2. Parties and polarization: National- and state-level actors

Many scholars emphasize that polarization has been driven by elites rather than by the mass public,Footnote ¹ pointing to evidence that the average voter continues to gravitate to the political center.Footnote ² Explaining the rise of partisan polarization in recent decades, other analysts have pointed to increases in income inequality,Footnote ³ party sorting,Footnote ⁴ the changing place of race, gender, and sexuality in American life,Footnote ⁵ and electoral parity between the two major political parties.Footnote ⁶ Understanding how parties change—specifically, how today’s parties have polarized so sharply and dramatically—is a central concern for scholars, indeed for anyone concerned about American democracy. How did the bipartisan consensus of the mid-twentieth century come undone?

Since 1955, when V. O. Key introduced the concept of critical elections, political scientists have developed multiple approaches to explain shifts in party alignments. The first body of theory emphasized abrupt shifts that occurred at specific moments, whether a single election or over a small set of years.Footnote ⁷ Modifying this approach, Carmines and Stimson, drawing on the civil rights revolution, proposed a model of “issue evolution” with two key features.Footnote ⁸ The first is that most position change occurs at relatively brief critical moments, which then set in motion a period of slower, path-dependent change. The second is that national elites play the key role in deciding their party’s position, and that voters ultimately take their cues from these elites. Adams, and also Carmines and Woods, utilize the “issue evolution” model to explain position-taking, by the two national parties, on abortion.Footnote ⁹ They argue that national elites drive partisan polarization on abortion, and that the mass public appears to respond, with a lag, to the new party positions.

Work growing out of the UCLA school of political parties also tends to view position-taking by parties as largely a top-down process, one that is driven by national party leaders and their interactions with the leaders of key interest groups. These groups can be members of the party’s existing electoral coalition or they can be potential partners, valued for the resources that they could bring to electoral campaigns.Footnote ¹⁰ Scholars in this tradition have considered party evolution on abortion, though their work usually focuses on roll call voting in Congress or position-taking by national political figures, rather than official party positions contained in platforms. Karol, for example, emphasizes the ways in which presidential nominees worked to shape their party’s position on abortion in order to gain electoral advantage,Footnote ¹¹ and Schlozman explores the ways in which “New Right” brokers at the national level brought social conservatives into the Republican Party.Footnote ¹² While the UCLA school’s approach is somewhat agnostic as to the speed of new position-taking, both Karol and Schlozman show that the partisan divide among elites on the issue of abortion unfolded over many years, even decades.

Feinstein and Schickler present an alternative approach to party shifts.Footnote ¹³ Studying party positions on race and Black civil rights, they argue that the initiative came from state leaders, not national leaders, and that there is no critical moment in which party positions changed. Examining state party platforms, Feinstein and Schickler show that state and local parties drove the process, that national leaders were constrained by position-taking by Democratic parties in northern states, and that this transformation had occurred by the mid-1940s.Footnote ¹⁴ In a recent article, Hopkins, Schickler, and Azizi use topic modeling and another large database of state party platforms—1,783 platforms covering the period 1918–2017—to study the timing and mechanics of political polarization, finding that the process was driven by state parties.Footnote ¹⁵ They operationalize polarization not as parties taking opposite positions on a given set of issues, but rather as parties discussing different topics.

These competing arguments yield two hypotheses:

3. Culture war: Timing

Since Hunter’s seminal book,Footnote ¹⁶ most scholarship on the culture war has focused on a small number of critical moments and decisions by national elites, including Ronald Reagan, Pat Buchanan, and Bill Clinton. Layman notes the importance of Reagan in bringing religious conservatives into the Republican Party in the late 1970s and early 1980s,Footnote ¹⁷ and Fetner argues that Clinton’s presidency represented the crucial moment in party-positioning on LGBTQ rights.Footnote ¹⁸ In the literature generally, scholars focus on those two eras—the late 1970s and the 1990s—in their assessments of when the culture war emerged.

Some point to the ways in which parties began diverging on social issues in the late 1970s, in the aftermath of Roe v. Wade, the battle over the Equal Rights Amendment, the organization of the Moral Majority, and the opposing positions taken by Carter and Reagan in the 1980 election. Schlozman argues that it was in the late 1970s that the national Republican Party forged its alliance with white evangelical Christians and began to embrace a conservative social agenda.Footnote ¹⁹ By then, both national parties had staked out clear positions on abortion in their platforms, though partisan opinion at the mass level had not yet shifted.Footnote ²⁰

But others emphasize instead a dramatic shift in the early 1990s. Certainly by 1992, when Buchanan addressed the Republican National Convention—declaring, “There is a religious war going on in this country, a cultural war as critical to the kind of nation we shall be as the Cold War itself, for this war is for the soul of America”Footnote ²¹—the battle lines between the national parties were clearly drawn.Footnote ²² Only in that year did the two national parties adopt clear and opposing positions on LGBTQ rights.Footnote ²³ Hopkins et al., studying polarization more broadly but not the culture war, also find that there was a critical juncture at this time, “a sudden discontinuous rise in polarization in the mid 1990s.”Footnote ²⁴ Text analyses of congressional floor speech also suggest that partisan polarization increased sharply in the early 1990s.Footnote ²⁵

Research on the culture war, and party change more generally, yield three hypotheses relating to the timing of polarization on these issues:

Until now, most of the literature on the culture war has focused on events at the national level or, at least, local or state events with national resonance—such as Anita Bryant’s 1977 effort to roll back a gay rights ordinance in Florida’s Dade County, the 2003 decision of the Massachusetts Supreme Judicial Court to recognize same-sex marriage rights, and subsequent debates over same-sex marriage. Karol and Thurston’s 2020 study of abortion votes in the California State Assembly is a rare example of a state-level study.Footnote ²⁶ Although the national evidence is only suggestive, it appears that Democrats and liberals initiated conflict on these issues, upsetting a status quo that limited abortion rights and rejected the legitimacy of homosexuality.Footnote ²⁷

4. State party platforms

Since at least the mid-nineteenth century, most state political parties have met in biennial conventions to endorse nominees for office and to write and adopt platforms. Platforms are typically authored by a committee of individuals, including state and county chairs, members of the state party’s executive committee, state legislators and other state elected officials, and party activists.Footnote ²⁸ As we examine the platform-writing process in many of these states, there is evidence that issue activists sometimes play a prominent role in shaping platform positions. But, in almost every case, platforms are debated, amended, and approved by delegates at state party conventions. While we are still in an early stage of studying the process of platform writing to understand the relative roles of party regulars, elected officials, and activists in this process, the platform itself carries the imprimatur of the party.

The average voter may not read state party platforms, but interest groups and issue activists care deeply about the positions the party takes. Platforms have consequences, sometimes immediate. When in 1970 the California Democratic Party for the first time adopted a plank in its platform supporting abortion rights, the backlash was swift and public. “More than 500 members of St. Barbara’s Roman Catholic Church have announced that they have changed their voter registration from Democratic to Republican because of the California Democrats’ platform’s position on abortion,” the Sacramento Bee reported.Footnote ²⁹ “‘I’m a Democrat and I’m leading the van,’ said the pastor, the Rev. Michael Collins.”

Party platforms represent the party’s official issue positions. Of course, there are other ways to measure where a party stands on an issue, such as roll call votes and position-taking by its elected officials and candidates as well as the opinions held by a party’s voters. In other work, we are developing and analyzing new data sets that allow us to consider these other dimensions of where a party stands on a given issue. But only the platform represents the official positions of the party.

In comparative politics, scholars have regularly drawn upon national party platforms to study politics. The most noteworthy of these endeavors is the Comparative Manifestos Project. This effort, which began in 1979, has compiled a database of national party platforms from fifty countries covering all free elections since 1945. These data have generated a wealth of empirical results as well as numerous new insights about party competition and policymaking.Footnote ³⁰ This project has also spurred innovations in computer-assisted techniques for coding and interpreting political texts.Footnote ³¹

There have been fewer efforts to study political platforms in the context of the United States, though that is quickly changing. Gerring uses national party platforms to qualitatively trace the evolving ideologies of America’s national political parties,Footnote ³² while Levendusky uses national platforms to document increased polarization of the Democratic and Republican parties in recent decades.Footnote ³³ A number of scholars have also now begun studying state party platforms. Paddock collected and analyzed platforms from eleven states to study the transformation of state party organizations; in documenting the nationalization of American politics, Hopkins undertook content analysis on a large collection of state party platforms.Footnote ³⁴ Most relevant to our own work is the scholarship by Hopkins et al. using state party platforms to study partisan polarization in the 1918–2017 period and the work on race and Black civil rights in the article by Feinstein and Schickler and in the book by Schickler.Footnote ³⁵

5. Locating state party platforms, 1960–2018

The archive that we have constructed contains Democratic and Republican state party platforms written from 1960 through 2018. While other recent studies, above all Hopkins et al., draw on comparable numbers of platforms,Footnote ³⁶ our collection for this particular time period is considerably larger than any that has previously been assembled. By collecting all platforms since 1960, we ensure that our data begin prior to party position-taking on either abortion or LGBTQ rights. When we started this project, we benefited greatly from the generosity and research of other scholars, who shared with us all of the state party platforms that they had discovered in their own research.Footnote ³⁷ We subsequently proceeded to locate hundreds of additional state platforms scattered among historical societies, manuscript holdings in archives and special collections libraries, state parties, and party activists. The fruit of several years of intensive research, this archive now contains 1,771 unique state party platforms, along with newspaper summaries of another 206 platforms.

Collecting historic party platforms presents several challenges. Chief among these is that, unlike many other types of historic documents, state governments do not archive platforms. While we found two states that published party platforms in serial publications—Wisconsin and, for some years, Kansas—as a rule, state party platforms have not been systematically preserved. To find these platforms is to seek buried treasure—state by state, party by party, year by year—through a quest where no map exists as a guide and where, in many cases, much of the treasure lies forgotten or has been permanently lost or destroyed.

The state party offices themselves, which we initially thought might be among the best repositories of their own party history, almost never maintain any significant archives. As we came to learn, state parties are forward-looking organizations, focused on winning the next election rather than keeping records of past actions. There are, to be sure, encouraging exceptions. For instance, the attic of the Montana Democratic Party offices contains several old platforms, and the South Carolina Republicans maintain a shed that serves as an informal archive of old party documents (including platforms). Most state parties, however, have small staffs with limited record keeping or—and perhaps even more regrettable—excellent record keeping marked by periodic, wholesale purges. (We were told by one party that, just a year or two before we had reached out to them, they had discarded decades’ worth of platforms and documentation related to the platform-making process.) Some parties seem, in retrospect, to regret having disposed of these records: the Hawaii Democratic Party, for example, upon learning of our archival effort, requested that we send them copies of their old platforms.

Due to the general lack of preservation by the state parties, we largely turned to other sources. State platforms are rarely given a call number and put on a library shelf, but a small fraction of them are, and we obtained these through interlibrary loan. Many states have historical societies, a handful of which maintain historical state platforms in designated folders. For instance, we obtained several documents from the “platform folders” at the Iowa Historical Society. Very few libraries, however, maintain platforms as discrete, stand-alone collections.

Our most plentiful sources of platforms have been special collections libraries and archives. Every state has at least one major library (often attached to a university) that preserves the papers of past political figures (e.g., governors, members of Congress, state legislators, and nonelected activists) and—when we are particularly lucky—the state parties themselves. For many prominent politicians, often over 100 feet of documents will comprise their collection; we have found many platforms scattered across scores of such collections. The specificity of the collection’s finding aid determines how efficiently we can locate these documents. Many collections at the Marriott Library at the University of Utah, for instance, have document-level descriptions of the contents, allowing us to locate and request state platforms with relative ease. For many other libraries, however, the finding aid lacks such specificity, and boxes with generic headings such as “state party” and “campaign literature” must be searched with the hope that they contain missing platforms. In total, a member of our research team visited sixty-five archives across thirty states, with every visit in person.Footnote ³⁸ We list these archives in Appendix Figure A1.

Finally, we contacted present and former party leaders and activists and, in some cases, their surviving children, to inquire whether they had kept old platforms themselves. This proved successful in our quest to acquire a complete set of Vermont Republican platforms. After over a year of searching, we had obtained most of the party’s platforms, but became progressively convinced that the few we were missing were lost to history. Then, in one of our more exciting adventures, we made contact with a party activist who had decades-old state platforms in his barn in rural Vermont, which allowed us to complete our collection of Vermont Republican platforms. Two other historic platforms—in this case, what appear to be the only surviving copies of the 1964 and 1968 Virginia Democratic platforms—were shared with us by a long-time party activist in Virginia, who was given the platforms for safekeeping by a former cochair of the state party soon before the cochair passed away.

The likelihood of our finding a platform is mediated, in part, by how many copies were produced and distributed in the first place. The Connecticut platforms of the 1960s and 1970s were made by a professional advertising company and appear to have been widely shared. The Iowa Republican platforms of the 1980s direct that the document be mailed to all Republican state convention delegates, officeholders, and candidates as well as all Iowa government teachers and news media. Most platforms, unfortunately, were not so assiduously dispersed, and we fear that some are permanently lost. Many appear to be unique copies, typed out on sheets of paper or, literally, carbon copies with handwritten marginal notes reflecting changes made at the convention.

As Appendix Table A2 shows, we discovered in our research that two-thirds of all state parties were consistent platform writers. These parties, at least since 1960, have written and adopted a new platform every two years, almost without exception. A small number, including the Illinois Republican Party since 1996 and both West Virginia parties, have consistently written platforms, but only during presidential election years. Beyond this core group, an additional 12 percent of state parties wrote platforms throughout the entire period of study, but did so intermittently: the Virginia Republicans, for instance, adopted between one and three platforms during each of the past six decades. Another 14 percent of parties consistently wrote platforms at the start of our time series, but eventually stopped. Of those parties that stopped writing platforms, half did so after 1992, well after abortion and (to a lesser extent) LGBTQ rights had emerged as salient partisan political issues. In total, only eight state parties—both parties in Kentucky, Maryland, and Tennessee, and only the Democrats in Alabama and Louisiana—either never or almost never wrote platforms during the time period of our analysis.

When we could not locate any copy of a particular platform, we searched newspaper archives for coverage of the state party convention. This was a time-consuming process. In total, our efforts yielded quality summaries for 206 of our missing platforms. These summaries allow us to determine whether the party in question took positions on abortion and LGBTQ rights. If newspaper summaries did not mention either issue, we assume that the party took no official position. Given the controversial nature of abortion and LGBTQ rights throughout this entire period, it is unlikely that any news article would fail to note positions taken on these matters. Indeed, both abortion and LGBTQ rights tend to be featured prominently in these articles.

Adjusting the denominator for instances where we know, from a credible source, that no platform had been written, we are confident that we have obtained either a copy or a thorough newspaper summary of at least 85 percent of all state party platforms in the 1960–2018 period, with even distribution between the two parties. This is a conservative estimate. If we were to remove cases where we suspect, but cannot prove, that no platform was written, our success rate would be considerably higher.

Figure 1 presents, by year, a count of the total number of platforms. This times series ranges from a high of 88 in 1962 to a low of 55 in 1998. (Since there were eight state parties that never, or nearly never, wrote a platform, the maximum number possible, in a normal year, would be 92.) The biennial mean is 70. Despite the challenges in obtaining historical platforms, our efforts have yielded a large number of platforms for each election cycle. Perhaps surprisingly, we were able to locate a large number of platforms from the earliest years in our time series; indeed, the years for which we found the largest number of platforms were 1960 through 1970.

Figure 1.

Number of Platforms Collected by Year.

Figure 2 shows, by state party, the share of bienniums (election cycles) for which we have either a copy of the adopted platform or a newspaper summary.Footnote ³⁹ The top map is for Democratic parties, and the bottom map is for Republicans. As the figure suggests, most states in both maps are darkly shaded, indicating that we have a great deal of coverage. The figure also reveals a key pattern: the same states tend to be darkly or lightly shaded in both maps. This reflects the empirical reality that, if one party in a state writes a platform, its competitor party is also very likely to do so.Footnote ⁴⁰ Finally, these maps reveal some regional differences. The states for which we have the lowest levels of platform coverage tend to be a combination of southern and mid-Atlantic states that constitute (or border) the Appalachian region.

Figure 2.

Share of Bienniums Covered by a Platform, 1960–2018. (a) Democrats. (b) Republicans.

6. Coding platforms

We hand-coded every platform in this study. We recognize that this was an unusual decision: it was tremendously time- and labor-intensive. Indeed, for all practical purposes, this approach would be impossible to employ for a study looking at every position and issue in a platform. Hopkins et al., for example, employed an approach that relies heavily on automated topic modeling; this allowed them to include every significant issue in their platforms, but not to analyze what positions were taken on these issues.Footnote ⁴¹ But our questions and hypotheses necessitated a different approach. In limiting our focus to abortion and LGBTQ rights, we were able to use hand-coding to identify and analyze the details of every pertinent position articulated in these 1,977 platforms. Following the method laid out in this section, we determined what general positions each platform took on each issue, then utilized IRT models to take full advantage of the seventy-four specific policy positions that we identified as pertaining to LGBTQ rights and abortion.

To code the platforms in our archive, we used OCR (optical character recognition) software to transform each platform into a machine-readable text. We then uploaded every platform into NVivo, a software package that is designed for qualitative data analysis. NVivo enables researchers to manually code and classify the text of uploaded documents.

We coded platforms in NVivo using two approaches. First, we carefully read over 900 platforms, representing twenty-three states, from start to finish. Using this manual approach, we identified every reference to abortion or LGBTQ rights, then highlighted the text and assigned it to a “node” based upon whether it was supportive, opposed, or, in rare cases, ambivalent. We next used this text to construct extensive data dictionaries of relevant words and phrases. The abortion and LGBTQ rights dictionaries, created based on our manual reading of platforms, each contains around 150 words and phrases. We then used these dictionaries to conduct automated searches of the remaining platforms. These searches directed us to the locations in every platform in which a dictionary word or phrase appears. Upon locating this potentially relevant text, we then reverted back to the first method—we read the plank, determined whether it in fact addresses abortion or LGBTQ rights, and, if so, we highlighted the text and assigned it a node based on whether it was in support, opposition, or neutral.

After completing this initial review, we carefully read through the text placed in each of the NVivo nodes, identifying the full set of abortion and LGBTQ-related policies that were addressed in state platforms. In total, we identified forty-three unique abortion policies. These include policies such as abortion legalization, parental and spousal consent, public funding of abortions for low-income women, and judicial and candidate litmus tests. For LGBTQ rights, we identified thirty-one distinct policies. Examples of these are the legalization of consensual sodomy, the extension of traditional civil rights protections to LGBTQ individuals, the inclusion of sexual orientation in hate crimes laws, marriage equality, and ensuring access to desired and necessary medical care for transgender individuals. Among our set of LGBTQ policies we have included ten that broadly pertain to HIV/AIDS. For each of the seventy-four identified policies, we assigned the platform a +1 for taking the liberal position, a zero for taking no position, or a −1 for taking the conservative position, assigning values of +1 or −1 only if the position was explicitly stated.Footnote ⁴² We list all 43 abortion policies in Appendix Table B1 and the 31 LGBTQ policies in Appendix Table B2.Footnote ⁴³

While most policies are coded across the full six decades of our analysis, a handful are not. One of these is abortion “liberalization.” Efforts to liberalize state-level bans on abortion during the 1960s and early 1970s called for exceptions to be written into state penal codes, allowing abortions but only when certain conditions were met—such as if the pregnancy was the result of rape or incest or if a team of doctors certified that the mother’s physical or mental health was at risk. Abortion liberalization disappeared as an issue after the Supreme Court’s decision in Roe v. Wade, though we recognize that this is returning as an issue in the post-Dobbs era. For this reason, we only code abortion liberalization from 1960 through 1972. Similarly, all of the HIV/AIDS policies are coded starting in 1984, when the issue first appears in a state party platform, through 1996. We stop coding HIV/AIDS policies after 1996, the year in which highly active antiretroviral therapy (HAART) became the standard of care. HAART dramatically reduced the number of AIDS-related deaths in the United States and thereby the political salience of HIV/AIDS in U.S. domestic politics.Footnote ⁴⁴ After 1996, the number of references to HIV/AIDS in state party platforms falls dramatically.

7. Measures of party policy positions

We use our dataset of coded platforms to create two measures of state party position-taking on abortion and LGBTQ rights. The first measure is discrete and straightforward: whether a platform is supportive, opposed, or silent for each issue. The second measure is continuous and more nuanced, taking full advantage of the seventy-four distinct policy areas that we have identified. For this second measure, we employ IRT models to estimate a platform’s “liberalness” in each of the two issue realms.

7.2. Continuous measure: Estimating platform “liberalness”

To measure the more nuanced differences in position-taking between platforms, we employ Bayesian IRT models. IRTs have become a common measurement method for ideal point estimation in political science.Footnote ⁴⁵ The approach we take here is similar to Treier and JackmanFootnote ⁴⁶; using our hand-coded state party platform data as inputs, we estimate a latent variable of platform liberalness in each year for which we obtained a platform. These are conceptually, though not procedurally, similar to DW-NOMINATE scores.

More formally, for each state $s$, party $p$, and year $t$, we model the latent “liberalness” of the platform’s planks in a given policy domain (${\theta _{spt}}$). Our model is a static IRT using an ordered logit link function. We assume that each party decides to address (or not address) each issue $j$ in their platform by taking the conservative position (${Y_{sptj}}\, = \, - 1$), taking the liberal position (${Y_{sptj}}\, = \,1$), or stating no position (${Y_{sptj}}\, = \,0$). We assume that a party will take a given position, according to the following decision rule:

\begin{equation*}{Y_{sptj}} = \begin{cases} { - 1} & {{\text{if}}\ {z_{sptj}} \leq {\kappa _1},} \\ 0& {{\text{if}}\ {\kappa _1} \lt {z_{sptj}} \leq {\kappa _2},} \\ 1& {{\text{if}}\ {\kappa _2} \lt {z_{sptj}}} \end{cases} \end{equation*}

where ${z_{sptj}}$ is the product of the latent variable ${\theta _{spt}}$ and a discrimination parameter ${\beta _j}$ for each policy, and ${\mathbf{k}}$ is a vector of cutpoints to be used in the ordered logistic distribution. We model these parameters with the following priors:

\begin{align*}{\theta _{spt}} & \sim \mathcal{N}(0,1)\\ {\beta _j} & \sim {\mathcal{N}_{[0,\infty )}}(0,3)\\ {{\kappa}_j} & \sim \mathcal{N}(0,3).\end{align*}

The discrimination parameter takes a half-normal distribution because we have manually coded the platform planks such that a higher value (+1) is always the more liberal policy. This avoids any “switching” problems and means that the ${\beta _j}$ term only affects the magnitude of the relationship between the latent variable and expressed policy for each issue, not the directionality of that relationship. Because of the ordered nature of the underlying, hand-coded data, we can also assume that ${\kappa _1} \lt {\kappa _2}$.

We chose a static model, rather than a dynamic one that explicitly models trends over time, because our goal was to capture the expressed positions of parties within the platform documents themselves. Dynamic models are useful to capture latent variables when the results in time $t$ are a function of those in the previous period.Footnote ⁴⁷ However, in our case, dynamic modeling has the effect of flattening dramatic changes in parties’ stated policy positions. To the extent that party platforms are a function of those that came before, this should be reflected within the documents themselves, not as a result of a modeling choice by the researchers.

We fit three versions of the model in Stan using three sets of policies: abortion issues, LGBTQ issues, and all issues across both domains.Footnote ⁴⁸ In each case, we included all coded platform planks in the relevant domain. The main output of interest for each model is a platform liberalness score, ${\theta _{spt}}$. This score can be interpreted as a summary of the relative liberalness of a given state party’s explicit platform positions on abortion, LGBTQ rights, or across both domains in a given year.Footnote ⁴⁹

The great advantage of these platform liberalness scores is obvious: these scores allow us to utilize fully the detailed coding of a wide array of policy positions, giving us the ability to discern differences among state parties that receive the same score on the discrete measure. While a large number of state parties express opposition to abortion, for example, each of these parties expresses a distinctive mix of positions that separates it from other parties on the same side of the policy divide. These scores allow us to distinguish between the liberalness of two platforms that both oppose abortion and to track the ways in which platforms continue to evolve even after they take their initial positions for or against abortion and for or against LGBTQ rights.

No method is flawless, however. In working with the IRT models, we are sensitive to the fact that our measures reflect (in part) the number of positions a platform takes. That has two implications. First, longer platforms have more space to take more positions. Consequently, our platform liberalness scores are a function, at least in part, of a platform’s length.Footnote ⁵⁰ Second, we recognize that many positions related to LGBTQ rights—marriage equality, for example, or the rights of trans people—did not exist in mainstream political discourse in earlier decades. Consequently, recent platforms, in taking positions on more issues than older platforms, will, for that reason, seem more extreme by this measure than platforms from the first decades in our series. This is primarily a concern in the LGBTQ realm, where the number and variety of issues has expanded over time. It is much less of a concern in the abortion realm, where the primary issues have remained stable. For this reason, we need to be somewhat cautious in comparing IRT scores from one period of time to those from another period of time, especially in the LGBTQ realm. But IRT scores are very effective in comparing contemporaneous platforms from different states.

One way to validate the IRT scores of platform liberalness is to compare them to the basic, discrete measures of whether a platform is supportive, opposed, or silent on abortion and LGBTQ issues. Table 1 reports the correlation between our IRT models and the discrete measures. The high correlations confirm that the modeled liberalism scores are consistent with the latent concept we seek to measure, even as they introduce greater variance by accounting for the relative importance of diverse policy positions on the parties’ social liberalness. Taken together, these two measures give us greater confidence than either measure on its own.

Table 1.

Validating IRT Scores

8. Position-taking: Abortion and LGBTQ rights, 1960–2018

Drawing on our discrete measure, we can now examine the timing and speed with which state parties began taking positions on abortion and LGBTQ rights. Figure 3 plots, by year, the percentage of state party platforms that support or oppose abortion (the solid line) and that support or oppose LGBTQ rights (the dashed line). At the beginning of the time series, no state political parties took positions on these issues: abortion does not appear in any state party platform until 1968 (when four state parties adopted planks supporting the liberalization, but not the outright repeal, of abortion bans) and LGBTQ rights do not appear until 1970 (when two state parties called for the repeal of laws restricting consensual sexual activity, likely a reference to sodomy bans). Remarkably, there is no critical moment in which large numbers of parties suddenly took sides in these debates—not in the 1970s, following the Roe v. Wade decision; not with Reagan’s election in 1980; not in 1992, during the “culture war” battles in the Clinton–Bush campaign. Rather, as Figure 3 shows, the process by which state parties staked out positions was incremental and unfolded gradually over four decades. Only in 2016 did all written platforms take positions on both issues. Figure 3 also makes it clear that state parties were slower to take positions on LGBTQ rights than they were on abortion. The rates of position-taking on these issues were not equal until 2004, but have remained roughly equivalent since then.

Figure 3.

Position-Taking by Issue and Year.

But what positions did the parties take? Figure 4 shows position-taking on abortion, and Figure 5 shows position-taking on LGBTQ issues. In each figure, the top panel illustrates the discrete measure and the bottom panel the continuous measure derived from the IRT models. Here we draw on the richness of our entire database of 1,977 state party platforms to document how each issue exploded into the partisan world. The vertical lines in each of the two upper panels represent the years in which the national party platforms first took a position on these issues; we code the national platforms by the same criteria we use to code all of the state party platforms.

Figure 4.

Abortion Rights Positioning by Party.

Figure 5.

LGBTQ Rights Positioning by Party.

In the upper panel of Figure 4, we display the proportion of platforms, by party and year, that support and oppose legalized abortion. The percentage that support abortion are shown above zero on the y-axis, while the percentage that oppose abortion are displayed below zero. As we can see in the figure, Democratic state parties began taking positions first, with approximately 25 percent of the state Democratic parties staking out pro-choice positions in both 1970 and 1972. In the four years following Roe v. Wade (1973), however, the share of platforms supporting abortion fell—probably due both to the emergence of a more vocal antiabortion movement and the abrupt recognition of abortion rights by the Supreme Court, which obviated the need to call for legalized abortion. Beginning in 1978, explicit Democratic support for abortion rights began steadily to increase. By the early 1990s, over 75 percent of Democratic platforms expressly supported abortion, and by 2016 this figure had risen to 100 percent.

The rate of Republican parties adopting the antiabortion position looks similar, though somewhat delayed relative to the Democratic side. In 1972, when 25 percent of Democratic parties went on the record as pro-choice, only the Utah Republican Party opposed abortion. Indeed, in 1972 there were the same number of antiabortion Democratic state parties—one (the Massachusetts Democrats)—as there were antiabortion Republican parties. But, two years later, in the wake of Roe v. Wade, increasing numbers of Republican state parties began to articulate opposition to abortion. In 1974, 10 percent of Republican platforms adopted pro-life planks. The number continued to grow slowly: only in 1982 did a full 25 percent of Republican parties take the pro-life position. Despite this initial delay, support for the antiabortion position on the Republican side increased steadily, reaching 75 percent by the end of the 1980s and 100 percent by 2012.

The upper panel of Figure 4 also shows a phenomenon that occurred in the early part of the time series—a small number of parties staking out what today we would see as unorthodox partisan stances. This occurred more frequently among Democrats. At various times during the 1970s, the state Democratic parties in Massachusetts, Missouri, Rhode Island, South Dakota, and Utah all adopted platforms opposed to abortion rights. While on the other side only the Iowa Republican Party (in 1972) embraced a full pro-choice position, state Republican parties in Hawaii, Michigan, Minnesota, Nevada, and New York adopted platforms in the late 1960s or early 1970s (like a handful of Democratic parties in those same years) that called for the liberalization of existing abortion bans. However, because none of these platforms called for the repeal of such bans, we do not code them as supporting abortion. Later in the time series, the Maine (1990), New York (1990), and Hawaii (2000) Republican parties all briefly adopted platforms supporting abortion rights.

Turning next to the bottom panel of Figure 4, we see our continuous measure of platform liberalism, drawing on the forty-three distinct issue positions related to abortion. The y-axis here is the liberalism score. Platforms that adopt a strong set of pro-choice positions score high on this liberalism score, platforms that embrace a wide range of pro-life positions score low, and platforms that take no position score zero. What we see in this figure are averages across all Democratic parties and all Republican parties, so some details, such as the rogue abortion platforms of the 1970s, are obscured. But what is revealed, in high relief, is a general trend that mirrors that in the upper panel. The two parties, measured by their state platforms, are indistinguishable in the 1960s. They begin to separate in 1970, and by 1978 the differences in the average platform liberalness score between the two parties become statistically meaningful. Since then, those differences have grown incrementally into a wide gulf. As both panels in Figure 4 show, the change over time, in both parties, has been steady but inexorable. The state parties are more divided today on abortion than at any other time in history.

Figure 5 displays similar data regarding position-taking on LGBTQ rights. In the upper panel, drawing on the discrete measure, the percentage of platforms that support LGBTQ rights are shown above zero on the y-axis, while the percentage that oppose them are displayed below zero. In the lower panel, we graph the liberalism score, derived from the IRT models. This liberalism score shows that the two parties became meaningfully different on LGBTQ issues in 1980, just two years after the same had happened on abortion. In both the abortion and LGBTQ realms, as shown in Figures 4 and 5, this separation in 1978–80 appears to be driven largely by Democratic, not Republican, position-taking.

Democratic parties moved first when it came to taking positions on LGBTQ rights, just as they did with abortion. As both figures show, Democratic platforms began taking pro-LGBTQ positions in the 1970s and, as the upper panel shows, by 1980, a full 25 percent of state Democratic parties had adopted a platform that favored LGBTQ individuals. By 2000, nearly 75 percent of Democratic platforms had planks supporting LGBTQ rights, and by 2012 this figure had risen to 100 percent. As with the issue of abortion, the share of Democratic parties supporting LGBTQ rights increased incrementally and gradually over time: that is evident in the first panel. There was no critical moment of change. But, as the lower panel shows, the substance of Democratic support for LGBTQ rights did undergo a dramatic shift beginning soon after 2000. While the number of state parties supporting LGBTQ rights grew steadily over time, the number of positions embraced by these parties has expanded sharply over the last twenty years. Democrats, in short, have grown much more liberal in the range of policies they endorse. What began as a simple affirmation of gay and lesbian rights has, over time, become for state Democrats, a broad assertion of support for multiple issue positions favoring the LGBTQ community, above all robust and widespread advocacy for the rights of transgender individuals and for same-gender marriage.

The earliest Republican platforms to discuss homosexuality did not do so until 1978, and the share of Republican platforms expressing clear opposition to LGBTQ rights grew modestly for the next decade. As late as 1986, in dramatic contrast to the Democrats, just 15 percent of Republican platforms had taken a position on the issue. Between 1988 and 1994, the number fluctuated between 18 and 39 percent. Then, in 1996, the share of Republican platforms opposing LGBTQ rights jumped to 63 percent and remained at that level until 2004, when it increased to 84 percent. Since 2004, as the upper panel of Figure 5 demonstrates, opposition to LGBTQ rights in Republican platforms has remained constant. Notably, since 2014 a small number of Republican platforms have begun taking positions in support of LGBTQ rights, calling for sexual orientation to be added to the traditional list of attributes (e.g., race, sex, and religion) that cannot be the basis of discrimination. The only Democratic platforms to ever take an anti-LGBTQ position were a handful from the early 2000s that opposed same-sex marriage.

As the lower panel of Figure 5 suggests, illustrating the liberalness score derived from the IRT model, the mix of positions adopted by state Republicans closely mirrored the proportion of their platforms supporting or opposing LGBTQ rights. According to this continuous measure, Republican platforms became much more hostile to LGBTQ rights between 1980 and 1996. This was a function of Republican state parties taking positions on new issues, like same-gender marriage, as well as more expansive language addressing older issues, like the right of LGBTQ adults to adopt children, LGBTQ civil rights protections (or “special privileges”), and restrictions on what can be taught in schools relating to LGBTQ issues. Note that, by this measure, a party looks more extreme if it takes ideologically consistent positions on a larger range of issues, which is the case here. After 2004, by this measure, Republican platforms did not grow any more conservative. On the continuous, liberalism score, Republican platforms stayed constant between 2004 and 2018—though we speculate that the increasing prominence of issues relating to trans rights may lead to a new conservative shift in Republican platforms in the 2020s, as they begin taking positions on this set of issues.

9. Party polarization, 1960–2018

Next, in Figure 6, we consider the extent and speed with which state party systems have polarized. The units of analysis here are state-level party pairs (e.g., the California Democratic and Republican parties) in a given year. Using our discrete measure, we code a pair of parties as being polarized if they take opposing positions on an issue. As one can see, polarization of party pairs is virtually nonexistent in the 1960s and 1970s. From there, however, it increases in linear fashion across the full time series, from 1980 until the mid-2010s, when this measure approaches 100 percent. As we have observed previously, there does not appear to be a critical moment or election cycle that is driving this polarization; the sharpest sustained increase in the share of party systems that polarized occurred between 1986 and 1992. We also observe in Figure 6 that party pairs have been more frequently divided on the issue of abortion than LGBTQ rights. In fact, polarization on LGBTQ rights declined in the 2010s as some state Republican parties dropped hostile planks or, in a handful of cases, supported the inclusion of sexual orientation and/or gender identity in traditional civil rights protections.

Figure 6.

Polarization of State Party Pairs.

With Figure 7, we present platform liberalness scores for a combined IRT model, using party positions across both issue areas, rather than continuing to report liberalness scores separately for LGBTQ and abortion positions. The top panel of Figure 7 tracks the average platform liberalness score among all state Democratic parties (the blue line) and the average score among all state Republican parties (the red line) across the full period 1960–2018. As the figure shows, and as we have already seen for the issues separately, there is no difference between the parties near the beginning of the time series. Consistent with our separate findings on abortion and LGBTQ positions, meaningful differences between the parties emerge only in 1978. Over subsequent years, Democratic platforms became increasingly liberal, while Republican platforms became increasingly conservative.

Figure 7.

Combined Platform Liberalness Scores, 1960–2018.

The bottom panel of Figure 7 tracks the platform liberalness scores of the Democratic and Republican parties in a sample of six states—California, Massachusetts, Minnesota, Texas, West Virginia, and Wyoming. Gaps in the time series for any state party mean that we are either missing the platform for that year or no platform was written. In general, the patterns in the individual states are similar to those observed across all fifty states. But there are noteworthy differences in the liberalism scores of individual parties and in the speed and timing of polarization across state party systems. In California, Minnesota, and Texas, the liberalness scores of the two parties become meaningfully different in 1976–80, mirroring patterns across the whole universe of states, and in these three states the distance between the parties becomes very large, indicating high levels of partisan polarization. The story diverges in the other three states. In Massachusetts, West Virginia, and Wyoming, the state parties grow notably distinct only in the 1980s and 1990s, and the parties in these states are less polarized, even today, than those in the first three states. The differences between the two parties’ liberalism scores in the twenty-first century remain lower in Massachusetts, West Virginia, and Wyoming than in California, Minnesota, and Texas. In some cases, state parties remain relatively constant in their scores for long periods of time, and, in other cases, as the lower panel of Figure 7 suggests, there are periods when parties move toward greater liberalism or conservatism.

The differences between the blue and red lines represent the ideological polarization between the two parties on these issues. Figure 8 plots this difference for our entire data set, drawing on platforms from all states and all years. As one can see, there is a steady and steep increase in polarization from 1978 through 1998. After 1998, polarization continues to grow, but at a slower pace. Understanding the dynamics of this 1978–98 period is crucial to explaining the sources of contemporary partisan polarization. As Figure 8 shows, there are no dramatic inflection points in this era. Polarization instead grew steadily, ineluctably, from the late 1970s through the 1990s. As evidence from state party positioning on abortion and LGBTQ rights demonstrates, the American party system entered the twenty-first century utterly transformed.

Figure 8.

Platform Liberalism Polarization, 1960–2018.

10. Conclusion

In the summer of 2022, just days after the U.S. Supreme Court announced its decision in Dobbs v. Jackson Women’s Health Organization, the Texas Republican Party, at its convention in Houston, adopted a platform brimming with conservative positions on culture war issues. Delegates urged legislators to “abolish abortion” and demanded that the Fourteenth Amendment’s guarantee of equal protection of the laws be applied to “all preborn children from the moment of fertilization.” The platform also described homosexuality as “an abnormal lifestyle choice” and declared its opposition to ‘‘homosexual marriage” and to “all efforts to validate transgender identity.” The platform adopted by the Democrats stands in stark contrast. The Texas Democratic Party in 2022 affirmed its commitment to “equal protections and rights for all lesbian, gay, bisexual, transgender, queer, intersex, asexual, and other non-cis non-heterosexual or LGBTQIA+ people” and it also sought to “protect the constitutional right to an abortion.”

The degree of partisan polarization in 2024 on abortion and LGBTQ rights is without historical precedent. Fifty years ago, party leaders in states across the nation were only beginning to take positions on social issues, and there were no meaningful differences in party positions. The changes over this half-century have been explosive. As we show in this paper, the culture war, which until the 1970s hardly existed, has now come to every state in the union. Astonishingly, every state party drafting a platform in the 2010s had staked out a clear position on abortion, and Democratic and Republican positions, in state after state, were sharply at odds.

The vast state party platform database that we have assembled, and which we draw on for the first time in this paper, allows us to test multiple hypotheses regarding the origins and development of the culture war—and, more broadly, the polarization of the nation’s political parties. By hand-coding these platforms, then employing methods that yield discrete and continuous measures of issue positions, we find, contrary to nearly all preexisting scholarship, that polarization on the issues of abortion and LGBTQ rights occurred gradually. There was no sharp change in either the late 1970s or the 1990s—or at any other time. Rather, partisan polarization on culture war issues was gradual. While we find that the parties began to differ in meaningful ways on these issues in 1978–80, this was not a dramatic shift, but rather a point on a gentle slope.

State parties drove the process of polarization. By the time the national parties took positions on abortion and LGBTQ rights, parties in a significant number of the states had already acted. Democrats and liberals moved first. Given the strongly conservative status quo ante of the 1960s, when abortion and homosexuality were effectively outlawed in every state, it is not surprising to find that it was liberals who began to raise these issues in the political arena. But the extent to which the national culture war was rooted in state politics presents an important new perspective on the era. As was the case with race,Footnote ⁵³ we show that the issues of abortion and LGBTQ rights first emerged in partisan political conflict at the state level before spreading to the national party agenda.

We live in an era defined by partisan polarization. The degree of polarization differs in intensity from state to state, with some state party systems, like that in Texas, severely polarized, and others much less so. The politics of LGBTQ rights and abortion have never before been so central to our national political discourse and our private disagreements. Today’s Americans inhabit a party universe that is entirely foreign from that of the 1960s and 1970s, and that universe took shape, through the work of activists and party leaders, in state party platforms.