Do Family Wealth Shocks Affect Fertility Choices?

Do Family Wealth Shocks Affect Fertility Choices? Evidence from the Housing Market Boom Michael F. Lovenheim (Cornell University) Kevin J. Mumford (Purdue University) Purdue University SHaPE Seminar January 2012

Introduction There is a great deal of evidence showing that income and family size are negatively correlated high income families tend to have fewer children than low income families developed countries tend to have lower birth rates than developing countries over time, per-capita income has increased while fertility has decreased higher incomes from the industrial revolution have been linked to the demographic transition

Income and Fertility Correlation Births per 1,000 Women age 15 44 60 70 80 90 100 MS Panel B: Birth Rates and Income UT AK ID AZ NM TX SD HI WY CA LA OK NE KS NV AR ND IL MT GA CO SC AL INIA MO MN FL KY NC OR OH WA TN WI MI DE NY NJ VA MD WV PA ME RI NH VT MA CT DC.2.4.6.8 1 Log Real Income Per Capita State average birth rate and average log real income per capita, 1976-2008

Identifying the Causal Effect of Income on Fertility Cross-sectional data is problematic for identifying the causal effect because: fertility, male labor supply, female labor supply, and savings decisions are made jointly. higher wages imply a higher opportunity cost of child raising time higher cost of living implies a higher cost of raising a child (wages positively correlated with cost of living) selection of those with a low preference for children into high cost and high wage areas

Housing Price and Fertility Correlation Births per 1,000 Women age 15 44 50 60 70 80 90 100 Panel A: Birth Rates and Home Prices UT AK ID AZ TX NM SD WY LA MS OK KS ND NE NV AR MT IL GA CO IA INAL MOMN SC OH KY FL MI OR WA WI TN NC VA WV PA HI CA DE NJ MD DC CT NH VT ME RI 20 30 40 50 OFHEO Housing Price Index NY MA State average birth rate and housing price, 1976-2008

Literature The literature has looked for exogenous income shocks: Because of market conditions, some cohorts experience higher income than others. Heckman and Walker (1990) show that high income cohorts in Sweden have higher fertility. A job loss is a negative shock to income. Lindo (2010) and Amilachuk (2006) both show that fertility is negatively affected by lower income due to job loss. The 1970s coal boom in West Virginia caused an unexpected increase in income in counties with a lot of coal. Black et al. (2009) shows that it also caused an increase in fertility in those counties. Opportunity Cost of Time Criticism

Our Identification Strategy We use the housing wealth variation supplied by the recent housing market boom as a source of exogenous variation in household wealth. individual-level data from 1985-2007 on (self-reported) home value and fertility (natality files) from the PSID for women age 25-44 we use within and across MSA variation in housing price we compare homeowners to renters that live in the same MSA we find that homeowners who experience an increase in housing wealth have higher fertility.

Children and Housing Following Becker (1960), we think of families as choosing how many children to have in a utility maximizing framework. The cost of raising a child differs over families and depends on the value of parent time, the cost of relevant market goods, and the child production function. Housing is an important input to the child production function. Substitution Effect: increase in the price of housing should lead to a (weak) decrease in the demand for children Income Effect for Homeowners: increase in the price of housing should lead to an increase in the demand for children Income Effect for Renters: increase in the price of housing should lead to no change in the demand for children

Fertility Response Will Occur with a Lag Births per 1,000 Women age 15 44 64 66 68 70 72 housing fertility 1975q1 1980q1 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 Time.1.05 0.05.1 real housing price percent change

PSID Data Panel A: Homeowners Variable Mean Std. Dev. Min Max Birth in the past year 0.050 0.218 0 1 Home Value ($100,000) 1.5853 1.4415.0104 33.2274 2-Year Home Value Change ($100,000) 0.3496 0.7313-2.9668 4.9609 4-Year Home Value Change ($100,000) 0.6072 0.9881-2.8814 14.0415 Married 0.810 0.392 0 1 Real Family Income ($100,000) 0.8620 0.6910-0.8142 24.0098 Children 1.539 1.194 0 9 Obs = 32218 Panel B: Renters Variable Mean Std. Dev. Min Max Birth in the past year 0.054 0.226 0 1 Market Average Home Price ($100,000) 1.6367 0.8285 0.4212 6.9224 2-Year Market Price Change ($100,000) 0.0664 0.4068-2.8971 3.8146 4-Year Market Price Change ($100,000) 0.1150 0.5927-2.9074 7.5964 Married 0.457 0.498 0 1 Real Family Income ($100,000) 0.4051 0.3495 -.8652 15.5748 Children 1.560 1.386 0 9 Obs = 27252

Empirical Methodology birth ist = β 0 +β 1 HomeValue ist +γx ist +θ s +φ t +η ist, Birth ist = 1 if gave birth in the previous year X ist = vector of demographic characteristics including marital status, real family income, number of other children, woman s age, woman s educational attainment as well as unemployment and real income per capita at the state-by-year level. We also include MSA and year fixed effects. Standard errors clustered at the MSA-level (or state-level for rural sample). We estimate this model separately for homeowners and renters, using average home price measures for the renter regressions.

Homeowners Dependent Variable: Birth ist (Dummy = 1 if Birth in the Previous Year) Independent Variable (1) (2) (3) Home Value ($100,000) -0.0020.. (0.0013).. 2-Year Home Value Change ($100,000). 0.0088.. (0.0019). 4-Year Home Value Change ($100,000).. 0.0085.. (0.0020) Real Family Income ($100,000) -0.0009-0.0010-0.0013 (0.0020) (0.0020) (0.0020) All estimates include MSA and year fixed effects, age group dummies, educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA-level. The sample only includes those who live in an identifiable MSA at the time of the interview and own a home.

Renters Dependent Variable: Birth ist (Dummy = 1 if Birth in the Previous Year) Independent Variable (1) (2) (3) Average Home Value ($100,000) 0.0035.. (0.0030).. 2-Year Home Value Change ($100,000). 0.0001.. (0.0045). 4-Year Home Value Change ($100,000).. 0.0019.. (0.0031) Real Family Income ($100,000) -0.0049-0.0041-0.0046 (0.0055) (0.0060) (0.0063) All estimates include MSA and year fixed effects, age group dummies, educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA level. The sample only includes those who live in an identifiable MSA at the time of the interview and do not own a home.

How Large Are These Estimates? Marginal effect is about 0.85 percentage points for each $100,000 increase Baseline fertility rate = 0.05 This implies a $100,000 increase in housing wealth increases fertility by 17 percent Average 2-year home price increase from 2000-2005 was $48,025 9% increase in fertility over this period due to housing wealth. Average 4-year home price increase from 2000-2005 was $77,911 13% increase in fertility over this period due to housing wealth.

Simulated Home Price Changes Dependent Variable: Dummy=1 if Gave Birth in the Previous Year Current MSA Original MSA Independent Variable (1) (2) (3) (4) 2-Year Change ($100,000) 0.0073 0.0075 (0.0016) (0.0016) 4-Year Change ($100,000) 0.0057 0.0055 (0.0012) (0.0012) Real Family Income ($100,000) -0.0003-0.0007-0.0002-0.0001 (0.0020) (0.0020) (0.0020) (0.0020) R 2 0.068 0.069 0.067 0.068 All estimates include MSA and year fixed effects, age group dummies, educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA level. The sample only includes those who live in an identifiable MSA at the time of the interview and own a home. Simulated Home Price Growth: ˆ P ist P ist 4 where ˆ P ist = P is,t 4 hpist hpi s,t 4. hpi st is an MSA-level home price index (OFHEO).

Estimates by Age Estimated Mean Estimated % 4-Year Home Effect of Fertility Change in Birth Price Change $100,000 Increase Rate Probability Interacted with: (1) (2) (3) Age Group: 15-19 20-24 25-29 30-34 35-39 40-44 R 2 = 0.060-0.0059 (0.0016) 0.0327-18.04% -0.0049 (0.0037) 0.0740-6.62% 0.0175 (0.0054) 0.1172 14.93% 0.0144 (0.0040) 0.0865 16.65% 0.0081 (0.0020) 0.0263 30.80% 0.0042 (0.0014) 0.0056 75.00% Linear probability model estimates of the effect of the 4-year change in housing price ($100,000) on the probability of birth for women in the 1990-2007 PSID. MSA-level fixed effects are included, as well as educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA level.

Not just an effect on timing 4-year change Lagged 4-year change Four-Year Birth Births (1) (2) 0.0079 0.0286 (0.0024) (0.0110) 0.0009-0.0044 (0.0018) 0.0102 MSA-level fixed effects are included, as well as educational attainment dummies, and controls for marital status, the number of other children in the household, family income, state-by year unemployment rates and state-byyear real income per capita. Standard errors clustered at the MSA level.

Estimates by Family Size Estimated Mean Estimated % 4-Year Home Effect of Fertility Change in Birth Price Change $100,000 Increase Rate Probability Interacted with: (1) (2) (3) Number of Children: 0 Children 1 Child 2 Children 3+ Children R 2 = 0.074-0.0007 (0.0025) 0.0428-1.64% 0.0305 (0.0048) 0.0727 41.95% 0.0041 (0.0024) 0.0448 9.15% 0.0056 (0.0032) 0.0494 11.34% Linear probability model estimates of the effect of the 4-year change in housing price ($100,000) on the probability of birth for women in the 1985-2007 PSID. MSA-level fixed effects are included, as well as educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA level.

Estimates by Family Income Estimated Mean Estimated % 4-Year Home Effect of Fertility Change in Birth Price Change $100,000 Increase Rate Probability Interacted with: (1) (2) (3) Family Income: Top Quartile Third Quartile Second Quartile Bottom Quartile R 2 = 0.070 0.0103 (0.0023) 0.0543 18.97% 0.0105 (0.0049) 0.0605 17.36% 0.0076 (0.0038) 0.0515 14.76% -0.0018 (0.0034) 0.0427-4.22% Linear probability model estimates of the effect of the 4-year change in housing price ($100,000) on the probability of birth for women in the 1985-2007 PSID. MSA-level fixed effects are included, as well as educational attainment dummies, and controls for marital status, the number of other children in the household, state-by year unemployment rates and state-by-year real income per capita. Standard errors clustered at the MSA level.

Estimates by Decade and Sign of Price Change Estimated Mean Estimated % 4-Year Home Effect of Fertility Change in Birth Price Change $100,000 Increase Rate Probability Interacted with: (1) (2) (3) Decade: 1985-1989 1990-1999 2001-2007 R 2 = 0.070 0.0046 (0.0032) 0.0618 7.44% 0.0098 (0.0027) 0.0517 18.96% 0.0101 (0.0027) 0.0415 24.34% Sign of Price Change: Positive Change Negative Change R 2 = 0.070 0.0104 (0.0016) 0.0521 19.19% -0.0124 (0.0087) 0.0487-25.46%

Conclusion We find that fertility responds positively to housing wealth variation. These are the first estimates in the literature that examine wealth and do not use variation in household resources that impact the opportunity cost of time. This analysis fits into a larger and growing literature indicating housing wealth affects various aspects of household behavior - consumption, health insurance, education. Our results suggest that housing booms have important fertility and long-run demographic consequences.