No crowd, no roar, no home advantage: evidence from ice hockey

Figures & data

Table 1. Descriptive statistics for dependent variables. N = 856.

Dependent variables	Mean	SD	Control seasons: mean	Treatment season: mean
Goals: Home team	3.308	2.079	3.320	3.233
Goals: Away team	2.780	1.910	2.719	3.172**^a
Goal difference: H – A	0.528	3.123	0.601	0.060*
Suspensions: Home team	4.652	2.093	4.574	5.147***
Suspensions: Away team	5.214	2.356	5.197	5.319
Suspension difference: H – A	−0.562	2.329	−0.623	−0.172*

^a The asterisks refer to the p-value for a t-test of the difference in means between the control seasons (home crowd present) and the treatment season, i.e. the COVID-19 season (no home crowd present).

*p < 0.10; **p < 0.05; ***p < 0.01 (two-tailed).

Figure 1. Histogram of goal scoring for the away team by control seasons or no home crowd/COVID-19 season.

Figure 2. Histogram of suspensions for the home team by control seasons or no home crowd/COVID-19 season.

Table 2. Goal scoring by non-presence or presence of home crowd. Linear regression. N = 856.

	Goals: Home	Goals: Away	Goal difference: H – A
No home crowd/COVID-19 season (yes = 1; no = 0)	−0.156 (0.198)	0.558 (0.186)***	−0.713 (0.288)**
Fixed effects for home and away teams	Yes	Yes	Yes
Constant	2.352	2.854	−0.502
R^2	0.298	0.240	0.400

Note. Robust standard errors are in parentheses. The regressions also control for day of week of the match, month of the match, and a playoff versus regular match dummy.

* p < 0.10; ** p < 0.05; *** p < 0.01 (two-tailed).

Table 3. Suspensions by non-presence or presence of home crowd. Linear regression. N = 856.

	Suspensions: Home	Suspensions: Away	Suspensions: H – A
No home crowd/COVID-19 season (yes = 1; no = 0)	0.665 (0.226)***	0.214 (0.251)	0.451 (0.268)*
Fixed effects for home and away teams	Yes	Yes	Yes
Fixed effects for referees	Yes	Yes	Yes
Constant	3.747	4.740	−0.994
R^2	0.125	0.115	0.072

Note. Robust standard errors are in parentheses. The regressions also control for day of week of the match, month of the match, and a playoff versus regular match dummy.

*p < 0.10; **p < 0.05; ***p < 0.01 (two-tailed).

Figure 3. Model of how home crowd presence or not affects goal scoring for away teams: directly as in (i) and indirectly through more suspensions for the home team, as in via (ii) and (iii).

Table 4. Goal scoring for away teams and goal scoring difference between home and away teams by non-presence or presence of home crowd and number of suspensions. Linear regression. N = 856.

	Goals: Away	Goal difference: H – A
No home crowd/COVID-19 season (yes = 1; no = 0)	0.479 (0.192)**	−0.597 (0.279)**
Number of suspensions for home team	0.126 (0.035)***	-
Number of suspensions for away team	−0.035 (0.029)	-
Number of suspensions difference: H – A	–	−0.155 (0.041)**
Fixed effects for home and away teams	Yes	Yes
Fixed effects for referees	Yes	Yes
Constant	2.535	−0.632
R^2	0.282	0.425

Note. Robust standard errors are in parentheses. The regressions also control for day of week of the match, month of the match, and a playoff versus regular match dummy.

*p < 0.10; **p < 0.05; ***p < 0.01 (two-tailed).