their kin or friends to have close connections with those of the other party. While the effect is strongest for Republicans, Democrats are about as unwilling to have their kin or friends have a close connection with a Republican as Republicans are unwilling to have their kin or friends have a close connection with a journalist, atheist, or gay person.
  • 29John Sides, Michael Tesler, and Lynn Vavreck, Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America (Princeton, N.J.: Princeton University Press, 2018).
  • 30William Butler Yeats, “The Second Coming,” The Collected Poems of W. B. Yeats, ed. Richard J. Finneran (New York: Simon and Schuster, 1989).