How do you measure a constitutional moment? Using algorithmic topic modeling to evaluate Bruce Ackerman's theory of constitutional change.

• Author: Young, Daniel Taylor
• Position: Introduction into II. Results A. Hypothesis 1: Evidence of Constitutional Discourse 2. The 1868 Election, p. 1990-2030

INTRODUCTION I. QUANTIFYING CONSTITUTIONAL POLITICS A. Ackerman's Theory of Dualist Democracy B. The Importance of Popular Attention C. A Test Case: Ratification of the Fourteenth Amendment D. The Challenge of Measurement E. A New Approach II. RESEARCH DESIGN A. Quantifying Public Discourse B. The Use of Newspapers as a Proxy for Political Attention C. Methodology D. Data III. RESULTS A. Hypothesis 1: Evidence of Constitutional Discourse 1. The 1866 Election 2. The 1868 Election 3. Parsing Out Constitutional Topics 4. Hierarchical Modeling and the Structure of Discourse, 1866-1868 B. Hypothesis 2: Comparing Normal and Constitutional Politics, 1866-1884 CONCLUSION INTRODUCTION

When considering the long arc of American constitutional history, most legal historians recognize that constitutional change sometimes occurs outside the boundaries of Article V's formal amendment process. Three moments, in particular, have become touchstones of legal scholarship: (1) the ratification of the Fourteenth Amendment, purporting to comply with the requirements of Article V but in reality occurring under armed occupation as a precondition for Southern states' readmission to the Union; (2) the juridical revolution ushered in by the New Deal; and (3) the legal transformation wrought by the civil rights movement. Each of these periods of constitutional upheaval featured powerful coalitions purporting to represent "We the People." The battle over Reconstruction pitted Radical Republicans against President Johnson and recalcitrant Democrats; (1) the New Deal involved a showdown between President Roosevelt and the "Four Horsemen" of the Supreme Court; (2) and the civil rights movement pitted activists against the apartheid South. (3) If we accept the premise that constitutional change occurs outside the Article V amendment process, then we need a way to evaluate the claims of competing constitutional visions during times of political tumult. How are we to weigh the legitimacy of claims to speak for We the People?

Bruce Ackerman's theory of constitutional change is an attempt to systematize an answer to that question. Ackerman's initial move is to posit that the United States has a "dualist democracy," meaning that it operates in two distinct modes--a "normal politics" mode and a "higher lawmaking" mode. (4) The dualist-democracy thesis starts from the premise that most of the time American citizens do not grapple with higher-order questions of constitutional law. As Ackerman puts it, one half of the "cyclical pattern" of American history "is characterized by normal politics, during which most citizens keep a relatively disengaged eye on the to-and-fro in Washington." (5) During these periods of normal politics, those who advocate for transformation are "regularly rebuffed at the polls in favor of politics-as-usual." (6) On rare occasions, however, debates that occur among governing elites during times of "normal" politics spill into the national discourse and catch the attention of the American people. Ackerman argues that during these periods of constitutional transformation key elections push constitutional change forward. On Ackerman's telling, certain elections legitimate constitutional change because voters, in contrast to their behavior during times of normal politics, pay special attention to the constitutional questions on the national agenda. (7) Critically, then, Ackerman's argument about popular sovereignty hinges on voter attention. Its principal claim--that voters pay particular attention to constitutional issues during periods of higher lawmaking--should be amenable to empirical scrutiny. But what variable to measure? And how? Pure electoral returns are silent on the question of why individuals voted or what issues were particularly salient for them at the ballot box. Public opinion polling is also problematic (8) and, in any event, is not available to help us evaluate constitutional politics during earlier periods of American history, such as Reconstruction. In light of these challenges, how can we measure the national "constitutional dialogue" to determine if Ackerman's key moments really were different from normal political debate? To put it another way: can we quantify constitutional discourse?

This Note proposes that we can. I worked with Brandon Stewart, a Ph.D. candidate in Government at Harvard University focusing on statistical methodology, to apply a series of algorithmic topic models to study historical newspapers published between 1866 and 1884. (9) Topic modeling is a process of machine learning that uses a statistical algorithm to analyze a body of text--the "corpus"--by grouping together words that have a high probability of appearing together in the documents. (10) I deploy these models to test the validity of constitutional moments theory by looking for spikes in particular kinds of constitutional dialogue when Ackerman's theory predicts they should arise.

This Note uses Reconstruction as its test case, focusing in particular on the debate over ratification of the Fourteenth Amendment during the years 1866 to 1868. I rely on four datasets containing over 19,000 pages from U.S. historical newspapers published between January 1, 1866, and December 31, 1884. I find empirical support for the proposition that newspapers did in fact focus to an unusually high degree on constitutional-level questions during the critical period between 1866 and 1868. These findings lend support to Ackerman's conception of "constitutional politics" during the debate over ratification of the Fourteenth Amendment.

Part I of this Note begins by explicating Ackerman's dualist-democracy thesis and outlining how it applies to ratification of the Fourteenth Amendment. I argue that Ackerman's theory rests on the notion that voters were paying special attention to constitutional debates in the lead up to the 1866 and 1868 elections, which in turn legitimizes (on Ackerman's telling) the constitutional changes that these elections helped ratify. I survey the literature on "constitutional moments" in order to demonstrate how the lack of empirical rigor in this area has fueled an ongoing debate about what "counts" as a moment of significant constitutional change. I then explain how topic modeling can address this challenge by evaluating the popular salience of constitutional issues during this period. This Part ends by justifying the choice of Ackerman's treatment of the Fourteenth Amendment as a test case.

Part II explains this Note's methodology. This Part begins by exploring how topic modeling distills the structure of enormous amounts of text. I then seek to justify this Note's use of historical newspapers as a proxy for public attention, responding to potential objections about whether newspapers truly capture issue salience among voters. This Part concludes by explaining how Brandon Stewart and I converted the raw text from newspaper articles into usable data for our analysis.

Part III provides the results of our topic models. In particular, this Note examines two hypotheses. Section III.A explores whether there is evidence of constitutional politics in the period from 1866 to 1868 and concludes that there is evidence of constitutional discourse before the 1866 and 1868 elections. I also use a technique known as hierarchical topic modeling to illustrate how newspapers discussed various issues. Section III.B then examines change over time, seeking to discern whether the salience of constitutional politics declined in the period from 1866-1884. I find evidence that constitutional discourse peaked between 1866 and 1868 and then gradually declined, lending support to Ackerman's narrative about a gradual return to "normal politics" during this period.

This Note concludes by surveying these results and arguing that there is indeed empirical evidence for the high salience of constitutional issues during this period. This provides quantitative support for Ackerman's dualist-democracy thesis. It also illustrates that topic modeling is an innovative research tool ripe for further applications in legal scholarship.

1. QUANTIFYING CONSTITUTIONAL POLITICS

   Attempts to distill American constitutional history into a set of key moments are a common feature of legal scholarship. (11) Such efforts, of which Bruce Ackerman's dualist-democracy thesis is arguably the most prominent, raise a consistent and vexing methodological challenge: How can we develop objective criteria to distinguish "constitutional moments" from "normal politics"?

   This Part begins by explicating Bruce Ackerman's theory of constitutional change and exploring its application to the ratification of the Fourteenth Amendment. I argue that Ackerman's theory rests on a critical but untested argument regarding the high salience of constitutional issues before certain key elections, including the elections of 1866 and 1868. I then survey other scholarship in this area to demonstrate how the lack of empirical metrics about what "counts" as a constitutional moment is a continuing problem. Finally, I argue that topic modeling provides us with a way to address this challenge, using Ackerman's arguments about the Fourteenth Amendment as a test case.

   1. Ackerman's Theory of Dualist Democracy

      Bruce Ackerman's dualist-democracy thesis is one of the most prominent attempts to systematize a theory of constitutional change. In its scope and ambition, Ackerman's theory has left a lasting impression on the legal academy. (12)

      Ackerman's theory is, at bottom, a framework for understanding how constitutional change occurs outside the boundaries of the formal Article V amendment process. When evaluating how the U.S. Constitution changes over time, legal scholars generally begin by making one of two analytic moves. The first is to look at the text of the Constitution itself, and in particular Article V. According to this view, the American people change the Constitution whenever a would-be...