Electronic Supplementary Material for the Article: The Impact of Internet Diffusion on Marriage Rates: Evidence from the Broadband Market

Transcription

1 Electronic Supplementary Material for the Article: The Impact of Internet Diffusion on Marriage Rates: Evidence from the Broadband Market By Andriana Bellou 1

2 Appendix A. Data Definitions and Sources This section describes basic variable definitions and data sources. The baseline sample spans the years and and consists of white, non-hispanic men and women 21 to 30 years old at the survey date. Marriage Data: Marriage rates are measured by the share of individuals 21 to 30 years old at the survey date that have ever been married. This variable is calculated from the current marital status variable of the Current Population Survey March Supplements (IPUMS CPS variable MARST ). The crosssections and are used for the main analysis. Broadband Diffusion Data: Broadband diffusion is measured by the number of residential high speed lines per 100 people in a given state and year from 2000 until This is the number of high-speed lines for Internet access present in a given state/year divided by the state s population. The information on the lines in service is obtained from the Statistical Reports on Local Telephone Competition and Broadband published semi-annually (December and June) since 2000 by the Federal Communications Commission (FCC) ( These reports provide summary statistics of subscribership data that facilities-based providers of Internet access services faster than 200 kbps in at least one direction file twice a year on FCC Form 477. The FCC lumps high-speed lines in service by state in two broad categories on the basis of the type of the end-user customer: Residential/Small Business and Other. The last category includes medium and large business, institutional and government end-users. For my 2

3 analysis, I only employ the number of lines that connect to residential/small business customers. Unfortunately, within this group, there is no way to distinguish connections solely to households and hence when the term residential high speed lines is used in the definition of broadband diffusion in this paper, it should be understood that these are the combined lines in use by residential as well as by small business customers. Population Data: Population estimates by state and year are obtained from the United States Census Bureau (years and ). Online Dating Data: The question on online dating usage for partner search was asked in the 2000, 2002, 2003, 2004, 2005 and 2006 nationally representative surveys conducted by the Pew Research Center of the Pew Internet and American Life Project. The datasets are publicly available at the Pew Research Center s website: Time Use Data: The auxiliary analysis makes use of time use data from 2003 to 2006 accessed using the American Time Use Survey-X extract builder. The sample includes single (never married) individuals 21 to 30 years old at the survey date. Non-work time spent on socializing includes socializing, communicating, playing games, doing arts and crafts, listening to music, watching movies, attending social events with other people (parents, siblings, friends, coworkers, neighbors). Non-work time spent on using the computer includes time spent alone checking household and personal s and messages as well as alone time spent on computer use for 3

4 leisure (browsing on the Internet, participating in chat rooms burning CDs, downloading files, pictures and music but excluding playing games). Other State-Level Variables: The following state-level covariates are calculated from the Current Population Survey Annual March Supplements (cross sections and ): sex ratio, share of the population in various age brackets and education groups, share nonwhite, share living in a metropolitan area, average male and female wages, median household income, unemployment rate. Wage and income are deflated using the Consumer Price Index for all urban consumers and all items published by the BLS (base ). The number of firms operating in high-tech sectors is calculated from the County Business Patterns (years and ). To define high-tech industries, I follow the classification proposed by Hecker (1999; Table 1). Annual Gross Domestic Product by state is obtained from the Bureau of Economic Analysis (years and ). Residential telephone adoption is measured by the number of residential phones per family in a given state in The information on the number of residential telephones in 1955 and the number of families is obtained from the 1956 County and City Data Book. These files can be accessed through ICPSR. The share of farm households in 1930 and the share of households in 1960 having hot and/or cold piped water are calculated from the Integrated Public Use Micro Sample of the decennial Census of the United States (IPUMS USA). 4

5 B. Tables Appendix Table 1: Trends in Marriage Rates and Broadband Diffusion % Ever Married Broadband Change: % Married Change: Broadband State Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware DC Florida Georgia Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska

6 Appendix Table 1: Trends in Marriage Rates and Broadband Diffusion (Cont.) % Ever Married Broadband Change: % Married Change: Broadband State Nevada N. Hampshire N. Jersey N. Mexico N. York N. Carolina N. Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island S. Carolina S. Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming NA NA Notes: Broadband is measured by the number of residential high speed lines per 100 people in a state. Marriage rates are calculated as the share of ever married individuals in the population of years old (non-hispanic) whites. Data sources: Annual March CPS Supplements, FCC Statistical Reports on Broadband Deployment. 6

7 Appendix Table 2: Summary Statistics of Selected Variables Variables Sex ratio (21-30 yrs old) % college education (21-30 yrs old) Mean male weekly log wages (21-30 yrs old) Mean female weekly log wages (21-30 yrs old) Fraction residing in an MSA (21-30 yrs old) Fraction years old Fraction years old Fraction years old Fraction years old Fraction 56 years old and above Fraction nonwhite Log Population Median Income Log GDP Unemployment rate Average number of firm in high tech sectors Notes: Age-specific statistics are calculated for the baseline sample of years old non-hispanic whites. GDP is obtained from the Bureau of Economic Analysis. The number of firms in high tech sectors is calculated from the County Business Patterns. Sector classification follows Hecker (1999). Population is obtained from the Census. The remaining covariates are calculated from the Annual March CPS Supplements. 7

8 Appendix Table 3: Broadband Determinants DV: Residential lines per 100 people Log-population density (5.68)** Firms in high tech sectors (0.002)*** % with college education (6.98)*** Urbanicity (3.11)*** % yrs old (relative to 56 years old and above) (10.05)*** % yrs old (relative to 56 years old and above) (12.24)* State and year fixed effects x Observations 299 Notes: OLS estimates from a regression of state Internet penetration on state covariates. These are: population density, number of firms in high tech sectors, share with a college degree, share of nonwhite population, age structure, sex ratio, median income, urbanicity, log GDP, mean female and male wages, unemployment rate. Coefficients of statistically significant controls are only reported. Internet diffusion is measured by the number of residential high speed lines per 100 people in a state/year (FCC, ). Estimates are weighted by state population. Robust standard errors (parentheses) clustered at the state-year level. ***, **, * show significance at 1%, 5% and 10% respectively. 8

9 Appendix Table 4: Impact of Broadband Diffusion on Marriage Rates (Baseline) DV: Ever Married (1) (2) (3) Men Women All [Mean] [0.408] [0.559] [0.484] Internet diffusion (0.0015)*** (0.002) (0.0016)*** Respond. resides in an MSA (0.005)*** (0.004)*** (0.004)*** Respond. is male (0.020) Respond.: high school (0.005)*** (0.005)*** (0.004)*** Respond.: some college (0.005)*** (0.004)*** (0.003)*** Log-Population density (0.081) (0.077)* (0.070) Log-GDP (0.068) (0.061) (0.058) Firms in high-tech sector ( ) ( )* ( )** % Nonwhite (0.115) (0.126) (0.103) % yrs old (0.217)*** (0.232) (0.194)*** % yrs old (0.220)*** (0.198)** (0.183)*** % yrs old (0.226) (0.211) (0.188) % yrs old (0.240) (0.224) (0.199) Avg female log-wage (21-30 yrs old) (0.021)* (0.024) (0.019) Avg male log-wage (21-30 yrs old) (0.023) (0.025) (0.021) % High school (21-30 yrs old) (0.062) (0.065)** (0.055) % Some college (21-30 yrs old) (0.063) (0.071) (0.058) 9

10 Appendix Table 4: Impact of Broadband Diffusion on Marriage Rates (cont.) DV: Ever Married (1) (2) (3) Men Women All [Mean] [0.408] [0.559] [0.484] % yrs old residing in an MSA (0.041)*** (0.037)** (0.034)*** Sex ratio (21-30 yrs old) (0.017)*** (0.017)*** (0.017)** Sex ratio (21-30 yrs old)*male dummy (0.020)*** Unemployment rate (0.017) (0.183) (0.152) Log household income (0.039) (0.044) (0.036) Age dummies x x x State and year fixed effects x x x Observations Notes: Reported coefficients are OLS estimates of specification (1). Dependent variable equals 1 if the respondent has ever been married and 0 otherwise. Internet diffusion is measured by the number of residential high-speed lines per 100 people in a given state/year. Individual-specific covariates: age, gender, metropolitan status, education dummies. State and year-specific covariates: log-population density, log-gdp, median household income, number of firms in high-tech sectors, fraction of non-white population, unemployment rate, fraction of population in different age brackets (16-25, 26-35, 36-45, and above years old), share of years old whites residing in an MSA, average wage of years old white males and females, share of years old white population having highschool, some college and college education, sex ratio and its interaction with a gender dummy (Column 3). Sex ratio is calculated as the number of white men years old over the number of women in the same age and racial group. The sample includes non-hispanic white men and women years old. Estimation is performed using the available sampling weights. Robust standard errors (parentheses) are clustered at the state-year level. ***, **, * denote statistical significance at 1%, 5% and 10% respectively. 10

11 Appendix Figure 1: State Household Broadband Diffusion Predicted by the Share of Households with Telephones in ,25 0,23 Actual Residential Broadband Diffusion (2005) 0,21 CT MA 0,19 NH NJ DC RI 0,17 FL NV MD CA OR WA NY VI KS CO 0,15 ME NE NC AZ GA VT WI MN PA OH MI DE 0,13 OK TX IL WY MO SC TN 0,11 MT IN UT WV AR AL NM ID 0,09 LA 0,07 ND SD MS KY IA 0,05 0,08 0,1 0,12 0,14 0,16 0,18 0, *phones per family (1955) Notes: The graph compares actual residential broadband diffusion measured in 2005 (y-axis) with that predicted by the regression shown on the x-axis: Broadband Diffusion s, t=2005 = α + β*telephone Diffusion s, t= ε. Data sources: 1956 County and City Data Book, FCC Statistical Reports on Broadband Deployment. Actual Residential Broadband Diffusion (2005) Appendix Figure 2: State Household Broadband Diffusion Predicted by the Share of Households with Piped Water in ,25 0,23 0,21 CT MA 0,19 NH NJ DC RI 0,17 MD FL NV CA WA OR NY VI KS CO 0,15 ME NE NC AZ GA MN DE WI OH VT PA MI 0,13 OK TX IL MO WY SC TN 0,11 MT IN UT WV AR AL ID LA NM 0,09 ND 0,07 SD MS KY IA 0,05 0,06 0,08 0,1 0,12 0,14 0, *perc_piped water (1960) Notes: The graph compares actual residential broadband diffusion measured in 2005 (y-axis) with that predicted by the regression shown on the x-axis: Broadband Diffusion s, t=2005 = α + β*piped Water s, t= ε. Data sources: IPUMS USA, FCC Statistical Reports on Broadband Deployment. 11

12 Actual Residential Broadband Diffusion (2005) Appendix Figure 3: State Household Broadband Diffusion Predicted by the Share of Farm Households in ,25 0,23 0,21 CT 0,19 NH MA NJ DC RI 0,17 FL MD NV CA OR WA NY KS CO 0,15 NE VI AZ ME NC GA MN VT WI DE MI OH PA 0,13 OK TX IL WY SC TN MO 0,11 MT WV IN UT AR AL ID 0,09 LA NM ND 0,07 SD KY IA 0,05 0,08 0,1 0,12 0,14 0,16 0, *perc_farms (1930) Notes: The graph compares actual residential broadband diffusion measured in 2005 (y-axis) with that predicted by the regression shown on the x-axis: Broadband Diffusion s, t=2005 = α + β*farm Households s, t= ε. Data sources: IPUMS USA, FCC Statistical Reports on Broadband Deployment. 12