Matthew Rice ECON 203 Project
ð“ð¡ðž ðžðŸðŸðžðœð ð¨ðŸ ðð¡ðž â€œðð¢ð ð…ð¢ð¯ðžâ€ ð©ðžð«ð¬ð¨ð§ðšð¥ð¢ðð² ðŸðšðœðð¨ð«ð¬ ð¨ð§ ð©ðšð«ðð² ð¥ðžðšððžð«ð¬ð¡ð¢ð© ðšðððšð¢ð§ð¦ðžð§ð ð¢ð§ ðð¡ðž ð”.ð’. ð’ðžð§ðšððž.
Only 100 individuals select the party leaders of the U.S. Senate: the U.S. Senators themselves. These officials have not only served with one another for yearsâ€”they have learned each otherâ€™s personal tendencies, qualities that, when measured, form the compos- ite of oneâ€™s personality. While many factors influence the rise of certain Senators to leader- ship in the Congress, one current Senator explains that â€œpersonal relationships are para- mountâ€ in this process.
In this study, I seek to explore the determinants of party leadership attainment in the U.S. Senate. In particular, Iâ€™m interested in examining whether any of the â€œBig Fiveâ€ factors1 have a statistically significant effect on which Senators are elected leaders by their respective peers. Because newer Senators have yet to advance through the ranks, I also use data from the ð¶ð‘œð‘›ð‘”ð‘Ÿð‘’ð‘ ð‘ ð‘–ð‘œð‘›ð‘Žð‘™ ð‘…ð‘’ð‘ð‘œð‘Ÿð‘‘ to use the number of years Senators have served as another independent variable. I predict that Senate leaders tend to be more open to experi- ence, more conscientious, more extroverted, less agreeable, more emotionally stable, and have served for a longer period of time.
In order to evaluate the determinants of Senate leadership attainment, I use data collected by University of Illinois Professor Jeffery J. Mondak for this recent paper, â€œPer- sonalityâ€™s Role in the Worldâ€™s Greatest Deliberative Body.â€ His dataset estimates â€œBig Fiveâ€ scores on a 1-7 scale for a sample size of 106 Senators that have served from 2009 to the present. By creating a multiple regression model, I can measure how the â€œBig Fiveâ€ factors and years served affect my dependent variable, Senate leadership status, data also obtained from the ð¶ð‘œð‘›ð‘”ð‘Ÿð‘’ð‘ ð‘ ð‘–ð‘œð‘›ð‘Žð‘™ ð‘…ð‘’ð‘ð‘œð‘Ÿð‘‘, the official journal of proceedings for the Congress.
I demonstrate that my final model satisfies all of the underlying assumptions. The data has no serious outliers, there is no serious multicollinearity between the independent variables, and a histogram of the standardized residuals shows the error terms are normally distributed. While a scatter plot of the residuals against predicted values for leadership sta- tus indicates very slight heteroscedasticity, Profesor Petry agreed that a transformation of the dependent variable is not needed to move forward with my regression analysis.
Results suggest that Extraversion, Agreeableness, Emotional Stability, and the number of years served each impact Senate leadership attainment significantly, as expect- ed. To summarize, Senators serving in leadership are more likely to be extroverted and emo- tionally stable, less likely to be agreeable, and more likely to have served for a longer num- ber of years, clearly affecting who becomes the leaders of our legislative branch.
1 In psychology, the â€œBig Fiveâ€ factors are the most common way to measure personality. These factors are: (1) Openness to Experience, the desire to explore and imagine; (2) Conscientiousness, the drive for discipline and order; (3) Extraversion, the inclination toward activity; (4) Agreeableness, the ability to create positive relationships; and (5) Emotional Stability, the maintenance of stability and self-reliance.
2. ðƒðšððš ðƒðžð¬ðœð«ð¢ð©ðð¢ð¨ð§
In this study, I first use data published by Professor Jeffery J. Mondak that mea- sures the â€œBig Fiveâ€ factors of personality for 106 Senators that served from the 111th-115th Congress (2009-present). In Dr. Mondakâ€™s recent paper, â€œPersonalityâ€™s Role in the Worldâ€™s Greatest Deliberative Body,â€ hundreds of experts evaluate Senatorsâ€™ personali- ty traits through the Ten Item Personality Inventory (TIPI). This framework generates â€œBig Fiveâ€ scores for Senators in the sample. Because I believed these personality factors influence which Senators become party leaders, I defined my first independent variables as the â€œBig Fiveâ€ factors: Openness to Experience (ð‘œð‘ð‘’ð‘›ð‘›ð‘’ð‘ ð‘ ), Conscientiousness (ð‘ð‘œð‘›ð‘ ð‘ð‘–ð‘’ð‘›- ð‘¡ð‘–ð‘œð‘¢ð‘ ð‘›ð‘’ð‘ ð‘ ), Extraversion (ð‘’ð‘¥ð‘¡ð‘Ÿð‘Žð‘£ð‘’ð‘Ÿð‘ ð‘–ð‘œð‘›), Agreeableness (ð‘Žð‘”ð‘Ÿð‘’ð‘’ð‘Žð‘ð‘™ð‘’ð‘›ð‘’ð‘ ð‘ ), and Emotional Stability (ð‘’ð‘šð‘œð‘ ð‘¡ð‘Žð‘ð‘–ð‘™ð‘–ð‘¡ð‘¦). These five variables are assessed along a sliding 1-7 scale, where higher scores indicate higher levels of a given factor.
The â€œBig Fiveâ€ factors are not the only determinants of leadership attainment in the Senate, and one other critical independent variable is the number of years each Senator has served. As such, my sixth independent variable is years served (ð‘¦ð‘Ÿð‘ ð‘ ð‘’ð‘Ÿð‘£ð‘’ð‘‘), data which I obtained from my second source, the ð¶ð‘œð‘›ð‘”ð‘Ÿð‘’ð‘ ð‘ ð‘–ð‘œð‘›ð‘Žð‘™ ð‘…ð‘’ð‘ð‘œð‘Ÿð‘‘, the official record of proceed- ings for the U.S. Congress. I also use the ð¶ð‘œð‘›ð‘”ð‘Ÿð‘’ð‘ ð‘ ð‘–ð‘œð‘›ð‘Žð‘™ ð‘…ð‘’ð‘ð‘œð‘Ÿð‘‘ to measure my dependent variable, leadership attainment level (ð‘ ð‘’ð‘›ð‘™ð‘’ð‘Žð‘‘ð‘’ð‘Ÿð‘ hð‘–ð‘), an ordinal variable that gives each Senator a leadership score that ranges from 0 to 3. Senators scored as 0 have not served in a leadership capacity. Senators scored as 1 have only served in party leadership at a junior level. Senators scored as 2 have served as Chairs or Party Secretaries, and Senators scored as 3 are Leaders or Party Whips, the number two position. The ð¶ð‘œð‘›ð‘”ð‘Ÿð‘’ð‘ ð‘ ð‘–ð‘œð‘›ð‘Žð‘™ ð‘…ð‘’ð‘ð‘œð‘Ÿð‘‘ annual- ly publishes this data, which I then coded in my dataset.
Table 1 reports the Descriptive Statistics for these variables. The study has a sample of 106 Senators. Their average leadership attainment score is 0.8490, which indicates that the average Senator has served close to but below a junior level on the party leadership structure. On average, they have served for 16.4245 years in the Senate and score a 4.0784 for Openness to Experience, a 5.1860 for Conscientiousness, a 4.319 for Extraversion, a 4.3002 for Agreeableness, and a 5.037 for Emotional Stability. Figure 1 shows the scatter plots for these six independent variables against the dependent variable. As one can see, no outliers appear present in the data. While Figure 3, a scatter plot of the residuals versus predicted leadership attainment levels, shows very slight heteroscedasticity, Professor Petry informed me that none of the four transformations learned in class would improve this minimal issue. Thus, a transformation of the dependent variable is not necessary.
Before running the regression analysis, I also created a correlation table to determine whether any serious multicollinearity is present between the independent variables. As shown in Table 2, no two independent variables have a coefficient correlation higher than 0.8. As a result, serious multicollinearity shouldnâ€™t be an issue.
3. ð‘ðžð ð«ðžð¬ð¬ð¢ð¨ð§ ð€ð§ðšð¥ð²ð¬ð¢ð¬
In order to evaluate the determinants of Senate leadership attainment, I estimate an initial linear regression model to be:
ð’”ð’†ð’ð’ð’†ð’‚ð’…ð’†ð’“ð’”ð’‰ð’Šð’‘ð’Š = ðœ¶ + ðœ·ðŸð’ð’‘ð’†ð’ð’ð’†ð’”ð’”ð’Š + ðœ·ðŸð’„ð’ð’ð’”ð’„ð’Šð’†ð’ð’•ð’Šð’ð’–ð’”ð’ð’†ð’”ð’”ð’Š + ðœ·ðŸ‘ð’†ð’™ð’•ð’“ð’‚ð’—ð’†ð’“ð’”ð’Šð’ð’ð’Š + ðœ·ðŸ’ð’‚ð’ˆð’“ð’†ð’†ð’‚ð’ƒð’ð’†ð’ð’†ð’”ð’”ð’Š + ðœ·ðŸ“ð’†ð’Žð’ð’”ð’•ð’‚ð’ƒð’Šð’ð’Šð’•ð’šð’Š + ðœ·ðŸ”ð’šð’“ð’”ð’”ð’†ð’“ð’—ð’†ð’…ð’Š + ðœº
As previously discussed, I contend that all five â€œBig Fiveâ€ personality factors and years served are good indicators of Senate leadership attainment. First, Dr. Mondak shows Openness to Experience (ð‘œð‘ð‘’ð‘›ð‘›ð‘’ð‘ ð‘ ) shapes Senatorsâ€™ originality, penchant for divergent thinking, and intellectual drive, qualities that should all influence who is elected to leader- ship. Second, Conscientiousness (ð‘ð‘œð‘›ð‘ ð‘ð‘–ð‘’ð‘›ð‘¡ð‘–ð‘œð‘¢ð‘ ð‘›ð‘’ð‘ ð‘ ) corresponds to higher levels of disci- pline and strategic thinking, traits expected to be found among those in leadership as well. Third, Extraversion (ð‘’ð‘¥ð‘¡ð‘Ÿð‘Žð‘£ð‘’ð‘Ÿð‘ ð‘–ð‘œð‘›) is linked to leadership capacity in psychology. Fourth, Agreeableness (ð‘Žð‘”ð‘Ÿð‘’ð‘’ð‘Žð‘ð‘™ð‘’ð‘›ð‘’ð‘ ð‘ ) relates to weariness in resolving interpersonal conflict, a task of particular importance to Senate leaders. Fifth, Emotional Stability (ð‘’ð‘šð‘œð‘ ð‘¡ð‘Žð‘ð‘–ð‘™ð‘–ð‘¡ð‘¦) relates to self-confidence, and leaders are elected to stand up to the opposition, not crack under pressure. I also include the number of years served (ð‘¦ð‘Ÿð‘ ð‘ ð‘’ð‘Ÿð‘£ð‘’ð‘‘) as a sixth independent vari- able, because junior Senators have had less time to advance through the party ranks and run for leadership positions themselves.
The regression results from my initial model are shown in Table 3. When assessing the modelâ€™s overall fit, the coefficient of determination (R-Squared) indicates that the mod- el explains 15.3594% of the variation in Senate leadership attainment. At the same time, an F-statistic of 3.0025 indicates the overall model is valid because the null hypothesis can be rejected at a 5% level. While the model is proven to be valid overall, two independent vari- ables did not meet statistical significance: Openness to Experience (ð‘œð‘ð‘’ð‘›ð‘›ð‘’ð‘ ð‘ ) and Consci- entiousness (ð‘ð‘œð‘›ð‘ ð‘ð‘–ð‘’ð‘›ð‘¡ð‘–ð‘œð‘¢ð‘ ð‘›ð‘’ð‘ ð‘ ) show individual slope coefficients that are not different from zero at a 10% level.
Prior to creating my reduced model, I tried my best to make sure the five underlying assumptions were satisfied. First, a histogram of the standardized residuals presented in Figure 2 demonstrates that the error terms are normally distributed. As discussed in the ðƒðšððš ðƒðžð¬ðœð«ð¢ð©ðð¢ð¨ð§ section above, no serious outliers or serious multicollinearity appear present, and while a scatter plot of the residuals against predicted values for leadership sta- tus indicates minimal heteroscedasticity (a slim increase in the variance of the error term as the predicted value of the dependent variable increases), none of the four transformations of the dependent variable learned in class would increase the overall validity the regression model or the significance of the individual independent variablesâ€™ slope coefficients. Be- cause the data is cross-sectional, autocorrelation is not a concern either.
I created my reduced model by removing the two independent variables proven to be insignificant. Table 4 displays the regression results. Compared to the initial model, R- Squared slightly decreased and Adjusted R-Squared slightly increased. This difference indi- cates that the eliminated variables contribute little to explaining the variation of the de- pendent variableâ€”and when R-Squared is adjusted by the number of independent vari- ables, keeping them resulted in a greater penalty than benefit. The F-test statistic becomes more significant, and the remaining independent variables all become significant at the 5% level or below. I then performed a partial F-test to ensure that the eliminated variables did not produce significant results. The partial F-test statistic is 0.6771 and the corresponding p-value is 0.5014. As a result, the null hypothesis that the eliminated variablesâ€™ coefficients equal zero cannot be rejected. Because the error terms also appear normally distributed (Figure 4) and homoscedastic (Figure 5), aside from the slight concern addressed above, the reduced model should be utilized. My final regression model is the following:
ð’”ð’†ð’ð’ð’†ð’‚ð’…ð’†ð’“ð’”ð’‰ð’Šð’‘ð’Š = ðœ¶ + ðœ·ðŸð’†ð’™ð’•ð’“ð’‚ð’—ð’†ð’“ð’”ð’Šð’ð’ð’Š + ðœ·ðŸð’‚ð’ˆð’“ð’†ð’†ð’‚ð’ƒð’ð’†ð’ð’†ð’”ð’”ð’Š + ðœ·ðŸ‘ð’†ð’Žð’ð’”ð’•ð’‚ð’ƒð’Šð’ð’Šð’•ð’šð’Š + ðœ·ðŸ’ð’šð’“ð’”ð’”ð’†ð’“ð’—ð’†ð’…ð’Š + ðœº
4. ð„ð¦ð©ð¢ð«ð¢ðœðšð¥ ð‘ðžð¬ð®ð¥ðð¬
Applying the results from the reduced model in Table 4, I was then able to analyze the results of my model in terms of both statistical and economic significance. The coeffi- cients of the independent variables show a 0.2563 increase in Senate leadership attainment level for every point increase for Extraversion, a 0.2687 decrease in Senate leadership at- tainment level for every point increase for Agreeableness, and a 0.5075 increase in Senate leadership attainment level for every point increase for Emotional Stability. There is also a 0.0194 increase in leadership attainment level for every additional year a Senator has served. While these statistics may seem insignificant, not only are they all significant at the 5% level (and the 1% level for Extraversion and Emotional Stability), they also demonstrate how truly critical these personality factors can be in the rise of certain Senators to leader- ship positions over their respective colleaguesâ€”especially when putting into perspective the relatively small variances found in each of the independent variables shown in the Descrip- tive Statistics table (Table 1).
The strength of these effects is clearly shown when calculating the predicted values of the dependent variable based on the regression results. Senators ranked in the top 5% for Extraversion are 91 percentage points more likely to serve in party leadership than those in the bottom 5%, Senators ranked in the top 5% for Agreeableness are 47 percentage points less likely to serve in party leadership than those in the bottom 5%, and Senators ranked in the top 5% for Emotional Stability are 79 percentage points more likely to serve in party leadership than those in the bottom 5%.
After running the results, it was also interesting to see which â€œBig Fiveâ€ factors did not significantly influence Senate leadership attainment. Openness to Experience and Con- scientiousness did not appear to influence which Senate Members serve in leadership. While I expected both variables to affect leadership attainment, perhaps Senatorsâ€™ interactions with their peers (both in terms of Extraversion and Agreeableness) and their ability to re- main calm in contentious situations (Emotional Stability) contribute more to who is ulti- mately selected to lead the legislative branch, a finding with clear social, political, and eco- nomic consequences for our country as a whole.
5. ð’ð®ð¦ð¦ðšð«ð² ðšð§ð ðƒð¢ð¬ðœð®ð¬ð¬ð¢ð¨ð§
In this study, I examined the determinants of party leadership attainment in the U.S. Senate. In particular, I focused on the influence of the â€œBig Fiveâ€ factors of personality and years served to evaluate whether the personalities and seniority of the Senators themselves relate statistically to who serves in leadership. To conclude, I found that three of the â€œBig Fiveâ€ factors significantly affected which Senators become party leaders: while higher lev- els of Extraversion and Emotional Stability increase levels of leadership attainment, Agree- ableness decreases levels of leadership attainment among the 106 Members in my sample. In addition, Senators that have served for longer amounts of time tend to be more likely to serve in leadership as well.
One limitation of this study is the fact that my sample size did not include the House of Representatives. Because the House is larger, less structured, and more unruly than the Senate, it is quite possible that different personal qualities influence leadership attainment on the other side of the Capitol. In turn, the effect of the â€œBig Fiveâ€ factors on House lead- ership could show different results as well. While the data I used from Dr. Mondakâ€™s re-
search did not include provide â€œBig Fiveâ€ ratings for Members of the House, future studies should examine the role of personality factors in the lower chamber as well.
Another limitation is my dependent variableâ€™s inability to differentiate between the types of leadership positions available to Senators. While leadership attainment level mea- sures whether Senators have served in junior or more Senior roles (or none at all), certain personality factors might be more important in some roles over others. For example, Out- reach Chairs might be more extroverted than Policy Chairs, and Party Leaders might be more emotionally stable than their deputies (or vice versa). These separate roles could all be studied as dependent variables in the future.
...(download the rest of the essay above)