Legislators’ accountability for issue stances: evidence from Australia’s Marriage Law Postal Survey

Figures & data

Figure 1. Directed acyclic graph of factors affecting incumbent voting. z₁ and z₂ represent observed and unobserved causes of respondent and incumbent position on same-sex marriage; z₃ observed and unobserved causes of area attitudes.

Directed acyclic graph showing the theorised relationship before geographic support for same-sex marriage, individual support for same-sex marriage, MPs' support for same-sex marriage, congruence between these positions, and how the MP eventually voted.

Table 1. Logistic regression models of incumbent voting. Models shown for first-preference and two-party preferred vote, with and without controls.

	FP (1)	FP (2)	TPP (3)	TPP (4)
R support for SSM	−0.310	−0.148	−0.499	−0.404
	[−0.602, −0.022]	[−0.534, 0.243]	[−0.822, −0.185]	[−0.811, −0.012]
Incumbent support for SSM	−0.103	0.094	−0.258	−0.175
	[−0.517, 0.307]	[−0.432, 0.609]	[−0.712, 0.182]	[−0.744, 0.385]
Congruence	−0.095	0.041	−0.047	0.082
	[−0.426, 0.238]	[−0.399, 0.476]	[−0.402, 0.308]	[−0.363, 0.521]
N observations	668	668	668	668
LOOIC	908.1	605.1	881.5	612.6
Bayesian R-squared	0.07	0.48	0.11	0.47
Percentage correctly predicted	0.63	0.83	0.66	0.82

Note: Models 2 and 4 include incumbent-level and area-level controls as described in the text.

Figure 2. Change in the vote share for Coalition incumbents as a function of local support for same-sex marriage and MP stance. Plotted points are scaled in proportion to the number of citizens in each polling place catchment area.

Table 2. Multilevel linear regression models of incumbent voting. Models shown for first-preference and two-party preferred vote, with and without controls.

	(1)	(2)
Intercept	12.593	19.020
	[10.972, 14.371]	[15.386, 22.576]
Lagged vote share	0.736	0.708
	[0.719, 0.754]	[0.687, 0.728]
Incumbent support for SSM	0.584	0.677
	[−1.133, 2.157]	[−0.981, 2.389]
Local support for SSM	−0.151	−0.096
	[−0.177, −0.125]	[−0.130, −0.063]
Congruence	−0.003	−0.008
	[−0.031, 0.026]	[−0.036, 0.019]
N observations	3057	3057
LOOIC	17,699.0	17,430.1
Bayesian R-squared	0.81	0.83

Note: Model 2 includes incumbent-level and area-level controls as described in the text.

Table 3. Multilevel linear regression models of incumbent vote share. Models shown without and with controls.

	(1)	(2)
Intercept	11.219	24.905
	[9.411, 13.060]	[21.190, 28.597]
Lagged vote share	0.715	0.606
	[0.696, 0.734]	[0.587, 0.624]
Lagged vote share, Labor inc.	0.231	0.254
	[0.194, 0.270]	[0.214, 0.293]
Incumbent support for SSM	−0.506	−0.743
	[−2.066, 0.968]	[−2.171, 0.659]
Incumbent support for SSM, Labor inc.	0.960	0.787
	[−1.789, 3.596]	[−1.556, 3.099]
Labor incumbent	−5.495	−10.257
	[−8.691, −2.160]	[−15.950, −4.599]
Local support for SSM	−0.084	−0.038
	[−0.114, −0.055]	[−0.069, −0.009]
Local support for SSM, Labor inc.	−0.056	−0.029
	[−0.108, −0.001]	[−0.081, 0.026]
Congruence	0.026	−0.007
	[0.000, 0.052]	[−0.030, 0.017]
N observations	3839	3839
LOOIC	22,307.7	21,236.4
Bayesian R-squared	0.82	0.86

Note: Model 2 includes additional control variables described in the text.