<<

The GOP’s Suburban Dilemma

Sean Trende JULY 2021

AMERICAN ENTERPRISE INSTITUTE Executive Summary

or years, Republicans thought of their party as a In the United States, the Republican Party is Fthree-legged stool, joining social, economic, and in the midst of a transition from a party based in foreign policy conservatives. There was something to the suburbs to a party based in rural areas. The this, and indeed parties can be looked at by the ideo- growing urban-suburban-rural divide increasingly logical components. This is not, however, the only way explains American politics. We would expect an to examine and understand parties. Instead, they can urban county today to be around seven points more be thought of as coalitions of different demographic Democratic than a demographically identical rural groups. Exploring how these groupings shift over time county would be. can provide clues and insights to understanding why This report establishes this phenomenon and they behave in a particular way and what their future explores its genesis. Future installments will examine may look like. the consequences of this for party coalitions.

1 The GOP’s Suburban Dilemma

Sean Trende

uring the late 2000s, as George W. Bush’s pop- It’s a little-known fact that people who did not own Dularity plummeted and nostalgia for President guns cast similar shares of votes for Republicans hit its peak in the wake of Reagan’s pass- and Democrats in different Senate races in 2018. ing, the GOP’s political coalition was commonly called Sen. Kirsten Gillibrand (D-NY) won 73 percent of the “three-legged stool.” The trope was so popular the non-gun-owning vote, while Sen. Heidi Heitkamp that presidential aspirant Mitt Romney frequently fea- (D-ND) took 63 percent of the vote from non–gun tured a three-legged chair in his stump speeches as he owners.2 The difference is in composition: Two-thirds sought the 2008 Republican nomination for president, of the electorate reported owning guns, claiming to be the only candidate who could reunite while two-thirds of the New York electorate reported the social, economic, and foreign policy conservatives not owning a gun. Similar relationships arise with the who supposedly made up the Reagan coalition.1 size of a state’s White evangelical population. There was some truth to this view of the party. This report, however, is concerned with the demo- After all, Reagan was clearly a social, fiscal, and foreign graphic composition of the Republican electorate. policy conservative, possibly the only such Republi- The thesis is that Trump’s ascension in the Republi- can president since those terms became meaningful. can Party does elevate the Republican Party’s chances It’s also clearly true that some of the challenges that of success in some states, most obviously in the conservatism faces today stem from the difficulties Upper Midwest. This will surprise few readers, given inherent in uniting these groups. Libertarian-minded his success there in 2016 and how close he came to fiscal conservatives are often wary of social conserva- winning in 2020. At the same time, though, he truly tives, especially those with religious orientations. For- endangers the Republican Party’s chances in other eign policy conservatives frequently find themselves states, especially in the Sunbelt. This can be explored at loggerheads with social conservatives, especially in through multiple angles, but this report is ultimately the era, while the fiscal conservatives concerned with urbanicity as a predictor for Republi- sometimes blanch at the level of military spending can success or failure. that the foreign policy conservatives demand. Keep- This argument will be made over four reports. This ing these groups together was relatively easy in the first report focuses on the history of the Republican era of the threat of “godless Communism”; hawks coalition and its modern development. It echoes, were kept in line by the threat, social conservatives refines, and updates arguments made in my 2012 by the godlessness, and economic conservatives by book, The Lost Majority: Why the Future of Govern- the Communism. It is likely not accidental that the ment Is Up for Grabs—and Who Will Take It.3 It traces Soviet Union’s collapse coincided with the end of the development of the Republican coalition and the large Republican wins in the popular vote in the ensu- emerging urban-rural divide. The second report looks ing election. at how this has played out, using Texas and Ohio as There are other ways of thinking about the Repub- test cases for the strengths and weaknesses of the lican political coalition’s development over time. For emerging Republican coalition. The third report looks example, one could look at its ideological composition. at a potential future for the post-Trump Republican

2 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Party, while the fourth examines problems on the But the Democrats’ dominance began to fade in Democratic side of the ledger. the 1930s. FDR’s wins in 1932 and 1936 proved that Andrew Jackson’s party was no longer dependent on the South to win. Moreover, where would the South The Development of the Modern go? The GOP was still the party of civil rights, after all. GOP Coalition FDR embarked on an ill-considered purge of South- ern Democratic senators in the wake of his 1936 land- The modern Republican Party can be seen as a coali- slide victory. All survived, but the message had been tion of four geographic or demographic parts first put sent that the marriage of conservatism and the Dem- together during the Dwight Eisenhower years. Over ocrats was beginning to fray. Moreover, FDR’s threats a traditional Republican base of business owners and proved empty, which caused problems for him down “Main Street Republicans” in the Midwest and on the the road.9 Great Plains, postwar Republican candidates added In the wake of the 1938 midterm election, conser- three constituencies: the South, the suburbs, and the vative Southerners joined ranks with conservative White working class. This report discusses the rise of Republicans in the North to block most New Deal that coalition, foreshadows the forces that are now legislation. By 1942, the average Southern Democrat working to unwind it, and hints at what might arrive was positioned roughly in the middle of Republi- to take its place. cans and Northern Democrats when voting on eco- nomic issues. In fact, many were effectively voting like Republicans by then. The Republican position on The South civil rights and the fact that the stock market crash and even the Civil War itself were living memories for In 1936, the Republican Party was devastated in a many Southerners kept these members in the Demo- way that few American parties have ever been. In the cratic coalition.10 aftermath of Franklin D. Roosevelt’s (FDR) reelec- A combination of factors ultimately broke South- tion, Republicans were reduced to just 89 seats in erners away from the Democratic Party. First, the the House and 17 in the Senate; there were almost as South became increasingly wealthy, as what had pre- many Democratic senators as there were Republican viously been sleepy hamlets grew, almost overnight, House members.4 Yet within two years, the Repub- into cities. As the Democratic Party became increas- lican Party had bounced back enough that the New ingly liberal on economic issues, wealthy voters fol- Deal was effectively over.5 Within six years, it had won lowed suit. Samuel Lubell, who made a career out the popular vote for Congress, and within 10 years, it of predicting elections by interviewing voters across had won control of Congress.6 Six years later, Repub- the country, noticed that wealthy denizens of Hous- licans held control of Congress and the presidency.7 ton had gone solidly for FDR in 1936, giving him How did this happen? It was due to various por- 57 percent of the vote. By 1944, FDR had received just tions of the New Deal coalition splintering off over 18 percent of the vote among this group. (The poorest FDR’s final term. Perhaps the most surprising portion residents continually gave him more than 85 percent was the American South. Since the end of Reconstruc- of the vote.) tion, the South had been the core of the Democratic Second, the parties changed on civil rights. One coalition. Democratic dominance in the region was so may think of the relationship between Southerners thorough that in 1924, Sen. Coleman Blease (D-SC) and the Democratic Party as an unhappy marriage, mocked the 1,123 votes that Calvin Coolidge received in which they stayed together for the kids—the kids in his state by remarking, “I do not know where he got in this case being Jim Crow. In 1948, the Democratic them. I was astonished to know that they were cast Party put support for civil rights in its platform—to and shocked to know that they were counted.”8 torture the analogy, the kids went off to college—and

3 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

at that point Whites who were otherwise closer ideo- Although the White working-class voters who fled logically to Republicans had no reason to stay in the to the suburbs were originally heavily Democratic, Democratic Party. It is not accidental that, eight years they formed the bedrock of the Republican Party from later, Eisenhower won a plurality of the popular vote 1952 onward. While scholars debate whether the sub- in the South. Finally, the in-migration of voters from urbs helped make working-class voters more Republi- the North to the South (coincident with the increased can or whether the voters who moved to the suburbs growth of Southern cities) meant these voters were more Republican to begin with, their lean from imported their Republican voting habits with them. 1952 to 1988 was decidedly toward the GOP (Figure 1). Republican gains in the South proceeded in fits According to the American National Election Stud- and starts from that point. During the intervening ies, Republicans won the suburban vote in excess of years between 1962 and 1992, the Republican share 60 percent in half these years. Perhaps ominously for of the Southern congressional vote varied between Republicans, their vote share collapsed among sub- 32 percent and 43 percent, with a slightly upward urbanites in 1964, when they nominated someone slope. That changed in 1994, when Republicans won perceived as far to the right of American politics, par- a majority of districts in the South for the first time ticularly on civil rights issues. since Reconstruction. They have done so in every election since. We will save discussion of the mod- ern South for a later report; for our purposes, at this The White Working Class point, the first leg of the stool of the modern GOP coalition was built. For most of the late 20th century, the battle for con- trol of Congress and the presidency was largely fought over the votes of working-class Whites. Upper-middle- The Suburbs class and upper-class voters were largely Republican, while liberals and minority voters were Democratic. The second leg of the Republican stool was the sub- Whites without college degrees swung back and forth urbs. For decades, new immigrants had swelled the and often provided the margin in the presidential population of Manhattan and other cities, with ten- elections (Figure 2). From 1952, these working-class ements packed to the brim with immigrant families Whites often provided the margin between victory springing up over the urban landscape. But with the and defeat. invention and widespread adoption at the turn of the You can note two things in Figure 2. First, when century of streetcars, subways, trolleys, and then the Republicans lost the White working class, they lost automobile, it became increasingly feasible for the the elections. Specifically, in 1960, 1964, and 1976, tenements to empty out and the countryside to fill up. Democrats won the presidency and won Whites The Great Depression and World War II dampened without college degrees. In all other years, they lost. suburban growth, but their aftermath left Americans Second, the differential in voting patterns between in a mood to spend and grow. Whites with and without college degrees disappears In 1947, William Levitt created Levittown, a over time. The gap between the two was a yawning (racially exclusive) suburb situated in what was then 19 points in 1952 but began to close in earnest with the potato fields of Long Island. His idea of small, Richard Nixon’s two elections. (Whites without col- affordable homes captured the imagination of the lege degrees were more Republican than were Whites American middle and working classes, and between with degrees in 1972.) By the 1980s, it had consistently 1940 and 1955, the percentage of Americans living fallen to single digits. in suburbs tripled, from 9 percent to 28 percent.11 Similar trends appear in the congressional data, By 1988, a near majority of the electorate voted when Republicans were less successful in winning in suburbs.12 majorities during this period (Figure 3).

4 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Figure 1. Republican Share of Two-Party Vote in Suburbs, 1952–88

80

70

60

50

40 centage r

Pe 30

20

10

0 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988

Source: American National Election Studies.

Figure 2. Republican Share of Two-Party Vote Among Whites by Education, 1952–88

80

70 College 60 50 40 centage r 30 Pe Noncollege 20 10 0 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988

Source: American National Election Studies.

5 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Figure 3. Republican Share of Congressional Two-Party Vote Among Whites by Education, 1952–90

80 70 60 College 50 40 centage r

Pe 30 Noncollege 20 10 0 1952 1958 1962 1966 1970 1974 1978 1982 1986 1990

Source: American National Election Studies.

Republican strength among Whites with and The Growth of the Urban-Suburban-Rural without college degrees was less pronounced in Divide Congress during this period, as evidenced by the Republicans’ failure to capture the lower chamber By the end of the 1980s, Republicans were looking for any time during these years, save 1952. But it was like a “party of everyone,” at least at the presidential not for lack of strength among Whites with college level. They had won seven of 10 elections, six of which degrees, who usually handed a majority of their vote they won with popular vote majorities in excess to the GOP. The real weakness came among Whites of eight points and their standard-bearers had won without college degrees, a group the GOP never won in excess of 75 percent of the popular vote. They had during these years. seen their Senate majority slip away in the late 1980s, Suburbanites, Southerners, and “traditional” Repub- but that they had won it in the first place served as licans gave the Republican Party a solid base during a reminder that they were not necessarily a legisla- 1950–80. The majority makers were Whites without tive minority party. The Democrats boasted of a large college degrees. If Republicans could only figure out congressional majority, but the ongoing realignment a way to solidify their standing among Whites with- of Southern Democrats toward the Republican Party out college degrees while holding on to Whites with suggested that the coalition’s days were numbered. college degrees, they would have a lock on American Enter . This portion of the story will be politics. The latter, it turned out, would prove much dealt with in more detail in the fourth report of this trickier than the former. series, but a major focus of Clinton’s presidency was to increasingly bring college-educated Whites and suburbanites into the Democratic fold. This created tensions in the Republican coalition and contributed to the difficulties they faced in subsequent presiden- tial elections.

6 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Figure 4. Bill Clinton Vote in the Midwest, Normalized to National Vote, 1992

Source: Author.

It also, however, created a divide between urban in Appalachia, Arkansas, southeast Oklahoma, and and rural counties in the United States. Examine even Texas. Figure 4, which shows the partisan lean of Midwest By 2016, however, this all changed (Figure 7). counties. It is normalized to the national vote, which Now, almost all the rural areas are red. In addition, is another way of saying that the national vote share the major metro areas listed above are readily appar- has been subtracted out. This allows us to compare ent to anyone familiar with the South’s geography. In the partisan orientation of counties across elections other words, one of the contributions of both Clin- while accounting for differences in the national envi- ton campaigns was to accelerate the geographic sort- ronment and, hence, vote share. ing of America; the suburbs were no longer solidly in We see that, even in 1996, Clinton did well in mid- the Republican Party, but the rural areas were. At the western rural areas. Compare this to Figure 5. The same time, densely populated urban areas accelerated Democrats’ advantages in rural areas completely dis- their move toward the Democratic Party. appeared. Indeed, someone with a solid understand- To be more precise about our intuitions here ing of political geography could pick out almost every and to see an extension into 2020, consider the fol- blue county and identify it as either a major city, a col- lowing regression analysis. If you understand what lege town, or an Indian reservation. a regression analysis is, you can probably skip the A similar story emerges in the South in 1996 next few paragraphs, but if you don’t, this should (Figure 6). Most of the major metropolitan areas— provide enough information to help you understand Atlanta, Georgia; Birmingham, Alabama; Char- Tables 1 and 2. lotte, North Carolina; Dallas, Texas; and Houston, Perhaps in your nightmares you will recall having Texas—are barely noticeable on the map, while to solve problems like this: If Train A left at 2:45 a.m. Clinton retains significant strength in rural areas and Train B left at 3:15 a.m. and they traveled at set

7 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Figure 5. Vote in Midwest, Normalized to National Vote, 2016

Source: Author.

Figure 6. Bill Clinton Vote in the South, Normalized to National Vote, 1996

Source: Author.

8 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Figure 7. Hillary Clinton Vote in the South, Normalized to National Vote, 2016

Source: Author. velocities, you could (at least in theory) calculate three hours, it would probably be unreasonable to when and where they would meet. Almost all the math suggest that the next trip will take 10 hours. It might you learned in high school and college was like this. be reasonable to suggest it will take two hours and You were given certain information, and by using the 10 minutes. While statistics doesn’t necessarily give correct techniques, you could calculate a clear answer us “certain” answers and will occasionally give a flatly with certainty. incorrect answer, over time it provides better answers However, statistics is a different branch of mathe- than if we were to blindly guess. matics. Statistics is the mathematical study of uncer- Regression analysis is a type of statistical analysis. tainty. Suppose that the trains don’t travel at set Its concern is with quantifying how certain we are that velocities. Instead, you have a list of hundreds of trips a given input affects an output and to what degree it the trains have made. As it turns out, their velocities does so. In our example above, suppose we also have vary: Sometimes they encounter bad weather, per- data on rainfall recorded along the tracks for each trip. haps some drivers are more aggressive than others We could perform a regression analysis to confirm are, maybe a goat is on the track, and so forth. Given or disconfirm our intuition about the relationship (if this, we can’t confidently say when the trains would any) between rain and travel time and to give us the meet. Instead, a statistician might observe, we can best estimate for how much each inch of rain affects generate an “expectation” of when they would meet, the time in which the train completes its journey. given our data. By observing how much these trips This is all a regression analysis is. If you happen to vary from the average time (statisticians cleverly call remember from pre-algebra that the equation form of this “the variance”), we might even be able to create a line is y = mx + b, then you can understand regression bounds to separate a good guess from a bad guess. In analysis, because all a regression analysis really does is other words, if the trips vary between one hour and produce the equation for the line13 that best explains

9 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

the relationship in the data.14 The y and x are the data Ohio), “5” if it is part of a large city (think Richmond, you put in. (In our example, y is the time to complete a Virginia, or Hampton Roads, Virginia), and “6” if it trip, and x is the amount of rain.) The b represents the is part of a megacity (think Washington, DC, metro intercept (statisticians call it β0 or the “constant,” but area). I then impose some statistical controls and it’s the same thing), which is what the time would be if test whether, as a county becomes increasingly urban, there were no rainfall. The m is the most important. It is Democratic vote share increases (Table 1). the slope, or the “coefficient.” It tells you what the best The controls are mostly what we would expect. As estimate is for how much of an increase or decrease income increases, Democratic vote share declines. As you will get in y for each unit of x. In our example, it the non-White share increases, Democratic vote share tells us the best estimate for how much each inch of increases. More Whites with college degrees increase rain slows down (or speeds up) the train trip. Democratic vote share, consistent with Trump’s over- There’s one more important concept: the p value. performance among Whites without college degrees. The p value is probably the most misused statis- Newer housing is negatively correlated with Demo- tic in general literature, and I will avoid attempting cratic vote share, consistent with an understanding of a detailed description of it here (and in the process better Democratic performance in the inner suburbs will take some shortcuts that may drive a statisti- than in the exurbs. The two surprises are unemploy- cian reading this crazy). Suffice it to say, the p value ment rate, which we might expect to be negatively is a measure of how sure we are that the relationship correlated while Democrats hold the presidency, and we found between our x and y is, in fact, “real.” In percentage foreign born, which we might expect to other words, your sample might suggest that as rain be positively correlated. There are also controls for increases, the time to complete a trip increases, but which state the county is located in, which I have your sample is still just a sample of a larger universe omitted from Table 1. of possible trips; it therefore has an error margin. Just Most importantly, however, we see that “urban as polls that are “within the error” margin aren’t con- level” is positively correlated, suggesting that, after sidered different, from a statistical point of view, so imposing various controls, as you went from a too regression results with large p values aren’t con- rural county to a mega-urban county, Democratic sidered valid differences. By convention, a regression vote share did increase. The three asterisks tell us result with a p value that is larger than 0.05 is con- the p value is less than 0.01, showing confidence sidered to show weak evidence against the idea that in the conclusion. (In fact, the p value is less than there is no relationship between the variables, while a 0.000000000000002, suggesting that it would be p value in excess of 0.1 provides little-to-no evidence extraordinarily unusual to ever see these data if there against the suggestion that there is not, in fact, a rela- were not a relationship between urban level and Dem- tionship between the variables.15 ocratic vote share.) Performing regression analysis on US elections The coefficient is 1.227. This suggests that after from 1988 to the present shows how the Clintonian controlling for things such as race, income, and col- shift has raised the Democrats’ performance in large lege education, a county in a small town micropo- urban areas but weakened the Democrats in rural litan area would be expected to give an areas. To run this, every county in the US was coded additional 1.25 points (2.5 net points) compared to a to the metropolitan or micropolitan area of which it is similarly situated rural county. After controlling for a part, as defined by the US Census Bureau. A county traditional factors thought to influence vote choice, is coded “1” if it is not part of any metropolitan area, a megacity (such as Chicago) would still be expected “2” if it is part of a small town (think Martinsville, to give Biden an additional six points (12 net points) Virginia), “3” if it is part of a large town (think Char- than a rural county would. In other words, in 2020 the lottesville, Virginia), “4” if it is part of a small city rural-urban divide informed us how a county would (Virginia doesn’t have any of these, but think Dayton, ultimately vote.

10 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Table 1. Predictors of Democratic Vote Share in US Table 2. Strength of Urbanicity Counties in 2020 Variable in Predicting Democratic Vote Share, 1988–2020 Dependent Variable: Democratic Vote Share in 2020 Year Coeffcient P Value –0.0001*** Median Income (0.00002) 2020 1.23 2.7e–27 0.599*** 2016 1.05 7.7e–19 Percentage Black (0.014) 2012 0.87 2.2e–11 0.280*** 2008 0.73 1.8e–08 Percentage Hispanic (0.018) 2004 0.62 2.3e–06 0.766*** 2000 0.83 4.3e–10 Unemployment Rate (0.050) 1996 0.30 0.03 –0.127*** Percentage Foreign Born 1992 0.33 0.02 (0.042) 1988 0.29 0.04 New Housing (Percentage –0.309* Built After 2010) (0.170) Note: In scientific notation, “e” means “10 to the White College-Educated 91.262*** power of.” For example, 4.3e–10 represents a dec- imal point with nine zeroes followed by 43. Percentage (1.829) Source: Author. 1.227*** Urban Level (0.112) confidence that there is a relationship –7.183*** Constant between urban level and Democratic voting (1.290) share improves over time. The p value for 1988 barely crosses the threshold of signif- Observations 3,110 icant. (If we ran an experiment every day, R2 0.833 you’d expect to see a false positive every 2 Adjusted R 0.830 25 days with a p value of 0.04.) By 2016, it is 6.713 Residual Standard Error extremely significant. (If we ran an experi- (Degrees of Freedom = 3,053) ment every day, you’d expect to see a false 271.751*** F Statistic positive in about three quadrillion years.) (Degrees of Freedom = 56; 3,053) The Bill Clinton bid to weaken the three- Note: *p < 0.1; **p < 0.05; ***p < 0.01. Numbers in parentheses are stan- legged stool of the GOP’s coalition worked dard errors. in many respects. By emphasizing fiscal Source: Author. moderation and a quiet social liberalism, he expanded the Democrats’ foothold in the suburbs of major cities, which allowed them But it was not always this way. Instead, this gap to become more competitive in the Midwest and Flor- between urban and rural communities grew gradu- ida and to put much of New England and California ally over the 1990s, 2000s, and 2010s. This is shown out of Republicans’ reach. At the same time, he weak- in Table 2. ened the Democratic Party in rural areas, putting The large coefficient develops over time. As of places such as Kentucky, Tennessee, and West Vir- 1988, the difference between a megacity and a rural ginia out of his party’s reach. area is a little less than 1.5 points, compared to the The urban-suburban-rural divide, along with the gaudier six-point difference in 2020. Moreover, large urban area versus smaller urban area versus towns

11 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

divide emerging in American politics, continues to chal- the ashes of the Trump presidency. Finally, I’ll examine lenge the GOP. It is likely, for example, that the GOP the problems Democrats might have in holding their would have easily held on to its Senate seats in Georgia nascent coalition together. had the megacity of Atlanta not swung against them. But it is not necessarily the existential threat that many make it out to be. As I will explore in the second report About the Author in the series, the growing coalitions pose trade-offs for Republicans: While states like Texas might be moving Sean Trende is a nonresident fellow at the Ameri- against the party of Lincoln, states like Ohio are fall- can Enterprise Institute, where he works on elections, ing more solidly in its column. In the third report, I’ll American political trends, voting patterns, and demo- explore other potential outcomes for the GOP, includ- graphics. He is also the senior elections analyst for ing how a much broader coalition could be built from RealClearPolitics.

12 THE GOP’S SUBURBAN DILEMMA SEAN TRENDE

Notes

1. Michael Luo, “Romney’s 3-Legged Stool Takes the Stage,” New York Times, Caucus, July 7, 2007, https://thecaucus.blogs.nytimes. com/2007/07/27/romneys-3-legged-stool-takes-the-stage/. 2. CNN, “Exit Polls: New York Senate,” https://www.cnn.com/election/2018/exit-polls/new-york/senate; and CNN, “Exit Polls: North Dakota Senate,” https://www.cnn.com/election/2018/exit-polls/north-dakota/senate. 3. Sean Trende, The Lost Majority: Why the Future of Government Is Up for Grabs—and Who Will Take It (New York: St. Martin’s Press, 2012). 4. Walter Dean Burnham, Thomas Ferguson, and Louis Ferleger, Voting in American Elections: The Shaping of the American Political Universe Since 1788 (Washington, DC: Academica Press, 2009), 177. 5. Burnham, Ferguson, and Ferleger, Voting in American Elections, 178. 6. Burnham, Ferguson, and Ferleger, Voting in American Elections, 179–80. 7. Burnham, Ferguson, and Ferleger, Voting in American Elections, 180. 8. Burnham, Ferguson, and Ferleger, Voting in American Elections, 47. 9. Susan Dunn, Roosevelt’s Purge (Cambridge, MA: Belknap Press, 2010). 10. Trende, The Lost Majority. 11. Fred I. Greenstein and Raymond E. Wolfinger, “The Suburbs and Shifting Party Loyalties,”Public Opinion Quarterly 22, no. 4 (Winter 1958–59): 473, https://www.jstor.org/stable/2746595. 12. American National Election Studies, 1988 Time Series Study, https://electionstudies.org/data-center/1988-time-series/. 13. To be technical, when two variables are measured, you are creating a plane, and when three or more variables are created, you are making a hyperplane. You can still think of it as a line without losing too much understanding. 14. “Best fit” has a particular meaning for regression: It is the line that minimizes the sum of squared errors. For a casual reader, how- ever, it is sufficient to understand this as meaning that you get as close to the observations as possible. 15. Larry A. Wasserman, All of Statistics: A Concise Course in Statistical Inference (New York: Springer, 2004), 157. To be sure, the fore- going is a concise explanation of regression analysis; a deep understanding could fill several book-length treatments. For example, see Michael H. Kutner, Christopher Nachtscheim, and John Neter, Applied Linear Regression Models, 4th ed. (New York: McGraw Hill Edu- cation, 2004), 283.

© 2021 by the American Enterprise Institute for Public Policy Research. All rights reserved. The American Enterprise Institute (AEI) is a nonpartisan, nonprofit, 501(c)(3) educational organization and does not take institutional positions on any issues. The views expressed here are those of the author(s).

13