Fertility, Social Mobility, and Long Run Inequality Juan Carlos Córdoba1, Xiying Liu2, and Marla Ripoll3 Abstract We investigate social mobility and long run inequality in the presence of endogenous fertility choices made by altruistic individuals facing uninsurable idiosyncratic risk. The role of fertility choices in analyzing long run inequality is important for at least two reasons. First, there is a well-documented negative relationship between fertility and income. For example, Jones and Tertilt (2008) estimate an income elasticity of fertility of about -0.38 using US Census data. Second, differential fertilities among rich and poor families lead to differences in intergenerational wealth transmission, social mobility and long run inequality. As reported by Menchik (1979), the median child-parent wealth ratio in one-child families is three times that of families with three or more children. Earlier theoretical work by Barro and Becker (1988, 1989), but especially Alvarez (1999), finds that fertility choices by altruistic parents largely reduce intergenerational persistence and increase social mobility. The reason is that wealthier parents have more children and transfer to each child an amount that is independent of parental wealth. This lack of intergenerational persistence is counterfactual. A key goal of this research is to recover empirically plausible levels of persistence with altruistic models of endogenous fertility. We show that a calibrated version of a Barro-Becker dynastic altruistic model of fertility choice embedded into a Bewley framework of idiosyncratic risk is able to replicate three key aspects of the data: (i) a negative fertility-income relationship; (ii) a negative relationship between family size and savings rates; and (iii) a significant intergenerational persistence or lack of social mobility. We also show that the endogenous fertility model improves upon the exogenous fertility model in a number of other dimensions such as in generating larger wealth dispersion. Our calibration exercise also sheds light on the technology of raising children, the shape of altruism by parents, and on the “intergenerational elasticity of substitution.” The following are the elements explaining our results. First, we calibrate an intergenerational elasticity of substitution larger than one, which implies that children’s consumption can be easily substituted by parental consumption so that richer parents are not particularly driven to have more children. Second, the negative-fertility income relationship naturally arises within the model because of the time cost of raising children together with the less-than-perfect intergenerational persistence of abilities. As a result, low ability parents have more children because they expect their children to be of higher average ability than their own ability. Third, in our model richer families have higher saving and inheritance rates because they have fewer children and each child weights 1 Iowa State University. E-mail: cordoba@iastate.edu Iowa State University. E-mail: xiyingl@iastate.edu 3 University of Pittsburgh. E-mail: ripoll@pitt.edu 2 more into the utility of the parent. Different from Alvarez (1999), we assume exponential child discounting rather than hyperbolic which implies strong decreasing benefits to having extra children. Ours is the first dynamic altruistic model of fertility that is consistent with a number of features describing wealth inequality and intergenerational persistence in the United States. Please see attached Beamer presentation summarizing some of the results of our ongoing research. 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Fertility, Social Mobility, and Long Run Inequality Juan Carlos Cordoba Xiying Liu Marla Ripoll Iowa State University and University of Pittsburgh Nov 2014 Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Motivation Wealth is highly concentrated: the wealthiest 1% owns over 30% of the nation’s wealth. Family size is an important factor: "median child-parent wealth ratio in one-child families was 1.84; the median in families with three or more children was between 0.60 and 0.69." (Menchik 1979). Standard models of inequality ("Bewley" models) implicitly or explicitly assume equal fertility rates among individuals. Laitner 1992, Loury 1981, Aiyagari 1994, Krusell & Smith 1998, Castañeda & Diaz-Gimenez & Rioss-Rull (CDR 2003), Krueger and Perri (2006), Cordoba (2008), etc. But reproduction rates are higher among poorer individuals ... Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Fertility vs. Income Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Motivation: Barro-Becker Families in a Bewley World A Bewley + Barro + Becker (BBB) economy: Bewley = In…nite life+Idiosyncractic earning risk + Incomplete markets = optimal savings. BB = …nite life+dynastic altruism + cost of children = optimal fertility + optimal inheritance. Can BBB models explain: fertility di¤erentials among the poor and the rich? child-parent wealth ratios di¤erentials by family size? long run-inequality, wealth concentration, social mobility? consumption, labor supply, savings, inheritance .... Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions What we do Run horse races between various calibrated BBB models Calibration strategy follows CDR (2003): use cross-section evidence to identify parameters and time-series evidence for overidenti…cation. Calibration is a powerful tool for intergenerational research given the scarcity of longitudinal data. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Main Findings Properly calibrated BBB model predict: realistic negative fertility-income relationship a negative fertiliy-wealth relationship at certain levels of wealth higher inheritance rates by smaller families more wealth concentration and wealth dispersion than exogenous fertility models realistic persistence Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Main Findings Family planning policies: One child policy may induce perpetual growth Two children policy: increases average earnings, income, consumption and wealth but also their dispersion and persistence, particularly persistence of wealth: reduces social mobility. Estate taxation: Reduces average income, consumption and wealth and their dispersion. Reduces fertility Increases social mobility Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Main Findings EIS = Elasticity of Intertemporal Substitution EGS = Elasticity of Intergenerational Substitution EGS seems to be di¤erent from EIS EGS larger than 1 and probably between 1.5 to 2. EIS seems to be less than 1. Bewley models with EGS<1 generate too much correlation between earnings and income, too little persistence, and too little wealth to earnings ratio. BBB models with EGS<1 cannot replicate a negative fertility-income relationship for asset poor individuals. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Related Literature Papers on fertility + inequality + warm glow altruism: no assets: Kremer and Chen (2002), de la Croix and Doepke (2003). assets: Moav (2005), Scholz and Seshadri (2009). Fertility + deterministic inequality + altruism: Bosi, Boucekkine and Seegmuller (2011). Fertility+altruism+inequality+no assets: Mookherjee, Prina and Ray (2012) and Cordoba & Liu (2013). Fertility+altruism+inequality+assets: Alvarez (1999) Negative results: lack of persistence Optimal contracts: Hosseini, Jones, Shourideh (2012) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.1. Theory Bewley model with endogenous fertility Individual’s problem c1 σ + βn1 σ c, b , n 1 V (b; ω ) = Max 0 (1 + r ) b + ω (1 λn) α E V (b 0 ; ω 0 )jω c + nb 0 , b 0 0; n 2 [0, 1/λ]. b =parental "transfers, bequest, inheritance" ω parent’s ability, ω 0 children’s ability ω0 F (ω 0 jω ) (Markov process) Solution: c = c (b, ω ), n = n(b, ω ), b 0 = b (b, ω ). Alvarez (1999) studies a related "household" version of this problem where ω 0 is the same for all children. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.1. Theory Lack of persistence property (Alvarez) First order conditions: b 0 : ! u 0 (c ) βn n :! u 0 (c ) b 0 + ωλ α β (1 E Vb (b 0 ; ω 0 )jω α) n α E V (b 0 ; ω 0 )jω (1) (2) Dividing (2) by (1): b 0 + ωλ (1 α) E [V (b 0 ; ω 0 )jω ] E [Vb (b 0 ; ω 0 )jω ] 0 b 8= 9 0 = 0 ! No persistence! < b (ω ) in interior solution ! ∂b = ∂b 0 if constraint for b binds : ; b (b, ω ) if upper constraint for n binds Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.1. Theory Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.1. Theory Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.1. Theory Persistence through abilities Lack of persistence of inheritances is a big problem. Evidence suggest large persistence (Mulligan, Piketty and Saez, Clark, etc) Since bt +1 = b (ω t ) , persistence of inheritances could be driven by persistence of abilities, ω. Can this channel generate enough plausible persistence? We need a calibrated version. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Calibrating an Intergenerational Bewley model with n=1. Suppose ln ω 0 = ρ ln ω + e, e N (0, σ2ω ) Use Tauchen Method to discretize and create F (ω 0 jω ). Earnings(e)= ω, Income(i)=ω + rb, Wealth(b)= b. p (b, ω ) =mass of population with wealth b and ability ω. Wealth-Ability Distribution: pt +1 (b 0 , ω 0 ) = ∑ ∑ ω fb:b 0 =b (b,ω )g pt (b, ω )F (ω 0 jω ) 5 Parameters: [σ, β, r , ρ, σω ] . 5 targets? 1/σ is the Elasticity of Intergenerational Substitution (EGS): it describes the willingness to substitute consumption across generations (not time). Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Parameter Values for Exogenous Fertility Model Parameter r = Interest rate σ = curvature utility function β = discount factor ρ ω = persistence log ability σω = dispersion log ability Value 2.00 0.15 0.32 0.50 1.05 Target An annual interest rate of 4.5 for 25 years Gini wealth earnings income correlation Persistence of hourly wages (Mulligan) Gini earnings Data 2.00 0.82 0.84 0.50 0.64 Model 1 2.00 0.80 0.88 0.50 0.60 A high EGS (= 6.66) is required to replicate earnings-income correlation. β is closed to 1 1 +r . Bewley models needs β (1 + r ) < 1 to avoid unbounded savings. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Performance Exogenous Fertility Model Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(b)/average(e) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 Model 1 1.66 1.49 2.29 0.20 0.63 0.60 0.58 0.50 0.71 0.72 0.73 0.21 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Generational and intergenerational correlations Exogenous Fertility Model Variables in logs ωp ep ip bp cp lsp ωc fp ec ic bc cc lsc fc ωp 100 ep 100 100 ip 78 78 bp -6 -6 44 100 cp 72 72 100 52 100 lsp 0 0 0 0 0 100 fp 0 0 0 0 0 0 0 ωc 48 48 37 -3 34 0 0 100 ec 48 48 37 -3 34 0 0 100 100 ic 57 57 75 35 75 0 0 81 81 bc 48 48 89 68 92 0 0 23 23 68 100 cc 57 57 78 40 79 0 0 77 77 100 73 100 lsc 0 0 0 0 0 100 0 0 0 0 0 0 100 fc 0 0 0 0 0 0 0 0 0 0 0 0 0 100 100 100 Cordoba, LiuCorrelations & Ripoll computed generating a random sample of 300,000 drawsIowa University and University fromState the steady state distribution of of Pittsburgh Fertility, Social anda Long Inequality with non-zero components was used. This reduced the sample to the Mobility, model. Only subsetRun of observations 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Parameter Values for Exogenous Fertility Model 2 σ=1.5 Parameter r = Interest rate σ = curvature utility function β = discount factor ρ ω = persistence log ability σω = dispersion log ability Value 2.00 1.50 0.12 0.50 1.05 Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Target An annual interest rate of 4.5 for 25 years Gini wealth earnings income correlation Persistence of hourly wages (Mulligan) Gini earnings Data 2.00 0.82 0.84 0.50 0.64 Model 1 Model 2 2.00 2.00 0.80 0.80 0.88 0.97 0.50 0.50 0.60 0.60 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Performance Exogenous Fertility Models Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(w)/average(e) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 Model 1 Model 2 1.66 1.66 1.49 1.57 2.29 2.59 0.20 0.25 0.63 0.46 0.60 0.60 0.58 0.59 0.50 0.50 0.71 0.62 0.72 0.60 0.73 0.65 0.21 0.08 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Findings: exogenous fertility model Exogenous fertility model produces signi…cantly persistence of wealth and consumption... but low dispersion of earnings, income and wealth Explaining inequality requires high β (β ! curvature (σ ! 0) of the utility function 1 1 +r ) and low Otherwise precautionary savings are too large, correl (e, i ) is too strong, persistence falls signi…cantly and mean(b)/mean(e) is too low. Existing calibration typically use σ = 1.5 (CDR 2003, Restuccia & Urrutia 2004) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Calibration Alvarez endogenous fertility model Wealth-Ability Distribution: pt +1 (b 0 , ω 0 ) = 1 nt ∑ ω ∑ fb:b 0 =b (b,ω )g pt (b, ω )n(b, ω )F (ω 0 jω ) where nt = ∑ω,b pt (b, ω )n(b, ω ) is average population growth. 8 Parameters: [σ, β, r , ρ, σω ] + [α, λ] . Need 2 more targets. Additional targets: E [n] = 1 (similar to U.S. total fertility rate per-capita) Coe¢ cient of variation n = 0.6 (as suggested by Jones and Tertil, 2008) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Parameter Values for Alvarez Model 3 Parameter r = Interest rate σ = curvature utility function β = discount factor α = curvature altruism λ = cost of a child ρ ω = persistence log ability σω = dispersion log ability Value 2.00 0.72 0.25 0.57 0.40 0.50 0.85 Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Target An annual interest rate of 4.5 for 25 years Gini wealth earnings income correlation average fertility coefficient of variation fertility Persistence of hourly wages (Mulligan) Gini earnings Data 2.00 0.82 0.84 1.00 0.60 0.50 0.64 Model 3 2.00 0.84 0.78 1.00 0.52 0.50 0.59 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Performance Exogenous Fertility Model vs Alvarez's Model Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(w)/average(e) income elasticity of fertility Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 -0.20 Model 1 1.66 1.49 2.29 0.20 0.63 0.60 0.58 0.50 0.71 0.72 0.73 0.21 0.00 Model 3 1.49 1.47 2.90 0.16 0.74 0.57 0.56 0.35 0.58 0.27 0.59 0.30 -0.21 Problematic predictions: very low correlation between earnings and consumption, strong positive association between wealth and fertility, and therefore negative association between labor supply Cordoba, Liu & Ripoll State University and University of Pittsburgh and wealth. Negative association between Iowa consumption and labor Fertility, Social Mobility, and Long Run Inequality 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Generational and intergenerational correlations Alvarez Model Variables in logs ωp ep ip bp cp lsp fp ωc ec ic bc cc lsc ωp 100 ep 73 ip 93 46 100 bp 28 -28 58 cp 83 27 98 72 100 lsp 26 85 -8 -62 -27 100 fp -48 -89 -12 65 9 -90 100 ωc 48 36 45 13 40 13 -23 100 ec 19 15 18 5 16 6 -10 91 100 ic 63 46 58 17 52 17 -30 98 83 100 bc 97 71 92 28 82 25 -46 47 18 61 cc 71 52 66 20 59 19 -34 95 76 99 69 lsc -34 -25 -33 -10 -29 -8 16 43 76 28 -35 17 100 fc 27 19 25 8 23 6 -12 -68 -87 -53 27 -42 -86 fc 100 100 100 100 100 Correlations computed generating a random sample of 300,000 draws from the steady state distribution of the model. Only a subset of observations with non-zero components was used. This reduced the sample to 65129 observations. Cordoba, Liu & Ripoll Iowa State University and University of Pittsburgh Some problematic predictions: very low correlation between Fertility, Social Mobility, and Long Run Inequality 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Jones-Schoonbroodt (negative utility) version Parameter Values for Jones-Schoonbroodt Model 4 σ = 1.5 Parameter r = Interest rate σ = curvature utility function β = discount factor α = curvature altruism λ = cost of a child ρ ω = persistence log ability σω = dispersion log ability Value 2.00 1.50 0.16 1.80 0.30 0.50 1.00 Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Target An annual interest rate of 4.5 for 25 years Gini wealth earnings income correlation average fertility coefficient variation fertility Persistence of hourly wages (Mulligan) Gini earnings Data 2.00 0.82 0.84 1.00 0.60 0.50 0.64 Model 4 2.00 0.81 0.74 1.00 0.52 0.50 0.61 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration Performance Exogenous Fertility Model vs Alvarez vs JS Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(w)/average(e) income elasticity of fertility Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 -0.20 Model 1 Model 3 Model 4 1.66 1.49 1.68 1.49 1.47 1.59 2.29 2.90 3.05 0.20 0.16 0.08 0.63 0.74 0.73 0.60 0.57 0.60 0.58 0.56 0.58 0.50 0.35 0.14 0.71 0.58 0.60 0.72 0.27 0.23 0.73 0.59 0.60 0.21 0.30 0.21 0.00 -0.21 -0.12 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 2.2. Calibration The Negative Fertility-Ability Relationship and the EGS. First order condition: u 0 (1 + r ) b + ω = (1 α) n α n b 0 + λω b 0 + ωλ E V (b 0 ; ω 0 )jω In the i.i.d. case with b 0 = b = 0: λ (1 nλ) σ ω1 σ = (1 α) n α E V (b 0 ; ω 0 ) Ability a¤ects the marginal cost but not the marginal bene…t. σ 2 (0, 1) , or EGS > 1 is necessary for the the marginal cost to be increasing with ability. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.1. Theory Extended Model V (b; ω ) = u (c ) + Φ(n)E V (b 0 ; ω 0 )jω c + nb 0 + λ(n)ω, b 0 s.t. (1 + r ) b + ω n Max c , b0, n 0 2 f0, 0.5, 1, 1.5, 2, ..., nmax g Φ(n) =degree of altruism: Φ(0) = 0, Φ0 (n) > 0, Φ00 (n) < 0. λ(n) = time cost of n children Three channels to recover persistence: Non-isoelastic altruistic function: Φ(n). Non-constant costs of raising children: λ(n). Discrete number of children. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.1. Theory First order conditions for n and b 0 : E [V (b 0 ; ω 0 )jω ] b 0 + ωλ0 (n) = εn (n) | {z } E [Vb (b 0 ; ω 0 )jω ] | {z } 0 MRT (b ,n ) MRS (b 0 ,n ) where ε(n) = Φ0 (n) Φ(nn ) . In Barro-Becker both MRT = independent of n. ∂b 0 ∂n and MRS = ∂b 0 ∂n are Persistence is recovered if MRT increases with n or MRS decreases with n (Alvarez 1999). In both cases parents are less willing to use fertility as a way to obtain welfare as n increases. Some persistence is also recovered if fertility is discrete. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Calibration Extended Model Functional forms Altruistic weight: Φ (n ) = ε(nt ) = βχn 1 + χn 1 1 + χn (elasticity decreases with n). Cost of children: λ (n ) = λ (n + η ) θ λ η θ , 0 < θ < 1. 1 /θ λ (0) = 0, λ (nmax ) = 1 where nmax = 1/λ + η θ η. λ0 (n) > 0 and λ00 (n) < 0 if θ 2 (0, 1) (realistic and necessary). θ 1 λ0 (0) = λθη < ∞ if η > 0 (important!) calibrated model with η = 0 and θ < 1 implies an even higher cost of children than in Alvarez model. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Parameter Values - Full Model Parameter r = Interest rate σ = curvature utility function β = discount factor χ = curvature altruism λ = parameter cost of children η = parameter cost of children θ = elasticity cost of children ρ ω = persistence log ability σω = dispersion log ability Value 2.00 0.70 0.39 2.00 3.05 1.84 0.23 0.50 0.90 Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Target An annual interest rate of 4.5 for 25 years Gini wealth earnings income correlation average fertility average cost per child (per parent) time cost of first child (per parent) times cost of first two children (per parent) Persistence of hourly wages (Mulligan) Gini earnings Data 2.00 0.82 0.84 1.00 0.30 0.20 0.35 0.50 0.64 CL Model 2.00 0.85 0.75 1.00 0.27 0.20 0.37 0.50 0.58 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Performance CL Model vs Alvarez Model Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(b)/average(e) income elasticity of fertility wealth elasticity of fertility Average cost per child Efficient number of children Maximum number of children per couple Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 -0.20 CL 1.49 1.53 3.29 0.21 0.80 0.60 0.59 0.37 0.65 0.51 0.65 0.24 -0.18 0.08 0.27 0.66 7.00 Alvarez 1.44 1.40 2.43 0.28 0.73 0.57 0.56 0.44 0.58 0.29 0.59 0.26 -0.21 0.19 0.40 0.60 5.00 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Endogenous Discounting and the Accumulation of Wealth First order condition with respect to assets: u 0 (c ) = Φ (n ) n Φ (n ) (1 + r ) E u 0 c 0 n is the average discount factor. Higher fertility reduces average discounting and incentives to save. Krussell and Smith (1998). Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration The Negative Fertility-Wealth Relationship Fertility goes to 1 for everyone rather than to the maximum. Why does fertility decrease with inheritance for certain individuals in our model? Because of credit constraints In addition to the pure wealth e¤ect there is a substitution e¤ect from inheritances Individuals who receive low inheritances may …nd themselves credit constrained if their ability shock is low enough For those individuals, higher inheritance reduces the shadow price of credit (the e¤ective interest rate) By arbitrage, the implicit rate of return to children is tied to the rate of return to savings A lower rate of return reduce the incentives to save in the form of children Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Performance CL Model vs Alvarez Model Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini Income Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(b)/average(e) income elasticity of fertility wealth elasticity of fertility Average cost per child Efficient number of children Maximum number of children per couple Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Data 3.60 4.32 6.02 0.48 0.57 0.58 0.32 0.48 0.71 0.5-0.8 0.77 6.80 -0.20 CL 1.49 1.53 3.29 0.21 0.80 0.60 0.59 0.37 0.65 0.51 0.65 0.24 -0.18 0.08 0.27 0.66 7.00 Alvarez 1.44 1.40 2.43 0.28 0.73 0.57 0.56 0.44 0.58 0.29 0.59 0.26 -0.21 0.19 0.40 0.60 5.00 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions 3.2. Calibration Generational and intergenerational correlations CL model Variables in logs ωp ep ip bp cp lsp fp ωc ec ic bc cc lsc ωp 100 ep 97 ip 61 53 100 bp -16 -21 49 cp 49 40 99 62 lsp 70 86 26 -27 14 100 fp -84 -88 -36 43 -21 -79 ωc 41 40 25 -7 20 28 -35 100 ec 35 34 15 -12 10 24 -29 96 100 ic 61 58 68 17 62 39 -50 83 73 100 bc 58 54 93 49 91 35 -45 24 12 65 cc 63 60 74 22 69 40 -51 77 65 99 72 lsc 13 11 -12 -18 -15 6 -10 61 80 30 -16 22 100 fc -12 -11 17 24 20 -8 10 -79 -87 -47 21 -37 -79 fc 100 100 100 100 100 100 100 Cordoba, LiuCorrelations & Ripoll computed generating a random sample of 300,000 drawsIowa University and University fromState the steady state distribution of of Pittsburgh Fertility, Social anda Long Inequality with non-zero components was used. This reduced the sample to the Mobility, model. Only subsetRun of observations 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Experiments Experiment 1: One and two child policies One child policy: a policy that restricts n perpetual growth into this economy. Given that r is exogenous, Φ (n ) n 0.5 brings (1 + r ) > 1 for n For bounded wealth, Bewley models needs Φ (n ) n 0.5. (1 + r ) < 1 We consider instead a two children policy instead: n 1. Experiment 2: Estate taxes Suppose a estate tax such that individuals returns are 1 + r /2 instead of 1 + r . Collected taxes are used to …nance exogenous government expenses. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Experiments Two Children Policy (TCP) and Estate Taxation (ET) Coefficient variation earnings Coefficient variation income Coefficient variation wealth Correlation earnings-wealth (e,b) correlation income-wealth (i,b) Gini earnings Gini income Gini wealth Gini consumption persistence earnings persistence income persistence wealth persistence consumption average(b)/average(e) average fertility income elasticity of fertility wealth elasticity of fertility Mean(e) Mean(i) Mean(b') Mean(c) stdev(e) stdev(i) stdev(b') stdev(c) Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Benchmark 1.49 1.53 3.29 0.21 0.80 0.58 0.60 0.85 0.59 0.37 0.65 0.51 0.65 0.24 1.00 -0.18 0.08 0.99 1.32 0.24 1.33 1.48 2.02 0.68 1.99 TCP* 1.46 1.44 2.42 0.20 0.81 0.57 0.58 0.81 0.58 0.49 0.71 0.67 0.72 0.29 0.92 -0.10 0.01 1.06 1.54 0.31 1.56 1.55 2.22 0.76 2.22 ST 1.49 1.49 3.90 0.15 0.46 0.57 0.57 0.93 0.56 0.45 0.54 0.15 0.55 0.08 0.97 -0.09 0.09 1.06 1.11 0.09 1.11 1.58 1.66 0.37 1.59 Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Conclusions Properly calibrated BBB model predict: realistic negative fertility-income relationship a negative fertiliy-wealth relationship at certain levels of wealth higher inheritance rates by smaller families more wealth concentration and wealth dispersion than exogenous fertility models realistic persistence Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Conclusions Family planning policies: One child policy may induce perpetual growth Two children policy: increases average earnings, income, consumption and wealth but also their dispersion and persistence, particularly persistence of wealth: reduces social mobility. Estate taxation: Reduces average income, consumption and wealth and their dispersion. Reduces fertility Increases social mobility Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh 1. Introduction 2. Horse Race Existing Models 3. A Better Model 4. Policy Experiments 5. Conclusions Conclusions EGS seems to be di¤erent from EIS EGS larger than 1 and probably between 1.5 to 2. EIS seems to be less than 1. Cordoba, Liu & Ripoll Fertility, Social Mobility, and Long Run Inequality Iowa State University and University of Pittsburgh