Political Networks of Companies - AOM Conference Paper

Notes on Regulatory Uncertainty Variable

The following is a very brief summary of the effect of Including Q25 and Q26, Regulatory Uncertainty, into the predictor set of political networks.

For the full-rank regression, it turns out Q26 (but not Q25) *IS* significant for both Senate response, but not at all for the House, which is interesting.  None of the residuals are normal.

Let us check the reduced predictor sets and see how this turns out.

Changes:  Between all responses, both Q25/26 were needed.  With the addition of Q25/26, AllGive02 dropped out of Q11c (Senate personal meetings), although Allgive was not included in the other responses under the previous predictors.

Reduced regressions on Size of Network for both House and Senate were unchanged with the addition of Regulatory Uncertainty predictor.  However, the number of private meetings in both the Houses were explained by a significantly different set of predictors with the inclusion of Q25 and Q26 into the superset.  (Go back and see comparison chart for effects.)

Basic Correlations: Largest is somewhat significant (Q11c, Q25)

Q8  Q11c   Q17  Q20c   Q25   Q26

Q8  1.00  0.50  0.71  0.51 -0.18  0.06

Q11c  0.50  1.00  0.29  0.40 -0.23  0.13

Q17  0.71  0.29  1.00  0.62 -0.08 -0.06

Q20c  0.51  0.40  0.62  1.00 -0.17  0.00

Q25 -0.18 -0.23 -0.08 -0.17  1.00 -0.12

Q26  0.06  0.13 -0.06  0.00 -0.12  1.00

Check Significance of Largest Correlation:

cor(q11c, q25) = -0.25

Pearson's product-moment correlation

------------------------------------

data:  q11c and q25

t = -2.2769, df = 79, p-value = 0.0255

sample estimates:  cor  -0.248157

Kendall's rank correlation tau

-----------------------------

data:  q11c and q25

normal-z = -2.0938, p-value = 0.0363

sample estimates:  tau  -0.139198

Spearman's rank correlation

---------------------------

data:  q11c and q25

normal-z = -2.003, p-value = 0.0452

sample estimates:  rho   -0.223926