Skip to main content
. Author manuscript; available in PMC: 2013 Jul 16.
Published in final edited form as: Q J Econ. 2011;126(1):103–143. doi: 10.1093/qje/qjr001

Table IV.

Unexpected Emotional Shocks from Football Games and Male-on-Female Intimate Partner Violence Occurring at Home.

Poisson Regression
Intimate Partner Violence, Male on Female, at Home
Baseline
Model
(1) (2) (3) (4) (5)
(a) Loss × Predicted Win (Upset Loss) .112 (.034) .099 (.032) .100 (.032) .096 (.031) .100 (.031)
Loss × Predicted Close (Close Loss) .031 (.026) .030 (.024) .032 (.024) .025 (.024) .026 (.024)
(b) Win × Predicted Loss (Upset Win) .001 (.037) .007 (.027) .016 (.027) .010 (.029) .007 (.029)
Predicted Win −.014 (.028) −.019 (.025) −.018 (.025) −.009 (.024) −.081 (.035)
Predicted Close −.022 (.025) −.012 (.030) −.013 (.028) −.010 (.030) −.080 (.043)
Predicted Loss −.016 (.023) −.007 (.021) −.016 (.021) .006 (.021) −.071 (.039)
Non-game Day --- --- --- --- ---
Nielsen Rating .003 (.001)
Agency Fixed Effects X X X X X
Season, Week of Season, & Holiday Variables X X X X
Weather Variables X X X
Nielsen Data Sub-sample X X
Loss Aversion Test:
  p-value for row (a) = – row (b) .02 .01 .00 .01 .01
Number of Agencies 764 764 764 747 747
Observations 79,386 79,386 79,386 73,522 73,522

Notes: Standard errors in parentheses, clustered by team×season (62 groups). Predicted win indicates a point spread of −4 or less (negative spreads indicate the number of points a team is expected to win by); predicted close indicates a point spread between −4 and +4 exclusive; predicted loss indicates a spread of +4 or more. Sample is limited to Sundays during the regular NFL football season. Agencies are NIBRS law enforcement units reporting crime for a city or county; agencies are matched to the corresponding local NFL team for their state. The unit of observation is an agency–day (where a day runs from 12:00 PM to 11:59 PM ET). There are 12 football seasons included in the sample and 17 weeks in each season. The holiday variables include indicators for Christmas Eve, Christmas Day, New Year’s Eve, New Year’s Day, Halloween, as well as Thanksgiving, Labor, Columbus, and Veterans Day weekends. Weather variables include indicators for hot, high heat index, cold, windy, rainy, and snowy days. The Nielsen data subsample is limited to observations with available television ratings; for earlier seasons, not all local markets were tracked by Nielsen Media Research (see note to Table 3).