Evan Katz

*This post is a modified version of my POLS 2000 research paper, How Divided Government Affects Legislative Output.

The two previous U.S. Congresses—the 112th, serving from 2011 to 2012, and the 113th, serving from 2013 to 2014—passed fewer bills than any other Congresses in recent memory, earning the pejorative moniker of “least productive ever.” The numbers are astounding: while the 106th Congress passed 580 total bills between 1999 and 2000, the 112th saw a meager 283 bills pass, including only 208 substantive pieces of legislation, with the 113th producing slightly more, at 296 total bills (Quinn, 2014, n. pag.). Among the many similarities between the 112th and 113th Congresses, perhaps the most pertinent lies in the composition of both the Senate and the House of Representatives: between January 2011 and January 2015, Democrats and Republicans split control of Congress, with Democrats controlling the Senate and Republicans controlling the House. During the 106th Congress, Republicans controlled both chambers.

In line with this observation, the central question I’ll be addressing proceeds as follows: to what extent does the phenomenon of divided government—i.e. a government wherein the majority party or party in control of one of the three bodies active in the legislative process, the Senate, the House, and the President, differs from the party that controls the other two—affect the amount of legislation passed during a two-year session of Congress? When looking at previous Congresses, it seems gridlock always manages to take hold in instances of divided government, with polarization and nasty partisanship crushing the legislature’s ability to output legislation. However, this post will go beyond anecdotal evidence; I will employ a time-series observational research design to illustrate the relationship between divided government and legislative output. I hypothesize that, if the bodies of the government active in the legislative process are divided between two parties during a particular session of Congress, legislative output by that Congress will be lower than that of governments under the full control of a single party.

Understanding the way party composition of both chambers affects legislative output affords us important insight into the legislative process, including the kinds of items on a legislative agenda that will produce gridlock, and inversely, which items can pass without much partisan rancor. With public approval of Congress at all-time lows, new strategies for dealing with divided government and polarization would help to engender public trust in the institution, ensuring the stability of our democracy for posterity (Hughes & Carlson, 2015, p. 773).

Literature Review

Among the most groundbreaking of works written on the impact of divided government on policy is Yale political scientist David Mayhew’s Divided We Govern, published in 1991. Before Mayhew’s work, most political scientists believed that governments under the control of a single party operate more prolifically—and thus more effectively—than those under the control of multiple parties. Mayhew challenged this view by collating sets of data pertaining to passage of legislation from 1947 to 1990 and found that divided governments do not produce significantly less legislation than their unified counterparts (Saeki, 2009, pp. 587-588).

In 2009, Manabu Saeki, political scientist at Jacksonville State University, used Mayhew’s work and databases to argue that, while divided government does not play a noticeable role in slowing the rate of legislative output, the ideological leanings of key veto players in Congress, actors that must agree to change the legislative status quo, do (Saeki, 2009, p. 590). In the legislative process, four different veto, or pivotal, players determine whether or not the legislative status quo changes. The first, the median voter, has half of Congress ideologically to his/her right and half to his/her left. Second is the president, who has the power to veto legislation. The third, the override pivot, has two-thirds of Congress—the number of votes needed to override a presidential veto—to his/her right and one-third of Congress to his/her left in a right-leaning Congress. Finally, the filibuster pivot has three-fifths of Congress—the number of votes necessary to filibuster a bill—to his/her right and two-fifths to his/her left in a right-leaning Congress (Cost, 2009, n. pag.).

Screen Shot 2016-05-01 at 4.50.55 PM.png
Figure 1: Indifference Curve of Veto Player, P (Saeki, 2009, p. 591)

Saeki posits that, when plotting each veto player’s ideal policy points in two-dimensional space, with economic conservatism on one axis and social conservatism on the other, and assuming each veto player has single-peaked preferences, there will exist an indifference curve centered around the veto player’s ideal point and extending to the legislative status quo wherein he/she will prefer any policy to the status quo (Saeki, 2009, p. 591). Figure 1 above, excerpted from Saeki’s work, shows that the status quo, Q, lies at the coordinates (-0.25, 0.20), and that the ideal point of veto player P lies at the coordinates (0.25, -0.01). In a hypothetical scenario where P votes on a bill at point Q1, P would vote in favor of Q1 because point Q1 lies within P’s indifference curve—i.e. lies closer to P’s ideal point than the status quo. However, in a similar hypothetical scenario involving a bill at Q2, which lies beyond P’s indifference curve, P would vote to maintain the status quo because the status quo lies closer to P’s ideal point than Q2 (Saeki, 2009, pp. 591-592).

Screen Shot 2016-05-01 at 6.13.22 PM
Figure 2: Winset Among Veto Players in the 104th Congress (Saeki, 2009, p. 592)

When plotting the status quo, the ideal points of each veto player in Congress, and their indifference curves in two-dimensional space, there exists an area of overlap—in essence, a shared indifference curve—wherein any policy reflective of that area would be preferable to the status quo for all veto players, as shown in Figure 2. Saeki denominates this area the winset because all policies that fall within the area should become legislative “wins.” Using the winset, Saeki hypothesizes that legislative output is a function of the size of the winset during a particular Congress; gridlock sets in when the winset is relatively small, indicating a higher degree of polarization amongst the veto players because individual preferences are more ideologically scattered, whereas a large winset begets passage of numerous bills (Saeki, 2009, pp. 588, 592-593).

Political scientists Tyler Hughes of California State University, Northridge and Deven Carlson of the University of Oklahoma argue that, in spite of Mayhew’s findings, divided government significantly impedes the pace of legislative output, even if it does not reduce output altogether, resulting in legislative delay on crucial issues (Hughes & Carlson, 2015, p. 772). Specifically, Hughes and Carlson point to methodological flaws in Mayhew’s findings, notably including his failure to account for failed legislation, and shift the focus on the impact of divided government from the aggregate number of bills passed to the proportion of bills passed by a particular Congress (Hughes & Carlson, 2015, p. 774). Hughes and Carlson cite the Budget Control Act of 2011 as an example of the delay that results from divided government; despite its ultimate passage, the act monopolized congressional debate for months, requiring numerous bills, addenda, and votes within both chambers and distracting lawmakers from focusing on other priorities (Hughes & Carlson, 2015, p. 772).

Theory Development and Method

The works of Mayhew, Saeki, and Hughes and Carlson highlight a lack of consensus on the question of divided government impacting legislative output, with each introducing different—and in some cases contradictory—viewpoints and hypotheses. To further investigate the relationship between divided government and legislative output, I will conduct a study that analyzes levels of legislative output dating back to the 93rd Congress. This section discusses in depth the rationale behind the hypothesis advanced in the introduction and the methodology employed in proving that hypothesis.

combined-control_of_the_u-s-_house_of_representatives_-_control_of_the_u-s-_senate
Figure 3: Party Control of the House of Representatives, Senate, and White House, 1855-Present

The independent variable of this study, the party composition of both chambers of Congress during a particular session of Congress, operates as a categorical variable that can be expressed as a Boolean data type—i.e. either a single party controls both chambers of Congress, which would constitute a unified government, or not, which would constitute a divided government. The historical data cited for the independent variable comes from the work of Kathy Gill (2015), who constructed a table illustrating the party divisions of each government dating back to 1945 (see Figure 3). For the purposes of this study, I will ignore the role of the president in the legislative process because the focus of this paper is on the ability of the legislature to pass bills on to the executive, not the number of bills signed into law by the president.

The dependent variable of this study, legislative output, presents measurement challenges. Because legislative output is not unidimensional, it can be difficult to operationalize, especially because the literature base on the relationship between divided government and legislative output has not arrived at a consensus as to how to operationalize it. For the purposes of this study, I will use data provided by GovTrack (2016) on historical legislative output, which includes the amount of enacted legislation, passed resolutions, bills on which at least one chamber held a significant vote, failed legislation, vetoed bills that Congress did not override, and other legislation that did not reach the floor.

I will conduct two separate series of measurements: one using the raw number of enacted laws in each Congress from the 93rd to the 113th—ignoring the 114th because it is currently still in progress—and another using the number of enacted laws as a percentage of the total legislation introduced during that Congress. The latter measurement is necessary because not all Congresses introduce the same amount of legislation. For example, while the 93rd Congress passed 772 bills, a large number in absolute terms, it also saw 26,222 total bills introduced. Alternatively, while the 104th Congress only passed 337 bills, a small number in absolute terms, it also only saw 7,991 bills introduced, less than one-third of that of the 93rd Congress (GovTrack, 2016, n. pag.). Under the first measurement, it would appear as though the 93rd Congress was far more prolific than the 104th, but when expressed as a percentage of total legislation, the 104th (4.22% of total legislation enacted) enjoyed far more legislative success than the 93rd (2.94%).

For the first series of measurements, I will calculate the mean number of enacted laws under conditions of unified and divided government, and will compare them by expressing the number of laws enacted under divided government as a percentage of the number of laws enacted under unified government. I will then repeat the process for the second series of measurements. Additionally, I will show a normal distribution graph for each series of measurements to demonstrate a noticeable difference in legislative output between unified and divided government.

Results

Screen Shot 2016-05-03 at 12.07.39 PM
Figure 4: Normal Distribution of Enacted Laws, 93rd to 113th Congress, based on data from GovTrack (2016)

As shown in Figure 4, the mean number of laws enacted for unified governments between the 93rd and 113th Congresses is 581.8 laws, while the mean for divided governments during that same time period is 476 laws. To calculate the percent difference between divided and unified governments, I use the percent error formula (|OE|^2 / E) * 100, where O refers to the observed value—in this case, the mean number of laws enacted in divided governments—and E refers to the expected value—in this case, the mean number of laws enacted in unified governments:

(|476-581.8|^2 / 581.8) * 100 = 18.2%

The equation above shows that divided governments enact, on average, 18.2% fewer pieces of legislation than their unified counterparts, indicating a statistically significant difference.

Additionally, the normal distribution for legislation enacted under unified governments has a standard deviation of 159.9, while that of divided governments has a standard deviation of 200. Using two standard deviations from the mean to calculate the distribution of 95% of the data, approximately 95% of unified governments enact between 262 and 902 pieces of legislation, while approximately 95% of divided governments enact between 76 and 876 pieces of legislation, a much lower number.

Screen Shot 2016-05-03 at 12.07.22 PM.png
Figure 5: Normal Distribution of the Percentage of Total Legislation Enacted, 93rd to 113th Congress, based on data from GovTrack (2016)

Switching to measuring legislative output as a percentage of total legislation enacted, Figure 5 points to a much smaller difference between the mean percentages for unified and divided governments. Using the same formula mentioned above to calculate the difference between the means, (|OE|^2 / E) * 100, divided governments enact 8.1% less of their total legislation than unified governments. While this number appears to be statistically significant, it is less than half of the previous total of 18.2%.

When analyzing the normal distribution of the percentages of total legislation enacted, unified governments have a standard deviation of 1.157% while divided governments have a standard deviation of 1.484%. Using two standard deviations from the mean to calculate the distribution of 95% of the data, approximately 95% of unified governments enact between 2.059% and 6.687% of total legislation introduced, while approximately 95% of divided governments enact between 1.049% and 6.987% of total legislation produced, indicating significant overlap and a much weaker correlation.

Conclusion

The purpose of this study was to determine, using data on enacted laws and total legislation introduced, to what extent divided government affects the amount of legislation passed by particular Congresses. Depending on which measurement is used to determine the relationship between divided government and legislative output, the party makeup of Congress has either a relatively strong correlation with legislative output, or a relatively weak correlation. Based on my findings, not enough evidence exists to reject the null hypothesis that divided government has zero bearing on legislative output. Perhaps the hypotheses put forward by Saeki—that individual preferences of veto players matter more—and Hughes and Carlson—that divided government slows the pace of legislation, but doesn’t decrease legislative output—more accurately explain the role of divided government in affecting legislative output.

References

Cost, J. (2009, June 10). The Pivotal Politics of Health Care Reform, Part I. Retrieved April 30, 2016, from http://www.realclearpolitics.com/horseraceblog/2009/06/the_pivotal_politics_of_health_1.html

Gill, K. E. (2015). Visual Guide: The Balance Of Power Between Congress and The Presidency (1945-2015). Retrieved May 1, 2016, from http://wiredpen.com/resources/political-commentary-and-analysis/a-visual-guide-balance-of-power-congress-presidency/

GovTrack. (2016). Historical Statistics about Legislation in the U.S. Congress. Retrieved May 1, 2016, from https://www.govtrack.us/congress/bills/statistics

Hughes, T., & Carlson, D. (2015). Divided Government and Delay in the Legislative Process. American Politics Research, 43(5), 771-792. doi:10.1177/1532673X15574594

Quinn, M. (2014, December 30). Turns Out the 113th Congress Wasn’t the ‘Least Productive’ Retrieved April 30, 2016, from http://dailysignal.com/2014/12/30/turns-113th-congress-wasnt-least-productive/

Saeki, M. (2009). Gridlock in the Government of the United States: Influence of Divided Government and Veto Players. British Journal Of Political Science, 39(3), 587-607.