The Justices’ Words: The Relationship between Majority and Separate Opinions

Abstract

Majority and separate opinions reflect the justices’ deliberations and strategic decision-making. As justices try to shape the legal outcome, private disagreements during the opinion-writing process spill out into the open, becoming the written words of majority and separate opinions. In this article, I ask how justices use separate opinions to shape the law. I argue that the length of an opinion provides a reasonable proxy of the law and the Court’s decision-making at work. Using time series techniques on the number of words in majority and separate opinions from 1953–2009, I examine whether there is a relationship between the number of words in majority and separate opinions. I demonstrate there is a fractional cointegration relationship between majority and separate opinion length. The majority and separate opinion relationship means there will not be a time in which the Court produces incredibly long separate opinions and succinct majority opinions, or lengthy majority opinions and brief separate opinions. I also find that separate opinion length causes the majority opinion to be shorter or longer. Error correction model results indicate that discussions that occur in one term do not conclude when the Court’s term ends, the effects continue in subsequent terms and cases. The law, as the Court generates it in its majority opinions, is shaped by separate opinions.

Keywords:

• Supreme Court
• opinion writing
• judicial decision-making
• dissenting opinions

Acknowledgements

The author would like to thank Michael Fix, Rachael Hinkle, Matthew Hitt, Sojeong Lee, and attendees of 2018 Midwest Political Science Association conference, as well as the editor and anonymous referees for their helpful comments and suggestions. The article is better for their guidance.

Notes

1 I am not arguing that dissents and concurrences are the same—because they are not—but rather that their implications are. Dissents and concurrences are different forms of conflict over outcomes and policy (Caldeira and Zorn 1998). Separate opinions offer justices a vehicle to voice their disagreement with something the majority opinion said, and in the long-term, perhaps offer future justices’ additional information.

2 Per curiam opinions may affect the final aggregated distribution and since the data is aggregated, it may hide some justice-level dynamics.

3 The median majority opinion has 2,869 words and a standard deviation of 691 words. The median separate opinion has 1,368 words and a standard deviation of 667 words.

4 I also tested whether the residuals were white noise. OLS assumes that variance of error term is constant, i.e. homoskedastic or no serial correlation, over time. Applying the Portmanteau test for white noise, I fail to reject the null of no serial correlation at the 0.05 level and conclude that the residuals for both series are white noise.

5 Box-Steffensmeier and Tomlinson (2000) state that so long as d is within one standard deviation, they can be treated as the same I(d). Moreover, when a constraint is applied, d = 0.493 (standard error 0.068) for both series.

6 For majority opinions, d = 0.490 and the standard error is 0.010. For the separate opinions, d = 0.469 and the standard error is 0.028.

7 I also estimated the d parameter of the residuals using an ARFIMA model. For majority opinion residuals, d = 0.208 and the standard error is 0.093. For separate opinion residuals, where d = 0.039 and the standard error is 0.129.

8 To verify the finding, I also estimated the d parameter of the residuals using an ARFIMA model. For majority opinions with chief justices, d = 0.263 and the standard error is 0.153. For separate opinions with chief justices, d = –0.068 and the standard error is 0.125.

9 Because the Court does not hear interstate relation cases every year, it had too many gaps in the data to analyze.

10 There are likely many exogenous independent variables that affect the length of majority and separate opinions. Black and Spriggs (2008) have documented several of these variables on majority opinion length. For the purposes of this article, I focus exclusively on the cointegration relationship and leave further exploration of these variables’ effect for future work.

11 Lag length for all models was determined using the Akaike Information Criterion.

12 In the error correction model, I use the residuals from the cointegrating regressions to estimate the error correction variable.