May 18

Ghana Subgroup Regressions

Below, we see if there is heterogeneity in how temperature impacts mental health. We split our sample into 4, grouping them by their baseline mental health. Then, we run separate regressions for each group to see if the temperature-mental health relationship differs across these groups.

We generally see those with worse baseline mental health are more affected by temperature but this isn’t consistent across measures.

Subgroup Regressions

All Measures - Anderson

	All Measures - Anderson
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.012 (0.009)	0.023*** (0.009)	0.030*** (0.009)	0.026*** (0.009)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.43	0.431	0.377	0.405
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

PHQ-2 & GAD-2 - Anderson

	PHQ-2 & GAD-2 - Anderson
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.015 (0.010)	0.026*** (0.009)	0.017* (0.010)	0.028*** (0.010)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.349	0.352	0.314	0.327
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

PHQ-2

	PHQ-2
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.012 (0.009)	0.020** (0.009)	0.021** (0.009)	0.026*** (0.009)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.309	0.305	0.277	0.288
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

GAD-2

	GAD-2
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.013 (0.010)	0.025** (0.010)	0.007 (0.010)	0.023** (0.010)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.306	0.32	0.284	0.302
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

Perceived Stress

	Perceived Stress
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.002 (0.011)	0.007 (0.010)	0.018 (0.011)	0.005 (0.011)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.305	0.296	0.27	0.284
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

Worry

	Worry
	Baseline MH < p25 - Daily Mean	Baseline MH p25-p50 - Daily Mean	Baseline MH p50-p75 - Daily Mean	Baseline MH > p75 - Daily Mean
	(1)	(2)	(3)	(4)
coef
Daily Mean Temperature	0.004 (0.011)	0.009 (0.010)	0.021** (0.011)	0.020* (0.010)
fe
period	x	x	x	x
hh_id	x	x	x	x
stats
Observations	3,622	3,642	3,692	3,690
R-squared	0.326	0.352	0.303	0.317
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Format of coefficient cell: Coefficient (Std. Error)

Note

To try testing a similar story in the DHS, I tried proxying “poor mental health” with other variables that could reasonable cause worsening in how people feel. These are food security, substance use (particularly alcohol consumption) and domestic violence. In all cases, people in all quantile group respond similar to temperature which is having no effect.