Does broadband Internet affect fertility?

Figures & data

Table 1 Descriptive statistics for SOEP sample, Germany 2008–12 (n = 34,495 person-years)

	(1) Mean/proportion	(2) Standard deviation	(3) Minimum	(4) Maximum
Outcome variables
Birth in current year	0.065	0.247	0	1
Working from home	0.208	0.406	0	1
Part-time work	0.313	0.464	0	1
Full-time work	0.431	0.495	0	1
Working hours (weekly)	32.064	14.804	0	60
Not working	0.256	0.437	0	1
Hours spent on childcare (weekdays)	3.276	4.994	0	24
Hours spent on childcare (weekends)	5.358	6.134	0	24
Life satisfaction	7.303	1.646	0	10
High life satisfaction	0.207	0.405	0	1
Got married in current year	0.024	0.154	0	1
Broadband^1
DSL subscription in household	0.815	0.388	0	1
Threshold dummy	0.111	0.314	0	1
‘No closer MDF’ dummy	0.942	0.233	0	1
OPAL dummy	0.010	0.101	0	1
Control variables
Female	0.553	0.497	0	1
Age	33.763	8.031	17	45
Married	0.538	0.499	0	1
Single	0.398	0.490	0	1
Divorced	0.060	0.237	0	1
Number of children	1.234	1.166	0	11
Rural area	0.361	0.480	0	1
West Germany	0.820	0.384	0	1
Lower secondary education (basic track)	0.199	0.399	0	1
Medium secondary education (intermediate track)	0.344	0.475	0	1
Higher secondary education (academic track)	0.325	0.469	0	1
Apprentice	0.127	0.333	0	1
Unemployed	0.073	0.260	0	1
Blue-collar worker	0.222	0.416	0	1
White-collar worker	0.400	0.490	0	1
Entrepreneur	0.058	0.234	0	1
First-generation immigrant	0.130	0.336	0	1
Second-generation immigrant	0.109	0.312	0	1
Household income (log)	7.861	0.546	0	11.408

¹See ‘Empirical method’ section for details of broadband variables.

Notes: Data are for individuals aged 17–45 years. All the samples contain individuals for whom information on childbirth in current year, all observables, and the respective outcome variable are non-missing. The sample sizes for working from home, working hours, childcare (weekdays), childcare (weekends), and life satisfaction are, respectively: 11,590; 24,559; 33,273; 10,802; and 34,169 person-years.

Source: Authors’ analysis of SOEP (v32).

Figure 1 Annual probability of childbirth by DSL access and age: highly educated individuals, Germany 2008–12

Source: Authors’ analysis of SOEP data.

Figure 2 Annual probability of childbirth by DSL access and age: less-educated individuals, Germany 2008–12

Source: As for Figure 1.

Figure 3 Proportion of SOEP households with DSL access by county (Kreis): Germany 2008, 2010, and 2012

Notes: Darker areas correspond to higher levels of DSL access in the corresponding county (Kreis).

Source: As for Figure 1.

Figure 4 Total fertility by county (Kreis): Germany 2008, 2010, and 2012

Notes: Darker areas correspond to higher levels of fertility in the corresponding county (Kreis).

Source: INKAR administrative records.

Table 2 Effects of access to broadband on fertility by age group and sex, Germany 2008–12

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Age group and sex:	17–45			17–24			25–45
	All	Women	Men	All	Women	Men	All	Women	Men
Dependent variable: birth in current year
DSL subscription	0.071	0.105**	0.041	0.021	0.003	0.046	0.080	0.113**	0.044
	(0.046)	(0.049)	(0.050)	(0.051)	(0.083)	(0.039)	(0.055)	(0.055)	(0.064)
Mean of dependent variable	0.065	0.065	0.066	0.028	0.041	0.015	0.073	0.069	0.078
Standard deviation of dependent variable	0.247	0.246	0.248	0.165	0.198	0.121	0.260	0.254	0.268
First stage^1
Threshold	−0.132***	−0.124***	−0.142***	−0.163***	−0.142***	−0.187***	−0.127***	−0.122***	−0.131***
	(0.018)	(0.020)	(0.023)	(0.039)	(0.049)	(0.054)	(0.019)	(0.020)	(0.024)
‘No closer MDF’	−0.062**	−0.052**	−0.074**	−0.104**	−0.096	−0.103	−0.055**	−0.044*	−0.067**
	(0.024)	(0.026)	(0.030)	(0.050)	(0.060)	(0.069)	(0.025)	(0.026)	(0.031)
OPAL	−0.091**	−0.066	−0.117*	−0.085	−0.104	−0.061	−0.091**	−0.058	−0.128**
	(0.043)	(0.047)	(0.060)	(0.086)	(0.095)	(0.137)	(0.043)	(0.049)	(0.059)
F-test of excluded instruments	24.43	19.00	17.73	7.074	3.842	4.901	21.71	19.04	14.43
Overidentification test	0.895	0.861	0.494	1.083	1.243	0.541	1.328	1.039	0.793
χ^2 p-value	0.639	0.650	0.781	0.582	0.537	0.763	0.515	0.595	0.673
Observations	34,495	19,069	15,426	5,988	3,036	2,952	28,507	16,033	12,474

¹See ‘Empirical method’ section for details of broadband variables.

Notes: Standard errors are reported in parentheses and are clustered at the household level. All models include controls for education, age and its quadratic term, sex, an indicator for rural area, marital status, occupational status, migration background, and log of net household income. All regressions further include state and survey year fixed effects, as well as state-specific time trends. The F-test for excluded instruments refers to the Kleibergen–Paap F-statistic. The overidentification test is based on the Huber–White robust variance–covariance matrix without clustering.

*Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.

Source: As for Table 1.

Table 3 Effects of broadband on fertility by education and sex: individuals aged 25–45, Germany 2008–12

	(1)	(2)	(3)	(4)	(5)	(6)
Sex and education group:	All	All	Women	Women	Men	Men
	Highly educated	Less educated	Highly educated	Less educated	Highly educated	Less educated
Dependent variable: birth in current year
DSL subscription	0.108*	−0.032	0.133**	−0.074	0.070	0.034
	(0.059)	(0.112)	(0.062)	(0.124)	(0.065)	(0.147)
Mean of dependent variable	0.075	0.070	0.072	0.063	0.079	0.076
Standard deviation of dependent variable	0.263	0.255	0.258	0.244	0.269	0.265
First stage^1
Threshold	−0.132***	−0.105***	−0.124***	−0.115**	−0.147***	−0.092**
	(0.021)	(0.033)	(0.022)	(0.047)	(0.030)	(0.037)
‘No closer MDF’	−0.061**	−0.037	−0.044	−0.047	−0.090**	−0.023
	(0.028)	(0.045)	(0.029)	(0.057)	(0.037)	(0.055)
OPAL	−0.094**	−0.105	−0.036	−0.344**	−0.159**	0.016
	(0.047)	(0.103)	(0.051)	(0.150)	(0.065)	(0.135)
F-test of excluded instruments	18.52	5.049	15.52	4.948	12.08	2.826
Overidentification test	0.497	3.530	0.677	2.173	0.111	2.141
χ^2 p-value	0.780	0.171	0.713	0.337	0.946	0.343
Observations	19,818	8,689	11,710	4,323	8,108	4,366

¹See ‘Empirical method’ section for details of broadband variables.

Notes: Standard errors are reported in parentheses and are clustered at the household level. All models include controls for education, age and its quadratic term, sex, an indicator for rural area, marital status, occupational status, migration background, and log of net household income. All regressions further include state and survey year fixed effects, as well as state-specific time trends. The F-test for excluded instruments refers to the Kleibergen–Paap F-statistic. The overidentification test is based on the Huber–White robust variance–covariance matrix without clustering.

*Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.

Source: As for Table 1.

Table 4 County-level analysis: first-difference model of effects of broadband on fertility, Germany 1999–2008

	(1)	(2)	(3)	(4)
Dependent variable: change in TFR	OLS		2SLS
Change in DSL availability 1999–2008	0.065** (0.025)	0.063** (0.026)	0.467*** (0.177)	0.326** (0.146)
Control variables	No	Yes	No	Yes
Mean of dependent variable	1.406	1.406	1.406	1.406
Standard deviation of dependent variable	0.131	0.131	0.131	0.131
F-test of excluded instrument	–	–	20.94	20.17
Observations	763	763	763	763

Notes: Standard errors are reported in parentheses and are clustered at the county level. The dependent variable is the change in the total fertility rate (TFR) at the county level between 1999 and 2008. All regressions include survey year fixed effects. Columns (2) and (4) also include county-level controls, such as GDP per capita, population density, net migration rate, the population share of individuals with a university degree, and the population share of individuals of working age (18–64).

*Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.

Source: Authors’ analysis of INKAR data.

Table 5 Potential mechanisms for effect of broadband on fertility: highly educated women aged 25–45, Germany 2008–12

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Dependent variable:	Got married	Work from home	Part-time work	Full-time work	Working hours	Not working	Childcare (weekday)	High life satisfaction
DSL subscription	0.052	0.294*	0.350**	−0.240	−13.377**	−0.110	3.693*	0.270**
	(0.041)	(0.178)	(0.174)	(0.167)	(5.836)	(0.142)	(1.928)	(0.133)
Mean of dependent variable	0.026	0.243	0.474	0.320	28.21	0.207	5.353	0.219
Standard deviation of dependent variable	0.160	0.429	0.499	0.466	14.10	0.405	5.982	0.414
First stage^1
Threshold	−0.121***	−0.127***	−0.124***	−0.124***	−0.114***	−0.124***	−0.124***	−0.126***
	(0.022)	(0.029)	(0.022)	(0.022)	(0.024)	(0.022)	(0.022)	(0.022)
‘No closer MDF’	−0.044	−0.024	−0.044	−0.044	−0.025	−0.044	−0.047	−0.045
	(0.030)	(0.040)	(0.030)	(0.030)	(0.033)	(0.030)	(0.029)	(0.030)
OPAL	−0.037	0.010	−0.034	−0.034	−0.009	−0.034	−0.035	−0.026
	(0.051)	(0.064)	(0.051)	(0.051)	(0.051)	(0.051)	(0.051)	(0.051)
F-test of excluded instruments	14.33	9.849	15.13	15.13	11.62	15.13	15.35	15.74
Observations	11,710	4,067	11,710	11,710	9,164	11,710	11,393	11,615

¹See ‘Empirical method’ section for details of broadband variables.

Notes: Standard errors are reported in parentheses and are clustered at the household level. All models include controls for education, age and its quadratic term, sex, an indicator for rural area, marital status, occupational status, migration background, and log of net household income. All regressions further include state and survey year fixed effects, as well as state-specific time trends. The F-test for excluded instruments refers to the Kleibergen–Paap F-statistic.

*Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.

Source: As for Table 1.