Eco-Labeling and the Gains from Agricultural and Food Trade: A Ricardian Approach

Transcription

1 Eco-Labeling and the Gains from Agricultural and Food Trade: A Ricardian Approach Kari Heerman (ERS/USDA) Ian Sheldon (Ohio State University) Jihyun Eum (Ohio State University) Seminar University of Wyoming April 21, 2017 Laramie, WY

2 Motivation Analysis of agricultural system should recognize extent of vertical product differentiation, e.g., environmental claims (Sexton, 2013) Eco-labeling key to resolving information asymmetry associated with environmental credence goods Rapid growth of eco-labeling relating to food and agricultural products since 1970s (Gruére, 2013) Trade often expected to generate negative externalities (Copeland and Taylor, 2004) However, if production generates environmental benefits, eco-labeling beneficial (Swinnen, 2015)

3 Outline Develop Ricardian-type model drawing on Eaton and Kortum (2002), and others including, inter alia, Waugh (2010), Fieler (2011), Levchenko and Zhang (2014) Class of model already applied to agricultural trade by Reimer and Li (2010), Reimer (2015), and Heerman et al. (2015) Use to derive comparative statics concerning impact of labeling of and trade in eco-friendly products Lay out recipe for calibrating model, and initial estimation of gravity equation

4 Modern Ricardian Approach Difficult to adapt Dornbusch, Fischer and Samuelson (1977) to multi-country setting Contribution of Eaton and Kortum (2002): focus on parameters of productivity distribution Given country will be more productive than others at producing range of goods in continuum generates reason for trade Generates gravity-like structure between share of spending on imports and trade costs (Arkolakis, Costinot and Rodriguez-Clare, 2012)

5 Model I countries trade products j, produced along continuum, producers having access to LC and EF: q i LC j = z i j L i q i EF j = z i j L i α H i 1 α z i j distributed independently as Fréchet: F i z = exp T i z θ Prices offered by exporter i in n: p LC ni j = r iτ ni z i (j) p ni EF j = κr i α w i 1 α τ ni ζ ni z i j

6 Model Consumers in n buy LC and EF products at lowest price on offer: p n k j = min i p k ni Productivity distribution used to derive distributions of EF price offers by i in n, and prices of EF products offered in n : G EF ni p = 1 exp T i κr α i w 1 α θ i τ ni ζ ni p θ j G n EF p = 1 exp Φ n EF p θ where: Φ EF I n = l=1 T l κr l α w l 1 α τ nl ζ nl θ

7 Model Setting α = ζ ni = 1 : G LC ni p = 1 exp T i r i τ θ ni p θ G LC n (p) = 1 exp Φ LC n p θ where: Φ LC I n = l=1 T l r l τ nl θ Φ n k, k=ef,lc describe how average productivity, input costs, trade and labeling costs around world affect prices of each type of good in each import market Lower trade costs allow consumption with smaller environmental impact, even without reallocation of consumption to EF products

8 Model Using price distributions, probability i offers lowest prices of EF and LC products in n: π EF ni = T i κr α i w 1 α θ i τ ni ζ ni EF Φ n π LC ni = T i r i τ θ ni LC Φ n With continuum, these are also fraction of products that consumers in n purchase from i: k X ni k X = π k 1 ni 0 Q k j dj k pdgn 0 p k 1 n Q k π j dj k ni (1) pdgn 0 p 0

9 Model Consumers have preferences over products, choosing EF and LC to maximize: σ σ q i LC j σ 1 σ dj + ω i 1 σ 0 1 q i EF j σ 1 σ Implies total expenditure on EF relative to LC: X i EF X i LC = ω i P i EF P i LC 1 σ P i k is CES price index, Pi k = γφ n k 1/θ, k = LC, EF consumers only choose EF if labeled dj

10 Comparative Statics: Labeling Labeling increases EF trade flows: (i) Labeling increases share of EF expenditure on imports: π EF nn = T n κr α 1 α n w θ n EF = Φ n I l=1 T n κr n α w n 1 α θ T l κr l α w l 1 α τ nl ζ nl θ Without labeling ζ ni =, consumers do not recognize imported EF as distinct from LC products, therefore: Φ EF n = T n κr α 1 α n w θ n and π EF nn = 1 As labeling costs fall, Φ n EF increases and π nn EF falls, i.e., import share of expenditure on EF products rises

11 Comparative Statics: Labeling (ii) Labeling increases share of total expenditure allocated to EF products: By definition, X i = X i EF + X i LC, therefore: X i EF = ω i X i 1 + ω i p i EF p i EF p i LC 1 σ p i LC 1 σ Recall p n EF = γφ n EF 1/θ, so lower labeling costs implies lower prices for EF products Therefore, since lower labeling costs have no impact on Φ LC n, introducing EF labels lowers p EF LC i p i

12 Comparative Statics: Land and EF Optimal land allocation implies: EF L i LC L = n π EF EF ni X n LC EF i X n X n n π ni Already established that π EF ni increases with eco labeling, as does share of expenditure allocated to EF X n X n EF is also decreasing in import markets where labeling of i s EF products is introduced Therefore, share of land allocated to EF production increases for exporter i

13 Comparative Statics: Mutual recognition Recognition of i s labeling in n implies: Φ n EF π EF ni = T α 1 α θ i κr i wi τni ΦEF n = T n κr n α w n 1 α θ + T i κr i α w i 1 α τ ni θ + l i,n T l κr l α w l 1 α τ nl ζ nl θ Φ n EF increases, and given: Φ n EF Φ n LC = p EF i LC p i Relative price of EF products declines, EF trade flows increase for fixed level of expenditure θ

14 Model: Solution and parameterization Given T i, τ ni, ζ ni, H i and ω i, equilibrium is r i, w i, π LC ni, π EF ni, X LC i, X EF i and L LC i, L EF i, such that input markets clear and trade is balanced Solve for LC-type equilibrium variables, obtaining land rental rate r i, and then solve for equilibrium w i, and EF-type equilibrium values Parameterization/calibration requires values for T i, θ, τ ni, ζ ni, σ, and ω i Standard approach: log-linearize (1) and estimate gravity-like equation to get, T i, and τ ni, use values of θ and σ from literature, and solve for ζ ni and ω i

15 Model: Solution and parameterization Table 1: Key Parameters α Land s value-added share in organic 0.65 (OECD, 2009) production (1-average labor share of valueadded) w i Solve out assuming H i =1 for all countries Calibrate r i Country s agricultural output/hectare of World Bank (2012) arable land T i Mean parameter for productivity distribution Estimate θ Dispersion parameter for productivity 2.83 (Reimer and Li, 2010) distribution τ ni Bilateral trade costs Estimate ζ ni Organic labeling costs in market n in excess of exporter i s labeling costs Calibrate σ Elasticity of substitution 1.5 (Ruhl, 2008) ω i Consumer love of sustainability Calibrate

16 Model: Solution and parameterization Normalize π LC ni by π LC nn : LC LC -θ π ni X ni Ti ( riτ ni ) Ti ri = = = LC LC -θ π nn X nn Tn( rn ) Tn rn and taking the log: LC X ni Ti ri ln = ln -θ ln -θ lnτ LC ni X nn Tn rn Following Reimer and Li (2010), define: S i = ln T i θln(r i ) -θ τ -θ ni

17 Model: Solution and parameterization Substituting S i in for T i : LC X ni ln = -θ ln τ ni + S i - S LC n X nn Gravity-like structural relationship in LC: ln X LC ni LC = S i θ b ni + l ni + RTA ni + X nn where: c d cni + ex i S n θln τ ni = b ni + l ni + RTA ni + c d cni + ex i + ξ ni

18 Gravity Equation Estimates Variable D1 (0,375) *** (0.50) *** (0.45) D2 (375,750) *** (0.30) *** (0.28) D3 (750, 1500) *** (0.20) *** (0.20) D4 (1500, 3000) *** (0.14) *** (0.15) D5 (3000, 6000) *** (0.08) *** (0.09) D6 (6000, max) *** (0.17) *** (0.12) Border 1.30*** (0.45) 1.01*** (0.41) Language 1.35*** (0.18) 1.30*** (0.19) RTA 2.88*** (0.21) 3.35*** (0.21) Adjusted R Sample-size 11,955 12,099 *** Significant at 1 percent level

19 lnt i (2010, θ=2.83)

20 lnt i (2013, θ=2.83)

21 Next Steps Use parameterized model to evaluate impact of alternative eco-labelling policies: - Mutual recognition - Regulatory harmonization Allow for non-homothetic preferences to explore impact of income differences across i (Fieler, 2011), i.e., North vs. South and differing standards Introduce explicit environmental damage function Use pesticide standards to pin down weight ω i on consumer preferences in utility function