Measuring Political Positions from Legislative Speech

Abstract

Table A1 provides an overview of the speeches and debates included in our analysis. In an average electoral period, there were 8,707 speeches across 779 debates in the Irish Dáil and 7,464 speeches across 584 debates in the US Senate. The 112th and 113th Senates stick out at as the two least productive Senates in terms of number of speeches and debates, which is most likely the result of legislative gridlock after the Republican party took over the House in the 2010 election.The average debate in our data set consists of 12 speakers, ranging from a minimum of 5 speakers (our lower threshold) to as many as 73 speakers. At the level of individual speeches, we find an average of 683 and 552 words per speech and legislature. The longest individual speech in our data set is 37,610 words long, which is senator Ted Cruz's (TX-R) 21-hour filibuster speech in September 2013. Sources Additional Analysis of Classification and Uncertainty versus WordfishIn the main text, we show plots of the relationships between individual TDs Wordshoal score and coalition membership, and uncertainty estimates for the 30th Dáil. In Figure A1 we show the corresponding plots for the 29th Dáil. Figure A1: The association between the estimated positions of each legislator and their status as members of the coalition versus opposition, with correlation and local linear smooth, under Wordfish (left) and our approach (right), for the 29th Dáil. In the bottom row, we show the 95% intervals associated with the estimates for each legislator under Wordfish (left) and Wordshoal (right). A5 Dail 29Type of debate Beta root mean square Analysis of Variation in Disagreement by Legislative StageThe Wordshoal model facilitates assessments of the relative extent to which different kinds of debates align with the estimated common dimension. The greater the magnitude of the β j , the greater the extent to which the common dimension predicts speech variation in a given debate. As our summary statistic for β j , we use the root mean square of the β j , weighted by the number of speeches in each debate speeches j .In the paper, we utilize this statistic only for the US Senate, here we assess it across different types of parliamentary debates in Ireland. The majority of debates in our data take place during the second reading of a bill, which is the most important legislative stage after which a bill is formally accepted or rejected. The second most common type are motions, which are an instrument of parliament to scrutinize the work of the government. This includes ad hoc motions on topical issues, seasonal adjournment debates, and (less frequently) A6 motions of confidence in the government or in individual cabinet members (Gallagher 2010). Figure A2 shows the weighted root mean square of β j for motions, second-stage bills, and for debates during the remaining legislative stages, which we group together because of the small number of observations at each stage. We find high government-opposition division during debates on motions, which are mostly used by...

Measuring Group Differences in High‐Dimensional Choices: Method and Application to Congressional Speech

“…Peterson and Spirling (2018) studied trends in the partisanship of speech in the UK House of Commons. In contrast to Lauderdale and Herzog's (2016) analysis (and ours), Peterson and Spirling (2018) did not specify a generative model of speech. Instead, Peterson and Spirling (2018) measured partisanship using the predictive accuracy of several machine-learning algorithms.…”

“…We apply a different method that addresses finite‐sample bias and leads to substantially different conclusions. Lauderdale and Herzog () specified a generative hierarchical model of floor debates and estimated the model on speech data from the Irish Dail and the U.S. Senate. Studying the U.S. Senate from 1995 to 2014, they found that party differences in speech have increased faster than party differences in roll‐call voting.…”

Measuring Group Differences in High-Dimensional Choices: Method and Application to Congressional Speech

“…Peterson and Spirling (2018) study trends in the partisanship of speech in the UK House of Commons. In contrast to Lauderdale and Herzog's (2016) analysis (and ours), Peterson and Spirling (2018) do not specify a generative model of speech. Instead, Peterson and Spirling (2018) measure partisanship using the predictive accuracy of several machine-learning algorithms.…”

“…We apply a different method that addresses finite-sample bias and leads to substantially different conclusions. Lauderdale and Herzog (2016) specify a generative hierarchical model of floor debates and estimate the model on speech data from the Irish Dail and the US Senate. Studying the US Senate from 1995 to 2014, they find that party differences in speech have increased faster than party differences in roll-call voting.…”

The Talk of Norway: a richly annotated corpus of the Norwegian parliament, 1998–2016

“…Clustering and other unsupervised modeling techniques have become a staple of this kind of research. Notable examples in recent years include Eggers and Spirling (2014), who show that the level of conflict in the electoral districts of a given member of parliament (MP) is important for her participation in both voting and speechmaking; Bäck and Debus (2016), who use the Wordscores technique (Laver et al 2003) to explore what causes MPs to participate more or less actively in parliament and why they sometimes deviate from the party line; Lauderdale and Herzog (2016), who demonstrate that a hierarchical approach to the Wordfish algorithm (Slapin and Proksch 2008) greatly improves its quality when applied to parliamentary speeches; and Proksch and Slapin (2015), who study parliamentary speeches from the UK, Germany, and New Zealand, showing that backbencher MPs deviate more from their party line in majoritarian than proportional representation electoral systems.…”