## _wp04231

## Source details

**Canonical URL:** [_wp04231](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04231.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04231.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04231.pdf.json)

---

### I. Introduction — purpose and approach
- Research questions:
  - What are the sources of trade between the developed world (the North) and developing countries (the South)?
  - How are the gains from trade distributed?
  - How does trade affect factor prices?
- Core hypothesis: institutional differences are a key source of comparative advantage in North–South trade.
- Operational definition of institutions: contracting institutions that govern relationships between private economic parties (contract enforcement, property rights, investor protection).
- Two modelling approaches:
  - Ricardian view: institutions as productivity differences (benchmark).
  - Grossman–Hart–Moore view (preferred): institutions govern relationships via contract incompleteness, creating distortions parameterized in the Caballero and Hammour (1998) style.
- Empirical strategy: use U.S. import shares by country and 4-digit SIC industry; augment Romalis (2004) factor-content specification with industry institutional dependence and country institutional quality.

### II. Basic model structure (closed economy benchmark)
- Economy: two factors (K and L) and three goods (K-good, L-good, M-good).
- Preferences: U(C_K, C_L, C_M) = C_K^α C_L^β C_M^γ with α, β, γ positive and α + β + γ = 1.
- Numeraire: ideal price index P ≡ p_K^α p_L^β p_M^γ = 1.
- Production:
  - K-good: linear in K with productivity a.
  - L-good: linear in L with productivity b.
  - M-good: Leontief combining one unit of L and x units of K to produce y units; M is institutionally dependent.
- Ricardian institutions (benchmark): fractional loss τ of M output; better institutions → lower τ.
- Resource allocation summary (given E = share of labor employed in M-sector):
  - X_M = (1 − β) y E L
  - X_L = b (1 − E) L
  - X_K = a (K/L − x E) L
- Closed-economy equilibrium: set ⟨p_K, p_L, p_M, r, w, E⟩ satisfying first-order and market-clearing conditions.

### III. International extension — Ricardian interpretation and implications
- Two countries (North N and South S); negligible transport costs; K_N, L_N, K_S, L_S.
- Institutional productivity losses: β_N and β_S fractions of M output lost (β_N < β_S).
- With β_N = β_S model reduces to Heckscher–Ohlin; unequal β creates a sectoral Ricardian productivity difference (H–O–R special case).
- Integrated equilibrium and Factor Price Equalization (FPE) set:
  - Integrated equilibrium solved for world factor quantities.
  - In integrated equilibrium the North’s superior institutional technology is used to produce M; M-sector located in North for FPE-consistent endowments.
- Pattern under trade (FPE example point A):
  - South stops producing M and allocates endowment to K-good and L-good.
  - North expands M-sector labor to meet world demand.
- Welfare and gains from trade (Ricardian baseline):
  - Welfare measured by real income; autarky welfare of factor owners in country i is w_i L_i + r_i K_i.
  - Symmetric K/L case: trade prices equal Northern autarky prices (w_T = w_N, r_T = r_N); Northern aggregate welfare unchanged; gains accrue entirely to the South (w_T > w_S and r_T > r_S).
  - More general factor-proportions: standard H–O effects apply, but South always benefits relatively because trade confers access to North’s superior institutional technology.
- Summary: when institutions act as sectoral productivity differences, trade “bails out the South”: South gains more, North may be unchanged.

### IV. Grossman–Hart–Moore (GHM) specification — modeling and mechanisms
- Modelling of contract incompleteness:
  - When two parties invest jointly, fraction δ of capital investment is relationship-specific, producing hold-up and underinvestment.
  - δ captures contract enforcement/property-right quality; better institutions = lower δ; δ = 0 is frictionless.
- Participation/rationality condition (Nash bargaining split of surplus): K enters M production only if participation inequality in text holds (notation preserved in source).
- With contractual incompleteness:
  - Labor is segmented; rewards differ across sectors.
  - L earns rents in the M-sector of size δ r x.
  - Equilibrium is inefficient: underinvestment in M, lower E, w, and r relative to efficient case; effects monotonic in δ.
- Trade pattern and integrated equilibrium:
  - North with δ_N < δ_S produces M at lower price because satisfying K’s participation condition is easier.
  - Integrated equilibrium: M produced under North’s institutional setting; South may stop producing M.
  - FPE still useful but factor rewards equalize only within sectors; cross-sector rewards differ.
- Welfare implications (GHM view):
  - Symmetric K/L case: trade inherits North autarky prices and allocation (E_N = E_T); Northern base wages and r unchanged (w_N = w_T, r_N = r_T).
  - Northern gains from trade arise from shift into high-paying M-sector jobs: total labor rewards W_NT = w_T L_N + δ_N x r_T E_N L_N.
  - Southern autarky: W_SA = w_S L_S + δ_S x r_S E_S L_S; under trade: W_ST = w_T L_S; capital r_S K_S becomes r_T K_S.
  - Net labor change in South: W_ST − W_SA = ((wT − wS) − δS x rS ES) L_S — this can be negative; the country as whole may lose from opening.
  - Capital rewards equalize (r_T > r_S ⇒ capital wins); labor faces opposing forces: base wage convergence vs. loss of M-sector rents.
- Intuition and policy-relevant cases:
  - Trade-induced institutional improvement raises opportunity costs w and r (first effect) but often moves M-sector out of South (second effect) causing loss of M rents.
  - If δ_S ≈ δ_N + Π with small Π, South likely loses due to M-sector loss despite negligible aggregate institutional improvement.
  - If δ_N = 0 (North perfect institutions), trade can deliver first-best aggregate welfare to South despite loss of M rents.
- Equilibrium outside FPE:
  - w_NT = w_ST = b p_LT and r_NT = r_ST = a p_KT.
  - North produces integrated M quantity when p_M y = w_T + (1+δ_N) x r_T (North equality); South cannot produce when p_M y < w_T + (1+δ_S) x r_T.
  - Outside FPE some M-production may persist in South for extreme endowments, but South’s high-paying M-jobs shrink relative to autarky.

### V. Factor prices, factor movements, and political economy implications
- Contrasts with standard factor-abundance predictions:
  - Conventional: South (capital-scarce) opening ⇒ capital return decreases, wages increase; North opposite.
  - GHM model: Southern capital benefits from opening; Southern labor may lose. North: capital return unchanged; labor rewards increase.
- Migration and capital flows intuition:
  - Model rationalizes immigration pressure (labor seeks North for M-jobs) and limited capital flows South (institutional frictions limit capital attraction).
- Institutional choice model (political economy):
  - Policymaker maximizes G(δ,c) = V S(δ) + (1−V) c with V ∈ (0,1); interest group (labor) maximizes S_L(δ) − c.
  - Aggregate welfare S(δ) = r(δ)K + (w(δ) + δ x r(δ) E(δ)) L, maximized at δ = 0 and dS/dδ < 0.
  - Equilibrium δ* solves argmax δ∈[0,1] of weighted welfare (equation (13) preserved in original notation).
  - Comparative statics: better equilibrium institutions when 1) low corruption (higher V); 2) higher K/L.
  - Under autarky, bad institutions can persist because L can bribe policymaker to raise δ and capture rents (δ* redistributive and lowers aggregate welfare).
- Institutional choice under trade — “race to the top”:
  - With trade, best response dynamics yield unique equilibrium δ_N = δ_S = 0 (perfect institutions).
  - Each country has incentive to improve institutions marginally to capture M-sector rents; trade can force institutional improvement even if V = 0.
  - Contrasts with “race to the bottom” narratives; consistent with empirical findings of trade’s positive effect on institutions.

### VI. Empirical test — testable predictions, data, and main results
- Empirical prediction: countries with better institutions capture larger U.S. import shares in industries that are more institutionally dependent.
- Data and measures:
  - 1998 U.S. imports (4-digit SIC): initial universe 177 countries and 389 industries; final sample 117 countries and 389 industries where all data available.
  - Country institutional quality: Kaufmann, Kraay and Zoido-Lobaton (2002) index (range −2.5 to 2.5).
  - Capital and skill endowments: Hall and Jones (1999).
  - Industry institutional dependence (proxy): Herfindahl index of intermediate input use from U.S. Input–Output Use Table, 1992; Herfindahl multiplied by −1 so measure increases with institutional intensity.
  - Industry controls: capital intensity, skill intensity from U.S. Manufacturing database (1992); alternative institutional intensity measures used for robustness (share20, gini, number of intermediates, investment/output).
- Baseline regression specification (text-preserved form):
  - rel_share_ic = ϑ + K1 inst_dep_i * inst_c + K2 skint3_i * skill_c + K3 capint3_i * capital_c + Λ_c + Ν_i + Π_ic
  - rel_share_ic: country c’s share in U.S. imports in sector i normalized by average country i share.
- Main empirical findings:
  - (herfindahl index)*inst coefficient ≈ 2.51 with standard error (0.68)*** (column (1) Table 1).
  - (herfindahl index)*inst coefficient ≈ 2.36 with standard error (0.64)*** (column (2) Table 1).
  - Practical magnitude: moving from 25th to 75th percentile in institutional quality implies predicted relative import share decreases by 0.09 at the 25th-percentile institutional intensity good and increases by 0.18 at the 75th-percentile institutional intensity good (text statement).
  - skill intensity * skill endow coefficient ≈ 12.34 (0.53).
  - capital intensity * cap endow coefficient ≈ 0.53 (0.30)*.
  - Observations: 31,568; Industries: 389; Countries: 117.
- Robustness:
  - Alternative institutional intensity measures yield positive and significant interaction coefficients (Table 3).
  - Four-factor model including raw materials retains similar institutional interaction effect (Table 4).
  - Controlling for financial dependence and financial development: institutional coefficient similar and significant (Table 4 Column (5)).
  - IV using settler mortality interacted with Herfindahl (Table 4 Column (6)): coefficient similar in magnitude but p-value 23% due to smaller sample (80 countries).
  - South-only subsample: (herfindahl index)*inst ≈ 2.94 (1.12)*** (Table 5); coefficients often larger for South-only.
  - Excluding Africa or SE Asia or dropping industry outliers: coefficients remain similar and highly significant.
  - Institutions vs. per capita income (Table 6): (herfindahl index)*GDPPC coefficient 2.33 (0.61)*** when using log per capita PPP GDP alone; including both GDPPC and inst reduces significance and coefficients roughly halve (corr(inst, income) = 0.82).

### VII. Main conclusions and policy-relevant implications
- Two formalizations of institutional differences produce qualitatively different comparative-static implications:
  1. Ricardian view (institutions as productivity differences): trade bails out the South — South gains most; North unchanged.
  2. Grossman–Hart–Moore view (institutions as contract incompleteness and holdup): North certain to gain; South may lose; labor rewards can diverge and Southern labor may lose while Southern capital may gain.
- Empirical evidence from U.S. import shares by industry and country supports the model prediction that countries with better institutions capture larger import shares in more institutionally dependent (complex) industries: positive and statistically significant interaction between industry institutional intensity and country institutional quality across many specifications and robustness checks.
- Institutional change is slow; the GHM results are particularly relevant for short-run distributions of gains and losses.
- When institutions are endogenized, trade can create incentives for institutional improvement (a “race to the top”), reversing some short-run distributive losses over the long run.

### VIII. Extension: three-party production structure (entrepreneur, capital K, labor L)
- Organisation:
  - Entrepreneurs raise K and establish companies; companies hire workers.
- Institutional specificity parameters:
  - Fraction φ_K of K becomes relationship-specific (capital-market institutions).
  - Fraction φ_C of company value becomes specific when hiring a worker (labor-market/technological features).
- Ex post surplus split: equally between parties in both relationships.
- Entrepreneur outside option fixed at zero; R = (1 + φ_K) r required return on each unit of K to the company.
- Company participation constraint and sector pricing:
  - p_M y = w + (1 + φ_C) R x = w + (1 + φ_C)(1 + φ_K) r x.
- Worker rewards and inefficiency:
  - Workers earn rents in M-sector; outcome inefficient; baseline model is a reduced form of this fuller model.
- Insight: φ_K (capital-market institutional quality) has first-order effect on worker compensation by changing M-sector size and worker rents; differential φ_K and φ_C across North and South generate nuanced comparative advantage outcomes.

### IX. Key empirical and tabulated statistics (as presented)
- Table A2. Industry-Level Summary Statistics (single-line summary per variable):
  - capital intensity 0.61 0.11 0.18 0.95
  - skill intensity 0.11 0.06 0.01 0.48
  - herfindahl index of intermediate use 0.13 0.09 0.04 0.78
- Table A3. Country-Level Summary Statistics (single-line summary per variable):
  - Institutional quality -0.013 0.940 -2.166 1.909
  - log of physical capital per worker 9.241 1.586 5.763 11.589
  - log of human capital per worker 0.584 0.294 0.072 1.215
- Table A1: representative industries identified as Least Institutionally Intensive and Most Institutionally Intensive (industry names and codes listed in source).
- Table A4: economy list classifying countries into North and South following Romalis (2004); North defined as industrial countries with 1995 per capita PPP-adjusted GDP at least 50% of U.S. level (source lists country names).

*Source: _wp04231 (IMF working paper excerpt).*

### 1.        Baseline Specification........................................................................................

### 1.        Baseline Specification

### I. Introduction — purpose and approach
- Research questions:
  - What are the sources of trade between the developed world (the North) and developing countries (the South)?
  - How are the gains from trade distributed?
  - How does trade affect factor prices?
- Core hypothesis: institutional differences are a key source of comparative advantage in North-South trade.
- Operational definition of institutions: contracting institutions that govern relationships between private economic parties (contract enforcement, property rights, investor protection).
- Two modeling approaches presented:
  - Ricardian view: institutions enter as productivity differences (benchmark).
  - Grossman-Hart-Moore view (preferred): institutions govern relationships between factors via contract incompleteness, creating distortions captured through Caballero and Hammour (1998)-type parameterization.
- Empirical strategy:
  - Use U.S. import shares by country and 4-digit SIC industry.
  - Augment Romalis (2004) factor content of trade specification with industry institutional dependence and country institutional quality to test whether countries with better institutions capture higher U.S. import shares in more institutionally dependent sectors.

### II. Basic model structure (closed economy benchmark)
- Economy: two factors (K and L) and three goods (K-good, L-good, M-good).
- Preferences: identical Cobb-Douglas utility
  - U(CK, CL, CM) = C_K^α C_L^β C_M^γ with α, β, γ positive and α + β + γ = 1. (Equations (2)–(4) derive first-order conditions.)
- Numeraire: ideal price index P ≡ p_K^α p_L^β p_M^γ = 1.
- Production technologies:
  - K-good: linear in K; one unit of capital produces a units → price condition rap_K = a (equation (5) form).
  - L-good: linear in L; one unit of labor produces b units → wbp_L = b (equation (6) form).
  - M-good: Leontief combining one unit of L and x units of K to produce y units; M is the institutionally dependent good.
- Institutions in Ricardian view:
  - Institutional imperfection modeled as fractional loss τ of M output (better institutions → lower τ).
  - M-good price condition: p_M = (1 − τ)(y − x r − w) represented in profit condition (equation (7) form).
- Resource allocation summarized by E = share of labor employed in M-sector; given E, K, L, outputs are:
  - X_M = (1 − β) y E L,
  - X_L = b (1 − E) L,
  - X_K = a (K/L − x E) L.
- Market clearing conditions specified by equations (8)–(10). The closed-economy equilibrium is the set ⟨p_K, p_L, p_M, r, w, E⟩ satisfying (2)–(10).

### III. International extension and Ricardian interpretation
- Two countries: North (N) and South (S); negligible transport costs.
- Factor endowments: K_N, L_N, K_S, L_S with K = K_N + K_S and L = L_N + L_S.
- Institutional productivity losses in Ricardian case:
  - Fractions β_N and β_S of M output are lost in North and South respectively (β_N < β_S since North has better institutions).
- With β_N = β_S the model reduces to standard Heckscher-Ohlin; unequal β introduces a sectoral Ricardian productivity difference (Davis (1995) H-O-R special case).
- Integrated equilibrium and Factor Price Equalization (FPE) Set:
  - Integrated equilibrium solved for world factor quantities.
  - FPE set: country endowment partitions that allow replication of integrated equilibrium production pattern via trade (graphically represented in Figure 1).
  - In integrated equilibrium the North’s superior institutional technology is the one used to produce M; hence the M-sector in trade will be located in the North if world factor allocations fall in the FPE set (e.g., point A in Figure 1).
- Pattern of production under trade (illustrated in Figure 2 for point A):
  - South stops producing M; South allocates entire endowment to K-good and L-good.
  - North expands M-sector labor to meet world demand.

### IV. Gains from trade in the Ricardian baseline
- Welfare measure: with numeraire as optimal consumption basket, welfare is proportional to real income; autarky welfare of factor owners in country i is w_i L_i + r_i K_i; gains from trade are differences in factor rewards between trade and autarky.
- Simple symmetric K/L case (K_N/L_N = K_S/L_S = K/L) and trade equilibrium in FPE set (e.g., point B):
  - E_N = E_T, and trade prices equal Northern autarky prices: w_T = w_N and r_T = r_N.
  - Northern aggregate welfare unchanged; gains from trade accrue entirely to the South.
  - In the South both factors unambiguously gain: w_T > w_S and r_T > r_S.
- More general factor-proportions differences:
  - Standard H-O results apply: if North is capital-abundant, capital in North gains while labor loses.
  - However, the South always benefits relatively more because trade confers an effective technology improvement (M produced with superior Northern technology) on the South.
- Summary conclusion for Ricardian specification:
  - When institutional inferiority manifests as lower productivity in institutionally intensive sectors, trade leads to gains for both sides but the South stands to gain more (South gains relative to autarky because it ceases producing M and benefits from North’s superior institutions through trade).

### V. Contrast with Grossman-Hart-Moore approach (preview of preferred specification)
- Motivation: lack of proper contract enforcement causes distortions beyond pure productivity effects; contracts are more incomplete in countries with worse institutions.
- Key qualitative differences previewed:
  - Under contract incompleteness specification, factor market distortions generate industry-level differences in factor rewards even with perfect intersectoral mobility.
  - The institutionally dependent sector pays rents to one factor (labor in the paper’s calibration) — “good jobs”.
  - Under trade the North (with better institutions) specializes in the institutionally dependent good; the South loses its high-paying M-jobs.
  - Implications differ sharply from Ricardian view: North may gain more than South; South may lose overall; in South capital may gain while labor loses.
- Endogenous institutions extension (sketch):
  - Trade increases the cost of bad institutions; opening to trade leads to institutional improvement as countries compete to capture advantageous sectors (a “race to the top” in institutional quality).

### VI. Empirical test (description)
- Empirical question: do countries with better institutions capture larger U.S. import shares in industries that are more institutional-dependent?
- Data and method:
  - U.S. import shares disaggregated by country and 4-digit SIC industry.
  - Factor content trade methodology from Romalis (2004) augmented with industry institutional dependence and country institutional quality.
- Main empirical finding:
  - Institutional differences are a significant determinant of trade flows (detailed results and robustness in later sections).

*Source: _wp04231 - 1.        Baseline Specification (excerpt).*

### conclusion, then, is that trade bails out the South: the institutionally weak country no longer

### _wp04231 - conclusion, then, is that trade bails out the South: the institutionally weak country no longer

### B. Case II: The Grossman-Hart-Moore View of Institutions — modeling and mechanisms
- Modeling approach: follow Williamson (1985), Grossman and Hart (1986), Hart and Moore (1990); when two parties invest jointly, a fraction δ of capital's investment becomes relationship-specific, generating hold-up and underinvestment.
- Interpretation: δ captures quality of contract enforcement and property rights; better institutions = lower δ; limiting case δ=0 returns to frictionless setting.
- Key participation/rationality condition (Nash bargaining, split surplus): K enters M production only if
  - r(1−δ)x + 1/2 s ≥ rx
  - rearranged (as in text): (.)11(.)1(rxwypMφ++≥  (text preserves original notation/placement).
- With contractual incompleteness:
  - Labor (L) is segmented: rewards differ across sectors.
  - L earns rents in the M-sector of size δrx (see equation (12) structure in text).
  - Equilibrium is inefficient: underinvestment in M, lower E, w, and r relative to efficient case; effects monotonic in δ (higher δ ⇒ lower E, w, r).

### Trade pattern and integrated equilibrium
- Institutional comparative advantage: North with δN<δS can produce M at strictly lower price because satisfying K's participation condition is easier.
- Integrated equilibrium: only Northern institutional setting used for M production; South may stop producing M altogether.
- Factor Price Equalization set (FPE) still useful but factor rewards equalized across countries only within each sector; rewards differ across sectors.

### Welfare implications (Grossman-Hart-Moore view)
- When KN/LN = KL/LL = K/L, trade equilibrium inherits North autarky prices and allocation: EN = ET (text statement).
- Northern outcomes:
  - Base wage unchanged: wN = wT (also rN = rT; total reward to capital unchanged).
  - Total rewards to labor under trade: WNT = wT LN + δN x rT EN L (as in text).
  - Northern gains from trade arise because of shift toward high-paying M-sector jobs even if conventional comparative advantage gains are absent.
- Southern outcomes:
  - Autarky labor income: WSA = wS LS + δS x rS ES LS; capital: rS KS.
  - Under trade: labor WST = wT LS; capital rT KS.
  - Net labor change in South:
    - WST − WSA = ((wT − wS) − δS x rS ES) LS
    - This difference could be negative; for some parameters the country as a whole may lose from opening.
- Factor price convergence/divergence:
  - Capital rewards equalize (rT > rS ⇒ capital unambiguously wins).
  - Labor faces opposing forces: base wage convergence (wT>wS) vs. loss of M-sector rents (δSxrSES). Average wage in North unambiguously rises; South may increase or decrease.

### Intuition and comparative cases
- Two key effects of trade-induced institutional improvement in South:
  1. Raises opportunity costs w and r (first effect).
  2. Moves labor into or out of M-sector (second effect) — under trade this often moves M-sector out of South, opposing first effect.
- Extreme examples:
  - If δS = δN + Π with Π small, negligible aggregate institutional improvement but strong loss of M-sector ⇒ South likely loses.
  - If δN = 0 (North perfect institutions), trade can deliver first-best aggregate welfare to South despite loss of M-sector rents.

### Equilibrium outside the FPE set
- For any endowments, rewards in L- and K-sectors equalized under trade:
  - wNT = b pLT = wST
  - rNT = a pKT = rST
- If North can produce integrated quantity of M, South cannot produce M under trade when δN<δS (inequality in text):
  - pMT y = wT + (1+δN) x rT  (North, equality)
  - pMT y < wT + (1+δS) x rT  (South, cannot produce)
- Outside FPE some M production may remain in South if relative endowments sufficiently dissimilar or North small; but high-paying M-sector still shrinks in South and increases in North relative to autarky.

### C. Factor prices and factor movements
- Standard factor-abundance predictions (Dixit and Norman (1980)) differ from this model:
  - Conventional: South (capital-scarce) opens ⇒ return to capital decreases, wages increase; North opposite.
  - This model: Southern capital benefits from opening; labor in South may lose. North: return to capital unchanged; rewards to labor increase.
- Migration/immobility intuition:
  - Model rationalizes immigration pressure (labor wants North for chance at M-sector) and limited capital flows South: institutional differences prevent factor rewards equalization via mobility.

### III. Institutional choice — political economy model
- Setup: policymaker maximizes G(δ,c) = V S(δ) + (1−V) c with V ∈ (0,1); interest group (labor L) maximizes S_L(δ) − c.
- Aggregate welfare S(δ) = r(δ)K + (w(δ) + δ x r(δ) E(δ)) L; S(δ) maximized at δ=0 and dS/dδ < 0.
- Equilibrium institutional quality δ* solves (13) (as in text): argmax δ∈[0,1] of weighted welfare (expression preserved in original notation).
- Comparative statics:
  - Better equilibrium institutions when: 1) low corruption (higher V); 2) higher capital-labor ratios (K/L).
- Under autarky: bad institutions can arise because L can bribe policymaker to raise δ to capture rents; δ* redistributive and lowers aggregate welfare.

### Institutional choice under trade — a “race to the top”
- With trade, optimal δi is best response to partner δ−i; unique equilibrium δN = δS = 0 (perfect institutions).
- Intuition: locating M-sector requires being institutionally superior; each country has incentive to improve institutions slightly above partner to capture worldwide rents.
- Result: trade can force institutional improvement even when a country has high corruption (V=0) — trade-induced competition for M-sector yields institutional upgrading.
- Contrasts with "race to the bottom" narrative; supports empirical evidence (Rodrik, Subramanian and Trebbi (2002)) of trade's positive effect on institutions.

### IV. Empirical evidence — testable predictions and results
- Empirical strategy: J-country Armington extension; M-good varieties imperfect substitutes (α>1). Demand and import share expressions as in equations (14) and (15) from the text.
- Product complexity / institutional dependence proxy: Herfindahl index of intermediate input use (U.S. Input-Output Use Table, 1992); Herfindahl multiplied by −1 so measure increases with institutional intensity.
- Production extension with n intermediates yields participation condition (15): (.)15(,)1(rxwypnMφ++=  (text-preserved form). Higher n ⇒ higher pM for given δ ⇒ lower M production; better institutions (lower δ) raise M production.
- Approximate empirical relationship (16) in logs (with simplifying w≈0): (.)16(.)1ln()1()ln( lk kl Mk Dns++−≈φσ  (text-preserved form).
- Regression specification:
  - rel_share_ic = ϑ + K1 inst_dep_i * inst_c + K2 skint3_i * skill_c + K3 capint3_i * capital_c + Λ_c + Ν_i + Π_ic
  - Left-hand side rel_share_ic: country c's share in U.S. imports in sector i normalized by average country i share.
- Data:
  - 1998 U.S. imports (4-digit SIC) — 177 countries and 389 industries available; final sample: 117 countries and 389 industries where all data available.
  - Country-level institutional quality: Kaufmann, Kraay and Zoido-Lobaton (2002) index (range −2.5 to 2.5).
  - Capital and skill endowments: Hall and Jones (1999) for 123 countries.
  - Industry measures: capital intensity, skill intensity from U.S. Manufacturing database (1992); alternative institutional intensity measures (share20, gini, number of intermediates, investment/output).
- Baseline results (Table 1 summary):
  - (herfindahl index)*inst coefficient ≈ 2.51 (column (1)) and 2.36 (column (2)); both highly significant ((0.68)*** and (0.64)*** standard errors reported).
  - Practical magnitude: moving from 25th to 75th percentile in institutional quality implies predicted relative import share decreases by 0.09 at 25th-percentile institutional intensity good and increases by 0.18 at 75th-percentile institutional intensity good (text statement).
  - skill intensity * skill endow coefficient ≈ 12.34 (0.53) in table format; capital intensity * cap endow ≈ 0.53 (0.30)*.
  - Observations: 31,568; Industries: 389; Countries: 117.
- Robustness:
  - Alternative institutional intensity measures (share of 20 largest intermediates, gini, number of intermediates/1000, investment/output) yield positive and significant interaction coefficients (Table 3).
  - Four-factor model including raw materials (matint4) retains similar institutional interaction effect (Table 4).
  - Controlling for financial dependence and financial development (Column (5), Table 4) — institutional coefficient similar and significant.
  - Instrumental variables with settler mortality interacted with Herfindahl (column (6), Table 4): coefficient of interest similar in magnitude but p-value 23% (not conventionally significant) due to smaller sample (80 countries).
- Subsamples and further robustness (Table 5 and 6):
  - South-only sample: (herfindahl index)*inst ≈ 2.94 (1.12)***; results robust and coefficient often larger in magnitude for South-only subsample.
  - Excluding Africa or SE Asia, or dropping most institutionally intensive industry outliers: coefficients remain similar and highly significant.
  - Institutions vs. per capita income (Table 6):
    - (herfindahl index)*GDPPC coefficient 2.33 (0.61)*** when using log per capita PPP GDP alone.
    - Including both GDPPC and inst reduces significance and coefficients roughly halve — institutions and income highly correlated (corr = 0.82), preventing definitive separation.

### V. Conclusion (summary of comparative implications)
- Two formalizations of institutional differences in trade:
  1. Ricardian view (institutions as productivity differences): trade bails out the South — South gains most, North unchanged.
  2. Grossman-Hart-Moore view (institutions as contract incompleteness and holdup): results reversed — North certain to gain, South may lose; labor rewards can diverge under trade.
- Empirical evidence from U.S. import shares by industry and country supports the model prediction that countries with better institutions capture larger import shares in more institutionally dependent (complex) industries: positive and statistically significant interaction between industry institutional intensity and country institutional quality across many specifications and robustness checks.

*Source: _wp04231 - conclusion, then, is that trade bails out the South: the institutionally weak country no longer (IMF working paper content provided).*

### conclusions are reversed, and quite surprising. The North gains the most from trade, while the

### _wp04231 - conclusions are reversed, and quite surprising. The North gains the most from trade, while the

### Main conclusions on trade and institutions
- The North gains the most from trade, while the South may lose.
- When institutions are a source of trade:
  - Labor in the North and capital in the South are the factors that gain the most.
  - Labor in the South is likely to lose; in fact, wages can diverge as a result of trade.
- Institutional change is slow, so these results are appropriate in the short run.
- When institutions are endogenized (meant to capture long-run effects):
  - In autarky, bad institutions may persist indefinitely.
  - International trade can lead to a "race to the top" in institutional quality as countries compete to capture advantageous sectors; countries improve institutions to attract activity in those sectors.
- The paper contrasts two views of institutions:
  - Grossman-Hart-Moore view: captures contracting imperfections between private parties in production relationships.
  - Ricardian view: may better capture broader institutions such as government expropriation and political instability.
- Industries may differ in the kinds of institutions they require; interactions between institutions and trade are nuanced and context-dependent. What kinds of effects prevail in which circumstances remains an open question.

### Extension of the model: three parties to production (entrepreneur, capital K, labor L)
- Organization of production:
  - Entrepreneurs raise K and establish a company; the company then hires workers.
- Institutional specificity parameters:
  - A fraction φ_K of K becomes specific to the relationship (captures institutional quality in the La Porta et al. sense).
  - A fraction φ_C of company value becomes specific when the company hires a worker (captures labor market conditions and technological features).
- Ex post surplus split assumption: The ex post surplus is split equally between the parties in both relationships.
- Entrepreneur outside option and capital return:
  - Entrepreneur's outside option is fixed at zero; K becomes partly specific to the entrepreneur so K’s participation constraint holds with equality.
  - Given ex ante opportunity cost r, required return R that the company must earn on each unit of K is:
    - R = (1 + φ_K) r.
- Company participation constraint and sector pricing:
  - Because the company becomes partly specific to L, the participation constraint provides a joint restriction on w, r, and p_M analogous to equation (11):
    - p_M y = w + (1 + φ_C) R x = w + (1 + φ_C)(1 + φ_K) r x.
- Worker reward and inefficiency:
  - The reward to labor in the M-sector corresponds to the baseline expression (notationally described in the text) and both key consequences of the baseline model remain:
    - Workers earn rents in the M-sector.
    - The outcome is inefficient.
  - The baseline model without entrepreneurs can be interpreted as a reduced form of this fuller model.
- Insight from the fuller parameterization:
  - Institutional quality in capital markets, φ_K, has a first order effect on worker compensation by changing both the size of the M-sector and the size of workers' rents (equation A1 referenced).
  - The parameterization helps isolate North–South differences. The assumption δ_N < δ_S can be interpreted as a combination of φ_C^N = φ_C^S and φ_K^N < φ_K^S.
  - More nuanced outcomes arise if φ_C differs across North and South; e.g., if φ_C^N > φ_C^S but φ_K^N < φ_K^S, comparative advantage in the M-sector is inconclusive and depends on the interaction between union power (φ_C) and contracting environment (φ_K).

### Key empirical and tabulated statistics (as presented)
- Table A2. Industry-Level Summary Statistics (presented in the source as a single-line summary per variable):
  - capital intensity 0.61 0.11 0.18 0.95
  - skill intensity 0.11 0.06 0.01 0.48
  - herfindahl index of intermediate use 0.13 0.09 0.04 0.78
- Table A3. Country-Level Summary Statistics (presented in the source as a single-line summary per variable):
  - Institutional quality -0.013 0.940 -2.166 1.909
  - log of physical capital per worker 9.241 1.586 5.763 11.589
  - log of human capital per worker 0.584 0.294 0.072 1.215
- Table A1 lists representative industries identified as Least Institutionally Intensive and Most Institutionally Intensive (industry names and codes appear in the source).
- Table A4. Economy List: countries are classified into North and South following Romalis (2004). The North consists of industrial countries identified by Romalis as having in 1995 per capita PPP-adjusted GDP of at least 50% of the U.S. level (the source lists country names under North and South).

*Source: content unit _wp04231 - conclusions are reversed, and quite surprising. The North gains the most from trade, while the (PDF chapter/section).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04231.pdf_