## _wp03233

## Source details

**Canonical URL:** [_wp03233](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03233.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03233.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03233.pdf.json)

---

### I. Introduction — motivation and empirical puzzle
- Observed bilateral trade is substantially lower than predicted by complete-specialization gravity: "total trade (exports plus imports) between countries i and j is proportional to the product of the GNPs of the two countries" (equation (1)).
- Evenett and Keller (2002) estimate the coefficient of trade on the product of GNPs over world GNP to be 0.03–0.26, that is, 87 percent to 98 percent of the predicted trade is missing.
- The “border effect” (McCallum, 1995; Wei, 1996; Anderson and van Wincoop, 2001) shows trade is much lower across national borders than within countries, suggesting incomplete specialization may explain missing trade.
- Two competing explanations:
  - Trade costs/barriers (distance, borders, language, tariffs) — plausible trade costs appear too small to account for missing trade for reasonable elasticities of substitution (Haveman and Hummels, 1999; Hummels, 1999).
  - Incomplete specialization — models with incomplete specialization (e.g., HOV augmented with technology differences and home bias) may better explain bilateral trade volumes (Evenett and Keller, 2002).
- Paper’s strategy: use a Ricardian framework with uncertain productivity and incomplete international financial markets to generate incomplete specialization relying solely on technology differences as the source of trade.

### II. Related literature — channels linking financial integration and trade
- Four channels by which financial integration or financial development may affect goods trade:
  - Lower transaction costs from eliminating exchange controls (Tamirisa, 1999); empirical evidence mixed (Wei, 1999 finds no support for hedging instruments raising trade).
  - Financial development enables insurance against external (world) shocks, lowering costs of openness and enabling more liberal trade policies (Rodrik, 1998; Svaleryd and Vlachos, 2002b).
  - Financial development shifts specialization toward industries intensive in financial services (Svaleryd and Vlachos, 2002a); requires domestic financial development rather than globalization.
  - Improved international risk sharing encourages specialization by comparative advantage (Kalemli-Ozcan et al., 2002).
- Theoretical points:
  - Better risk sharing raises risk-taking and resources to riskier but more productive sectors (Obstfeld, 1994).
  - Incomplete financial markets produce inefficient diversification and under-specialization (Acemoglu and Zilibotti, 1997; Feeney, 1999).
- Modeling financial integration ranges from financial autarky (Ruffin, 1974a,b) to complete markets (Helpman and Razin, 1978a,b); intermediate approaches have limitations (Baxter and Crucini, 1995).
- Empirical relevance: international financial markets are far from complete; cross-border trade in individual shares is limited (French and Poterba, 1991; Coval and Moskowitz, 1999; Huberman, 2001).

### III. Main argument and intuition
- Core tradeoff: international versus domestic diversification — incomplete international markets induce domestic diversification, leading to incomplete specialization and lower trade.
- Illustrative example: two countries (Canada, U.S.), two goods (wheat, ice cubes); with high risk aversion and financial autarky both produce both goods → little cross-border trade; opening financial markets permits international diversification and specialization, increasing trade.
- Contribution: a general equilibrium Ricardian model with uncertainty and incomplete markets where asset prices are endogenous and industry-specific, allowing analysis of how financial market completeness affects specialization and trade volumes.

### IV. The model — setup and key relations
- Environment and assumptions:
  - Static model (no dynamics, no investment, no growth, no current account).
  - No trade frictions (law of one price holds).
  - J countries, S goods, representative consumer per country, homothetic preferences.
- Preferences and consumption:
  - Utility: u_j(C_j,1, ..., C_j,S) = v_j(g(C_j,1, ..., C_j,S)), g(·) homogeneous of degree one (equation (2)).
  - Indirect utility: V_j(p, I_j) = v_j(I_j / g*(p)); with normalization g* = 1, V_j(p, I_j) = v_j(I_j) (equations (3)–(4)).
  - Consumption shares identical across countries: C_{j,s} = β_s I_j / p_s (summary).
- Production, trade shares and trade volume:
  - Define α_{j,s} = p_{j,s} q_{j,s} / I_j.
  - Net trade of good s: T_{j,s} = I_j (α_{j,s} - β_s) (equations (5)–(6)).
  - Aggregate trade for country j: T_j = Σ_s I_j |α_{j,s} - α_s^w| (equation (7)).
  - Bilateral trade (two-country case): T_{ij} = (I_i I_j / I_w) Σ_s |α_{i,s} - α_{j,s}| (equation (8)); in complete specialization the specialization index sums to two and standard gravity recovered.
- Production technology and shocks:
  - Ricardian, single-factor capital with CRS:
    - q_{j,s} = k_{j,s} A_{j,s} θ_{s}^{%} (equation (9)).
  - A_{j,s} is country-specific productivity; θ_{s}^{%} are industry-specific productivity shocks identical across countries (mean 1).
  - Revenue: R_{j,s}^{~} = p_{s}^{%} k_{j,s} A_{j,s} θ_{s}^{%} (equation (11)).
  - Normalized revenue shock φ_s (mean 1); Q_{j,s} = E(revenue) (equation (12)).
- Risk preferences and portfolio choice:
  - Indirect utility CARA: v_j(I_j) = - exp( - γ_j I_j ) (equation (13)).
  - Equivalent mean-variance objective: max E(I_j) - (γ_j/2) Var(I_j) (equation (14)).
  - CARA coefficient varies: γ_j = γ / Y_j (approximation to CRRA around Y_j, with Y_j = E(I_j)).
- Asset markets and decomposition:
  - Forward contracts, N risky assets (zero mean, identity covariance), and riskless asset at world rate R_0.
  - Joint normality and key covariance B = Cov(φ, ω).
  - Decomposition: φ = 1 + B ω + u (equation (15)), Var(φ) = B B' + Σ where Σ is uninsurable variance.
- Equilibrium definition:
  - Equilibrium comprises consumption and production {c_{j,s}, q_{j,s}}, goods prices {p_s}, financial holdings {h_{j,n}}, and asset prices {π_n} satisfying optimality, market clearing, and zero net world demand for existing assets.
- Theorem 1 (industrial composition):
  - World production composition: q^w = γ^{-1} Ω^{-1} m (equation (16) form).
  - Country specialization: q_j - q^w = Y_j^{1-γ} Σ^{-1} ( m_j - m ) (equation (17) structure).
    - m_{j,s} = 1 - R_0 / [ E(p_s θ_s) a_{j,s} ].
  - Intuition: allocation parallels mean-variance portfolio choice; uninsured risk Σ dampens specialization; without uninsured risk small productivity differences yield complete specialization.
- Corollary: world production structure and distribution of goods prices are not affected by financial markets (product prices depend only on world composition).

### V. Financial globalization, insured vs uninsured risk, and trade implications
- Financial globalization modeled as increase in number of tradable assets N:
  - Systematic (insured) risk increases with N: systematic risk = sum_{n=1}^{N} b_n' b_n.
  - Overall risk Ω does not depend on N (Proposition 1); uninsured risk Σ = Ω - sum_{n=1}^{N} b_n' b_n decreases as N increases.
- Implications:
  - As Σ declines, countries' industrial structures diverge according to productivity differences → more specialization.
  - Financial integration expands export industries (more productive than world average) and contracts import-competing hedging industries → increases trade flows.
  - In the limit Σ → 0 (complete markets) specialization can reach corner (full specialization).
  - Aggregate world composition still determined by total variance Ω (unchanged by market completeness).

### VI. Sectoral-specialization results (diagonal shocks) and comparative statics
- With independent sectoral shocks (Σ and Ω diagonal), denote:
  - σ^2_s = total variance of sector s; τ^2_s = uninsured variance of sector s.
- Specialization equations (as presented):
  - (20) α_{j,s} - α_{w,s} ∝ (m_{j,s} - m_s) / τ^2_s (implied by displayed structure).
  - (21) α_{w,s} ∝ m_s / σ^2_s (implied).
  - Trade-to-consumption ratio:
    - (22) T_{j,s} / C_{j,s} ∝ (m_{j,s} - m_s) (σ^2_s / τ^2_s).
- Interpretations:
  - Country j is net exporter of good s if m_{j,s} > m_s.
  - Volume of trade increases as τ^2_s declines (financial globalization insures more sectoral shocks).
  - World trade in good s is high relative to world output if productivity is dispersed and uninsured risks τ^2_s are small (equation (23) variant).
- Reduced-form relation:
  - (25) (OPEN_{j,s}, RISK_{j,s}) PROD_{j,s} , js js js js T f C = ⋅ with f>0, f_1>0, f_2<0, f_{12}>0.
    - Interpretation: f_1>0 (more integrated countries respond more to productivity differences); f_2<0 (risk reduces specialization); f_{12}>0 (openness mitigates negative effect of risk).
- Aggregate trade:
  - (26) multilateral trade relative to GNP ∝ Σ_s Σ_j (m_{j,s} - m_s)^2 / τ^2_s (as in text structure).
  - Prediction: multilateral trade volume is high relative to GNP when countries differ from world in productivity and uninsured risks τ^2_s are small.

### VII. Empirical design, measures, and data
- Panel: 175 countries, years 1980 to 1997.
- Data sources:
  - World Trade Database (Statistics Canada) for trade flows.
  - UNIDO Industrial Structure Database for productivity.
  - Index of financial openness compiled at IMF by Abiad and Mody (2003).
  - NBER Productivity Database (Bartelsman, Becker and Gray, 2000) for TFP fluctuations.
- Sectoring: 3-digit ISIC (rev. 2) aggregated to 22 industries (Appendix Table A1).
- Key measures:
  - Labor productivity = value added per employment (UNIDO).
  - PROD = percentage difference of labor productivity relative to world average (definition preserved).
  - COMPADV = 1/S Σ |PROD_{is}| (cross-country comparative advantage).
  - Dependent variables: trade dependence ratio ≈ (X_{is} − M_{is}) / (α_{w,s} Y_i); multilateral trade = (exports + imports with rest of world) / country GNP.
  - Financial openness (Abiad and Mody, 2003): 0 = fully repressed, 1 = partially repressed, 2 = largely liberalized, 3 = fully liberalized; index available for 36 countries (Appendix Table A2).
  - Alternative measure: consumption risk sharing β (β = 0 no risk sharing, β = 1 complete consumption insurance).
  - Sectoral risk proxy: standard deviation of annual TFP growth of the sector in the United States (Appendix Table A3).
- Econometrics:
  - Country × year fixed effects used to filter persistent trade imbalances.
  - Export- and import-specific regressions estimated separately.
  - Scaling of dependent variable pinned by theory (equation (22)).

### VIII. Empirical findings (preserving reported numeric estimates)
- General: Tables 1–6 support theoretical predictions that financial openness increases trade and the responsiveness of trade to comparative advantage; sectoral risk reduces trade responsiveness.
- Selected quantitative results (as reported in the source text):
  - Table 1:
    - log(GNP) coefficients: 0.889 (0.010)** and 0.914 (0.009)**.
    - COMPADV coefficients: -0.191 (0.038)** and -0.762 (0.075)**.
    - Weighted COMPADV: -0.206 (0.051)** and -0.658 (0.097)**.
  - Table 2:
    - log(GNP) coefficients: 0.781 (0.022)** and 0.757 (0.020)**.
    - OPEN category coefficients examples: OPEN = 1: -0.011 (0.146), 0.462 (0.227)*, 0.051 (0.129), 0.487 (0.182)**.
    - COMPADV (OPEN = 3): 4.918 (0.435)** and 3.131 (0.731)**; weighted COMPADV (OPEN = 3): 3.545 (0.770)** and 0.316 (1.122).
  - Table 3:
    - PROD coefficients across specifications range 0.055 to 0.103 with standard errors (0.005)** to (0.007)**.
    - Observations: 25,649; R-squared: 0.01, 0.12, 0.16 across columns.
    - A 10 percent labor productivity advantage leads to an extra 0.6 to 1 percent of domestic consumption exported (reported magnitudes).
  - Table 4a (PROD by OPEN):
    - PROD (OPEN = 0): 0.085 (0.018)**, 0.104 (0.014)**, 0.084 (0.026)**, 0.107 (0.020)**, 1.987 (0.043)**.
    - PROD (OPEN = 3): 0.120 (0.025)**, 0.125 (0.021)**, 0.170 (0.035)**, 0.172 (0.029)**, 0.686 (0.057)**.
    - Example observations: 10,189; R-squared range 0.01–0.27.
  - Table 4b (weighted PROD by OPEN):
    - Weighted PROD (OPEN = 0): 0.045 (0.013)**, 0.064 (0.021)**, 1.297 (0.034)**.
    - PROD (OPEN = 3): 0.068 (0.019)**, 0.063 (0.031)*, 0.739 (0.051)**.
    - Observations: 10,189 and 11,512; R-squared: 0.06, 0.18, 0.23.
  - Table 4c (PROD × Beta interactions):
    - PROD coefficients e.g. 1.467 (0.026)**, 1.403 (0.064)**, 0.113 (0.011)**.
    - Beta*PROD: -0.340 (0.091)** and 1.549 (0.495)** in different specs.
    - Observations vary (e.g., 18,672; 7,694).
  - Table 5 (RISK × OPEN):
    - RISK (OPEN=3): 9.886 (1.527)** and 9.220 (0.765)** in reported columns.
    - PROD (OPEN=0): 0.091 (0.027)**.
    - Observations: 10,251 and 10,189; Within R-squared: 0.00–0.02.
  - Table 6 (PROD and RISK interactions by OPEN):
    - PROD overall: 0.089 (0.007)**.
    - PROD by OPEN: OPEN=0: 0.110 (0.011)**; OPEN=1: 0.141 (0.019)**; OPEN=2: 0.260 (0.025)**; OPEN=3: 0.221 (0.044)**.
    - PROD * RISK interaction coefficients range from -1.924 (0.450)** to -8.431 (2.322)** across subsamples.
    - Observations vary: 25,649; 2,881; 2,197; 2,341; 2,770.
- High-level empirical conclusions:
  - Financial openness tends to increase trade responsiveness to comparative advantage and total trade volumes.
  - Sectoral risk reduces trade driven by productivity differentials; openness mitigates some adverse effects but results vary across specifications and sectors.
  - Notable heterogeneity: in some specifications closed economies appear more specialized when specialization measured via value-added shares; highly open countries may display intra-industry trade and product differentiation effects not captured by the Ricardian model.

### IX. Conclusion and research directions
- Theoretical framework: a micro-founded general equilibrium Ricardian model with uncertainty and incomplete markets yields explicit predictions on specialization and trade and links financial globalization to specialization via risk sharing.
- Empirical validation: evidence supports that financially open countries trade more, respond more to productivity differences, and are less sensitive to industry risks.
- Directions for future work recommended in the source:
  - Identify specific industry factors and countries’ exposures to them.
  - Conduct event-study analyses around dates of capital account opening.
  - Include additional controls that may determine trade flows.

### X. Appendix — data notes and selected tables (as presented)
- Table A1: sector definitions (examples): Food products ISIC 311; Beverages ISIC 313; Tobacco ISIC 314; Textiles + apparel ISIC 321, 322; Leather + footwear ISIC 323, 324; Paper and products ISIC 341; Transport equipment ISIC 384.
- Table A2: financial openness index examples (selected country time series, 1982–1994; 0 = closed, 3 = fully liberalized):
  - ARG: 1982–1990 = 0; 1991–1994 = 3.
  - AUS: 1982–1983 = 0; 1984–1994 = 3.
  - CAN, DEU, USA: 1982–1994 = 3.
  - IND: 1982–1990 = 0; 1991–1994 = 1.
  - PER: 1982–1990 = 0; 1991–1994 = 3.
- Table A3: sectoral risk (σ(dTFP) by 2-digit SIC, U.S.):
  - SIC 20 Food: 0.0132
  - SIC 21 Tobacco: 0.0563
  - SIC 22 Textile Mill Products: 0.0192
  - SIC 23 Apparel and Other Textile Products: 0.0154
  - SIC 24 Lumber and Wood Products: 0.0250
  - SIC 25 Furniture and Fixtures: 0.0241
  - SIC 26 Paper and Allied Products: 0.0238
  - SIC 27 Printing and Publishing: 0.0211
  - SIC 28 Chemicals and Allied Products: 0.0271
  - SIC 29 Petroleum and Coal Products: 0.0360
  - SIC 30 Rubber and Miscellaneous Plastics Products: 0.0238
  - SIC 31 Leather and Leather Products: 0.0277
  - SIC 32 Stone, Clay and Glass Products: 0.0225
  - SIC 33 Primary Metal Industries: 0.0307
  - SIC 34 Fabricated Metal Products: 0.0234
  - SIC 35 Industrial Machinery and Equipment: 0.0332
  - SIC 36 Electronic and Other Electric Equipment: 0.0512
  - SIC 37 Transportation Equipment: 0.0192
  - SIC 38 Instruments and Related Products: 0.0207
  - SIC 39 Miscellaneous Manufacturing Industries: 0.0283

*Source: Excerpt from _wp03233 (pages 3–10) — Introduction, Related literature, Model, Empirical design, Results, Conclusion, and Appendix.*

### References  ............................................................................................................

### _wp03233 - References

### I. Introduction — motivation and empirical puzzle
- Observed bilateral trade is substantially lower than predicted by complete-specialization models that imply the gravity equation: "total trade (exports plus imports) between countries i and j is proportional to the product of the GNPs of the two countries" (equation (1)).
- Evenett and Keller (2002) estimate the coefficient of trade on the product of GNPs over world GNP to be 0.03–0.26, that is, 87 percent to 98 percent of the predicted trade is missing.
- The “border effect” (McCallum, 1995; Wei, 1996; Anderson and van Wincoop, 2001) shows trade is much lower across national borders than within countries, suggesting incomplete specialization may explain missing trade.
- Two competing explanations discussed:
  - Trade costs/barriers: distance, borders, language, tariffs; but plausible trade costs appear too small to account for missing trade for reasonable elasticities of substitution (Haveman and Hummels, 1999; Hummels, 1999).
  - Incomplete specialization: models with incomplete specialization (e.g., HOV augmented with technology differences and home bias) may better explain bilateral trade volumes (Evenett and Keller, 2002).
- Paper’s strategy: use a Ricardian framework with uncertain productivity and incomplete international financial markets to generate incomplete specialization while relying solely on technology differences as the source of trade.

### II. Related literature — channels linking financial integration and trade
- Four empirical/theoretical channels by which financial integration or financial development may affect goods trade:
  - Lower transaction costs from eliminating exchange controls (Tamirisa, 1999); empirical evidence mixed (Wei, 1999 finds no support for hedging instruments raising trade).
  - Financial development enables insurance against external (world) shocks, lowering costs of openness and enabling more liberal trade policies (Rodrik, 1998; Svaleryd and Vlachos, 2002b).
  - Financial development shifts specialization toward industries intensive in financial services (Svaleryd and Vlachos, 2002a); this requires domestic financial development rather than globalization.
  - Improved international risk sharing encourages specialization by comparative advantage (Kalemli-Ozcan et al., 2002 provide empirical confirmation at regional level using consumption-GDP “beta” and Krugman’s specialization index).
- Theoretical contributions on real effects of financial access:
  - Better risk sharing increases risk-taking and resources to riskier but more productive sectors (Obstfeld, 1994).
  - Inefficient diversification under incomplete financial markets leads to under-specialization and scale inefficiencies (Acemoglu and Zilibotti, 1997; Feeney, 1999).
- Modeling financial integration:
  - Extremes: financial autarky (uncertainty destroys neoclassical trade results; Ruffin, 1974a,b) vs. complete markets (most results resurrect; Helpman and Razin, 1978a,b).
  - Intermediate models (trade in riskless bonds) are empirically limited (Baxter and Crucini, 1995).
  - Proportional transaction cost models of asset trade (Martin and Rey, 2000, 2002; Heathcote and Perri, 2002) imply globalization lowers those costs but are at odds with evidence that transaction costs are small (Tesar and Werner, 1995).
- Paper’s approach: model international financial segmentation with incomplete markets where not all risks can be traded; financial globalization reduces the uninsurable portion of risk.
- Empirical relevance: international financial markets are far from complete; trade in individual shares across borders is limited (French and Poterba, 1991; Coval and Moskowitz, 1999; Huberman, 2001).

### III. Main argument and intuition
- Core tradeoff: international versus domestic diversification — incomplete international markets imply investors maintain domestic diversification, leading to incomplete specialization and lower trade.
- Simple illustrative example:
  - Two countries (Canada, U.S.), two goods (wheat, ice cubes); productivity is random.
  - With high risk aversion and financial autarky, both countries diversify domestically, producing both goods → little or no cross-border trade.
  - Opening international financial markets enables international diversification (via tradable assets), allowing specialization according to comparative advantage and increasing trade flows.
- Contribution: provides a general equilibrium Ricardian model with uncertainty and incomplete markets where asset prices (cost of insurance) are endogenously determined and depend on industry characteristics, enabling analysis of how financial market completeness affects specialization and trade volumes.

### IV. The model — setup and key relations
- Assumptions:
  - Static model focusing on uncertainty; no dynamics (no investment, growth, current account).
  - No trade frictions (law of one price holds for each commodity).
  - J countries, S goods, representative consumer per country with homothetic preferences.
- Preferences and indirect utility:
  - Utility: u_j(C_j,1, ..., C_j,S) = v_j(g(C_j,1, ..., C_j,S)), where g(·) is homogeneous of degree one (equation (2)).
  - Indirect utility with price index g* and income I_j: V_j(p_1, ..., p_S, I_j) = v_j(I_j / g*(p_1, ..., p_S)) (equation (3)); with g* normalized to one, simplifies to V_j(p, I_j) = v_j(I_j) (equation (4)).
- Consumption and production shares:
  - Consumption shares identical across countries due to homotheticity: C_{j,s} = β_s I_j / p_s (summary form).
  - Define α_{j,s} as share of production in sector s in income of country j: α_{j,s} = p_{j,s} q_{j,s} / I_j (text definition).
  - Net trade of good s: T_{j,s} = p_{j,s} q_{j,s} - C_{j,s} = I_j (α_{j,s} - β_s) (equation (5) and (6)).
  - World consumption shares equal world production shares: β_s = α_s^w = Σ_j (I_j / I_w) α_{j,s}.
- Volume of trade:
  - Aggregate trade for country j: T_j = Σ_s |T_{j,s}| = Σ_s I_j |α_{j,s} - α_s^w| (equation (7) indicates trade scales with income and specialization divergence).
  - Bilateral trade between i and j (two-country case): T_{ij} = (I_i I_j / I_w) Σ_s |α_{i,s} - α_{j,s}|, resembling the gravity equation augmented by an index of specialization differences (equation (8)). In the case of complete specialization the specialization index sums to two and the standard gravity equation (1) is recovered.
- Pattern of specialization:
  - Determined by portfolio choice: investors trade off higher expected returns (comparative advantage) against sector riskiness; financial market completeness affects the feasibility and cost of risk sharing and thus the equilibrium α_{j,s}.

### V. Implications for trade and financial globalization
- Incomplete international financial markets can quantitatively account for lower-than-predicted trade by inducing incomplete specialization.
- Financial globalization (reducing uninsurable risk) should:
  - Reduce domestic diversification incentives,
  - Increase specialization according to comparative advantage,
  - Raise the volume of international trade.
- Endogenous asset prices in the model allow assessing the cost of insurance and how that cost varies with industry composition and financial market completeness.

*Source: Excerpt from _wp03233 (pages 3–10) — Introduction, Related literature, and The model.*

### 1.  Investors allocate capital to each sector and enter into financial contracts with foreign

### 1.  Investors allocate capital to each sector and enter into financial contracts with foreign investors.

### Production: technology, shocks, and revenues
- Production is Ricardian, single factor of production (capital) with constant-returns-to-scale technology.
- Output in sector s, country j:
  - q_{j,s} = k_{j,s} A_{j,s} θ_{s}^{%}  (equation (9) notation: ,,,
θ
jss    js   js
qAk=%
%)
- Features:
  - A_{j,s} is country-specific (drives specialization).
  - Productivity shocks θ_{s}^{%} are industry-specific but identical across countries (mean 1).
- Equilibrium goods prices are functions of world production:
  - p_{s}^{%} = f_s( q_{1,s}^{%},..., q_{J,s}^{%} ) and f_s is homogeneous of degree zero (equation (10) structure).
- Revenue in sector s, country j (before rewriting):
  - R_{j,s}^{~} = p_{s}^{%} q_{j,s}^{~} = p_{s}^{%} k_{j,s} A_{j,s} θ_{s}^{%}  (equation (11))
- Combined primitive revenue shock notation:
  - φ_{s}^{%} denotes normalized revenue shock of sector s (mean 1).
  - Q_{j,s} denotes expected revenue of country j from sector s.
  - Rewritten relation (equation (12) style): Q_{j,s} = E( p_{s} θ_{s} ) a_{j,s} k_{j,s} with φ notation introduced.

### Portfolio choice: preferences and problem setup
- Indirect utility assumed CARA (constant absolute risk aversion):
  - v_j(I_j) = - exp( - γ_j I_j )  (equation (13))
- Maximizing exponential utility equivalent to mean-variance criterion:
  - max E(I_j) - (γ_j/2) Var(I_j)  (equation (14) structure)
- To capture scale invariance, CARA coefficient varies across countries:
  - γ_j = γ / Y_j  (approximation to CRRA around Y_j; Y_j = E(I_j))
- Decision sequence:
  - (1) choose portfolio to maximize expected indirect utility (max E v_j[I_j]),
  - (2) after uncertainty resolves, allocate real revenue to consumption via homothetic sub-utility.

### Asset markets: assets, payoffs, and decomposition
- Available financial contracts:
  - Forward contracts bought in period 0 with zero money down; payoff and predetermined price in period 1.
  - N risky assets in zero net supply, random cash flows ω^{~} = (ω_1^{~},...,ω_N^{~}), zero mean, identity covariance matrix (uncorrelated with variance 1).
  - Riskless asset available in infinitely elastic supply at world interest rate R_0 (borrow/lend freely).
- Joint distribution of financial payoffs and revenue shocks is normal with mean/variance normalization:
  - Revenue shocks have unit mean; financial payoffs have zero mean and identity covariance.
  - Key object: covariance of revenue shocks with financial payoffs, B = Cov(φ, ω).
- Decomposition of revenue shocks (equation (15) form):
  - φ = 1 + B ω + u
    - sure component: 1
    - hedgeable component: B ω
    - uninsured component: u
  - Variance decomposition:
    - Var(φ) = B B' + Σ  (Σ denotes variance matrix of uninsurable risk; notation in text: Var'   E'φ≡= +ΩBBuu %%% with Σ as uninsurable variance)
- Vector notation for country j:
  - q_j : S×1 vector of expected industry revenues
  - h_j : N×1 vector of financial asset holdings
  - k_j : S×1 vector of capital allocations
  - A_j : S×S diagonal matrix with a_{j,s} on diagonal
  - P : S×S diagonal matrix with E(p_s θ_s) on diagonal

### Equilibrium definition and characterization
- Equilibrium consists of:
  - consumption ({c_{j,s}}) and production ({q_{j,s}}) for each country-sector (J×S pairs),
  - goods prices ({p_1,...,p_S}),
  - financial asset demands ({h_{j,n}}) and asset prices ({π_1,...,π_N}),
  - satisfying optimal consumption shares, product market clearing, utility-maximizing capital allocation and financial trades, exhausted capital stock, and zero world net demand of existing financial assets (conditions (i)–(v)).
- Theorem 1 (industrial composition):
  - World production composition:
    - q^w = γ^{-1} Ω^{-1} m   (equation (16) style: 1 w w Y − = γ q Ω m)
      - notation aligns with text where world vector q and m appear with γ and Ω factors.
  - Country specialization:
    - q_j - q^w = Y_j^{1-γ} Σ^{-1} ( m_j - m )   (equation (17) structure: ()1 1 . j w j jw YY − −=− γ q q Σ m m)
    - m_j is S×1 with elements m_{j,s} = 1 - R_0 / [ E(p_s θ_s) a_{j,s} ] (text: m_j = 1 - R_0 P^{-1} A_j^{-1} 1)
    - m is income-weighted average of m_j across countries (weights Y_j)
- Intuition:
  - Allocation rule parallels mean-variance optimal portfolio: fraction of wealth in each asset ∝ inverse covariance × expected excess returns.
  - A country has higher share in sector s than world average if it has absolute productivity advantage (a_{j,s} > average).
  - Uninsurable risk (Σ) dampens specialization: higher Σ or higher risk aversion → smaller deviations from world average industry shares.
  - Without uninsured risk, infinitesimal productivity differences would yield complete specialization (corner solution).
- Corollary (Proposition 1):
  - World production structure and distribution of goods prices are not affected by financial markets.
  - Because goods prices depend only on composition of world output and utility is homothetic, product prices and their joint distribution are determined by world composition (equation (18) style).

### Financial globalization, insured vs uninsured risk, and trade implications
- Financial globalization modeled as increase in number of internationally tradable assets N:
  - Systematic (insured) risk increases with N:
    - systematic risk = sum_{n=1}^{N} b_n' b_n  (text notation: 1 ', N nn n = = ∑ ' BB b b)
  - Overall risk Ω does not depend on N (Proposition 1).
  - Therefore uninsured risk Σ decreases as N increases:
    - Σ = Ω - sum_{n=1}^{N} b_n' b_n  (text: 1 . N nn n = = − ∑ ' Σ Ω bb)
- Implications:
  - As uninsured variance declines, countries' industrial structures diverge according to productivity differences (move toward specialization).
  - Financial integration expands industries that are more productive than world average (export industries) and contracts less-productive (import-competing) industries that previously served as hedges.
  - Net effect: increase in trade flows as export industries expand and import-competing industries shrink.
- Limits:
  - In complete markets (Σ → 0), specialization intensifies up to the corner solution of full specialization.
  - Aggregate portfolio and world industry composition still determined by total variance Ω (unchanged by financial market structure).

### Key equations and expressions (as presented)
- Output and shocks:
  - q_{j,s} = k_{j,s} A_{j,s} θ_{s}^{%}  (equation (9) format)
- Revenue:
  - R_{j,s}^{~} = p_{s}^{%} q_{j,s}^{~} = p_{s}^{%} k_{j,s} A_{j,s} θ_{s}^{%}  (equation (11))
- Normalized revenue / expected revenue notation:
  - φ_s (mean 1), Q_{j,s} = E(revenue)  (equation (12) introduction)
- Utility:
  - v_j(I_j) = - exp( - γ_j I_j )  (equation (13))
  - Equivalent objective: max E(I_j) - (γ_j/2) Var(I_j)  (equation (14))
  - γ_j = γ / Y_j  (approximation to CRRA)
- Asset payoff decomposition:
  - φ = 1 + B ω + u  (equation (15))
  - Var(φ) = B B' + Σ  (variance decomposition)
- Specialization and world composition:
  - q^w = γ^{-1} Ω^{-1} m  (equation (16) form)
  - q_j - q^w = Y_j^{1-γ} Σ^{-1} ( m_j - m )  (equation (17) structure)
  - m_j elements: m_{j,s} = 1 - R_0 / [ E(p_s θ_s) a_{j,s} ]
- Financial globalization effect on uninsured risk:
  - uninsured risk Σ decreases with N as Σ = Ω - sum_{n=1}^{N} b_n' b_n

*Italic: Content based on _wp03233 - 1.  Investors allocate capital to each sector and enter into financial contracts with foreign investors.*

### 0. Then the economy will exhibit

### 0. Then the economy will exhibit

### Theoretical results and characterization of specialization
- Complete specialization: industries will only exist in the region in which they have the highest expected return (lowest cost of production). (Footnote: This special case is not included in equation (17), which is a first-order condition for equilibrium. Complete specialization is attained as a corner solution, where the first-order condition is not binding.)
- Relationship between expected revenues and specialization: q_j is the vector of expected revenues, E(p_s q_{j,s}), and in equilibrium Y_j equals expected GNP, E(I_j). Equation (17) gives an explicit formula for the expected vector of specialization, and hence the pattern of trade.
- Key equations (as presented):
  - Equation (17’) (variant of (17) when countries have differential access to world financial markets):
    - 1 1 () j w jj jw YY γ − − = − q q Σ m m                                          (17’)
  - Sufficiency condition for (17’):
    - () () i iJn w iw iJn YY ∈ ∈ = ∑ ∑ q q                                                     (19)                                                     
    - Interpretation: for all n ∈ N(j), the average production structure of countries trading asset n must equal the world average so asset prices are the same as under uniform financial integration.
- Proposition 2: If country j can trade set of assets N(j) and asset n is traded by set J(n), then (17’) characterizes specialisation of country j provided condition (19) holds.
- Effect of incomplete financial integration: the only difference between (17) and (17’) is that the variance matrix of uninsured shocks may be country specific:
  - () . jnn nN j ∈ =− ∑ Σ Ω b b
  - Countries trading fewer financial assets internationally will have a larger variance of uninsured shocks.

### Intuition on scope and limits
- Condition for Proposition 2 is restrictive only if opening financially closed countries affects world asset prices; approximation holds if those countries are small in GNP or have productivity structures similar to the world; it fails for large, highly specialized countries.

### Sectoral-specialization results with diagonal shocks
- Assumption: sectoral shocks are independent → total variance matrix Σ and uninsured variance matrix Ω are diagonal.
  - Notation: σ^2_s denotes total variance, τ^2_s the uninsured variance of shocks in industry s.
- Specialization equations:
  - (20) 2 1 . jss jss s mm αα γ τ − − = 
  - (21) 2 1 . s s s m α γ σ =
  - Ratio (trade to consumption, trade dependence) from dividing (20) by (21):
    - (22) 2 2 () . jsjss jjss s jssjss TImm CIm α − α − σ == α τ
- Interpretation of (22):
  - The ratio of trade to consumption is proportional to country j’s percentage expected-return differential relative to the world; the coefficient is the ratio of total variance and uninsured variance in the sector.
  - Country j is net exporter of good s (T_{j,s} > 0) if its expected return is higher than world average (m_{j,s} > m_s).
  - Volume of trade depends on fraction of productivity shocks insured internationally: as financial globalization progresses and τ^2_s declines, |T_{j,s}| rises.
- World production share special case:
  - (23) 2 1 . J jssj J jss j wss jw j wss wss Y mm T YY CY m = = α − α − σ == α τ ∑ ∑
  - World trade in good s is high relative to world output if productivity is dispersed and uninsured risks in the industry are small.

### Impact of increasing number of internationally traded assets
- Uninsured portion of risk declines as number of traded assets rises:
  - (24) 222 () . jsssn nN j b τ σ ∈ =− ∑
- Reduced-form expression linking trade dependence to openness, risk, and productivity:
  - (25) (OPEN_{j,s}, RISK_{j,s}) PROD_{j,s} , js js js js T f C = ⋅ with f>0, f_1>0, f_2<0, f_{12}>0.
  - Interpretation of partial derivatives:
    - f_1>0: Financially more integrated country responds more to a given percentage productivity difference.
    - f_2<0: Riskiness of sector reduces specialization (decreases exports in exporting industries, increases import-competing industries).
    - f_{12}>0: Openness mitigates negative effect of risk on specialization (open countries can diversify via international markets).
  - Caveat: no direct mapping between these measures and structural parameters; measures imperfectly capture total and uninsured risks.

### Volume of trade (aggregate)
- Aggregate multilateral trade relative to GNP:
  - (26) 2 11 1 SS s js jjsj ss s mm T T Y γ τ == − == ∑ ∑
- Prediction: multilateral trade volume is high relative to GNP if country differs from the world in productivity and if uninsured risks are small; trade volume should be higher in more integrated countries, especially those more different from the world in productivity.

### Empirical implications summarized
- Cross-sectoral: openness increases responsiveness to comparative advantage; risk reduces trade but less so for open countries; risk and productivity interact so risky sectors show diminished response to productivity differences.
- Cross-country: financial integration should increase total trade volumes and amplify trade responses for countries with larger productivity differences from the world.

### Empirical design and data (as used in the analysis)
- Panel dataset: 175 countries, years 1980 to 1997.
- Data sources: World Trade Database (Statistics Canada) for trade flows; UNIDO Industrial Structure Database for productivity; index of financial openness compiled at IMF by Abiad and Mody (2003); NBER Productivity Database (Bartelsman, Becker and Gray, 2000) for TFP fluctuations.
- Sectoring: 3-digit ISIC (revision 2) industries aggregated to 22 industries (listed in Appendix Table A1).
- Productivity measure:
  - Labor productivity = value added per employment (UNIDO).
  - Percentage difference of labor productivity relative to world average, then subtract average productivity advantage of manufacturing:
    - PROD ≡ // PRO D log log .// is is ii is ws ws ww YL YL YL YL ≡ −
  - Cross-country comparative advantage: COMPADV = 1/S ∑ |PROD_{is}|.
- Dependent variables:
  - Cross-industry estimations: trade dependence ratio ≈ (X_{is} − M_{is}) / (α_{ws} Y_i) where α_{ws} is world share of value added in sector s and Y_i is country GNP.
  - Cross-country estimations: multilateral trade volume = (exports + imports with rest of world) / country GNP.
- Financial openness index (Abiad and Mody, 2003):
  - Values: 0 = fully repressed, 1 = partially repressed, 2 = largely liberalized, 3 = fully liberalized.
  - This index is available for 36 countries (Appendix Table A2).
- Alternative measure: degree of consumption risk sharing β defined as fraction of variance in idiosyncratic GDP growth not passed to idiosyncratic consumption; β = 0 corresponds to no risk sharing, β = 1 to complete consumption insurance.
- Sectoral risk proxy: standard deviation of annual TFP growth of the sector in the United States (Appendix Table A3).
- Econometric considerations:
  - Country × year fixed effects used to filter out persistent trade imbalances.
  - Export- and import-specific regressions also estimated to avoid net-export confounding.
  - Scaling of dependent variable is theoretically pinned down by equation (22).

### Empirical findings (high-level, with key numeric points preserved)
- Tables 1–6 generally support theoretical predictions.
- Table 1: simple OLS of multilateral trade on COMPADV shows negative correlation; effect stronger when controlling for country size (small countries more specialized).
- Table 2: negative relationship disappears for financially open countries (OPEN = 2, 3); trade of financially more integrated countries depends more on comparative advantage.
- Table 3: productivity → net exports:
  - A 10 percent labor productivity advantage leads to an extra 0.6 to 1 percent of domestic consumption exported.
  - Fit is poor: productivity alone explains little variation in trade dependence ratios.
- Table 4a/4b (productivity × openness interaction):
  - Productivity matters more for trade in financially more open countries.
  - Robust difference between OPEN = 0 and OPEN = 2.
  - Two notable exceptions:
    - When specialization measured as production per consumption (domestic value added share per world share), closed economies may specialize more according to comparative advantage (possible measurement error in value added).
    - Sometimes a drop in PROD coefficient when moving from OPEN = 2 to full openness (OPEN = 3), possibly due to product differentiation, imperfect competition, and intra-industry trade in highly developed countries.
- Table 4c (consumption risk sharing interactions):
  - Generally, countries with more consumption risk sharing do not respond more to productivity differentials; interaction of β with productivity usually negative (mild significance).
  - Exception: specification (2) where β instrumented with openness dummies yields a highly significant positive coefficient.
  - Note: theory predicts no direct relationship between financial globalization and consumption risk sharing; industry-specific risk-sharing measures would be more appropriate.
- Table 5 (openness × sectoral risk):
  - Risk hampers exports only in financially closed economies (not statistically significant).
  - More open countries show even less negative effects of sectoral risk; fully open countries tend to export and import “too much” in risky sectors (positive coefficient on sectoral risk).
  - Possible explanation: two of riskiest sectors (electronics and machinery) have product differentiation and intra-industry trade not modeled.
- Table 6 (controlling for productivity):
  - Risk has a significant and robust negative impact on trade: for a given productivity differential, riskier sectors show less trade.
  - Sign and magnitude are similar for closed and open countries.
  - However, risk is found to have a more negative impact on trade (for given relative productivity) in more open countries.

- Overall empirical conclusions:
  - Financial openness tends to increase trade responsiveness to comparative advantage and increase trade volumes.
  - Sectoral risk reduces trade driven by productivity differentials; openness moderates some adverse effects of risk but results vary across specifications and sectors.

*Source: Excerpt from IMF working paper content unit _wp03233 - 0. Then the economy will exhibit*

### CONCLUSION

### CONCLUSION

### Theoretical framework
- Developed a fully micro-founded general equilibrium model to assess the impact of financial globalization on industrial specialization and international trade.
- Model's main merit: provides easily interpretable explicit predictions for the patterns of specialization and trade.

### Empirical results
- Empirical evidence confirms most model implications:
  - Trade in financially open countries is (i) higher,
  - (ii) more dependent on productivity differences, and
  - (iii) less sensitive to industry risks,
  - supporting a view of Ricardian trade with incomplete insurance.

- Supporting empirical material (tables reported in the source):
  - Table 1: The Impact of Comparative Advantage on the Volume of Trade — dependent variable: total trade (export+imports) over GNP or log trade; COMPADV is an index of absolute percentage labor productivity difference relative to the rest of the world.
  - Table 2: The Impact of Comparative Advantage by Financial Openness (OPEN Index) — OPEN is an index of financial openness, 0 meaning fully repressed capital flows, 3 corresponding to full liberalization.
  - Table 3: The Impact of Productivity on the Structure of Trade — dependent variable: net trade over consumption or export over consumption in the given sector; PROD is the percentage difference in labor productivity relative to the world average.
  - Table 4a/4b/4c: The Impact of Productivity by Financial Openness (OPEN Index), including weighted estimates and inclusion of consumption risk sharing measures; Beta is a measure of consumption comovement with GDP (Beta=0 means perfect comovement, 1 indicates no comovement).
  - Table 5: The Impact of Risk and Financial Openness — RISK is the standard deviation of annual TFP growth of the given sector in the United States.
  - Table 6: The Impact of Risk and Productivity — interactions of PROD and RISK by OPEN categories.

### Methodological notes from empirical analysis
- Key variables and definitions preserved from source:
  - COMPADV: index of absolute percentage labor productivity difference relative to the rest of the world.
  - PROD: percentage difference in labor productivity relative to the world average.
  - OPEN: index of financial openness, 0 meaning fully repressed capital flows, 3 corresponding to full liberalization.
  - RISK: standard deviation of annual TFP growth of the given sector in the United States.
  - Beta: measure of consumption comovement with GDP (Beta=0 means perfect comovement, 1 indicates no comovement).
- Estimation features reported in tables include: year fixed effects, country fixed effects, country*year fixed effects, sample sizes (observations), number of years, R-squared, and standard errors with significance markers (* at 5 percent, ** at 1 percent).
- Some reported quantitative estimates (as presented in tables) include:
  - Table 1: log(GNP) coefficients reported as 0.889 and 0.914 with (0.010)** and (0.009)** standard errors respectively; COMPADV coefficients reported as -0.191 (0.038)** and -0.762 (0.075)**; weighted COMPADV reported as -0.206 (0.051)** and -0.658 (0.097)**.
  - Table 2: log(GNP) coefficients 0.781 (0.022)** and 0.757 (0.020)**; OPEN category coefficient examples: OPEN = 1 coefficients -0.011 (0.146), 0.462 (0.227)*, 0.051 (0.129), 0.487 (0.182)**; COMPADV (OPEN = 3) coefficients 4.918 (0.435)** and 3.131 (0.731)**; weighted COMPADV (OPEN = 3) 3.545 (0.770)** and 0.316 (1.122).
  - Table 3: PROD coefficients across specifications range from 0.055 to 0.103 with standard errors (0.005)** to (0.007)**; observations 25,649; R-squared values 0.01, 0.12, 0.16 across columns.
  - Table 4a: PROD (OPEN = 0) coefficients 0.085 (0.018)**, 0.104 (0.014)**, 0.084 (0.026)**, 0.107 (0.020)**, 1.987 (0.043)**; PROD (OPEN = 3) coefficients 0.120 (0.025)**, 0.125 (0.021)**, 0.170 (0.035)**, 0.172 (0.029)**, 0.686 (0.057)**; observations and R-squared reported (e.g., Observations 10,189; R-squared 0.01–0.27 depending on column).
  - Table 4b: Weighted PROD (OPEN = 0) 0.045 (0.013)**, 0.064 (0.021)**, 1.297 (0.034)**; PROD (OPEN = 3) 0.068 (0.019)**, 0.063 (0.031)*, 0.739 (0.051)**; Observations 10,189 and 11,512; R-squared 0.06, 0.18, 0.23.
  - Table 4c: PROD coefficients e.g. 1.467 (0.026)**, 1.403 (0.064)**, 0.113 (0.011)**; Beta*PROD coefficients include -0.340 (0.091)** and 1.549 (0.495)**; observations vary by specification (e.g., 18,672; 7,694).
  - Table 5: RISK (OPEN=3) coefficients e.g. 9.886 (1.527)**, 9.220 (0.765)**; PROD (OPEN=0) 0.091 (0.027)**; Observations 10,251 and 10,189; Within R-squared values 0.00–0.02.
  - Table 6: PROD coefficients by OPEN category: 0.089 (0.007)** (overall), 0.110 (0.011)** (OPEN=0), 0.141 (0.019)** (OPEN=1), 0.260 (0.025)** (OPEN=2), 0.221 (0.044)** (OPEN=3); RISK coefficients and PROD*RISK interactions reported with standard errors and significance (e.g., PROD * RISK -1.924 (0.450)** through -8.431 (2.322)**); observations vary (e.g., 25,649; 2,881; 2,197; 2,341; 2,770).
- The tables document heterogeneous effects of PROD and RISK across financial openness categories and show significant interaction terms indicating that the effect of productivity on trade shares depends on sector risk and financial openness.

### Directions for future work
- Identify specific industry factors and investigate countries’ exposure to these factors.
- Concentrate empirical analysis around dates of capital account opening in an “event-study” fashion.
- Include additional controls that may be important determinants of trade flows.

*Source: CONCLUSION (from the author’s paper).*

### 1. From the definition of x

### 1. From the definition of x

### Hedging, forward price π, and expected return
- A portion B'q_j of investments will be hedged: the agent sells the marketable component of her non-financial assets in the forward market for a price of π.
- The second term in the parentheses, Bπ, is equal to the forward value of the marketable portion of one unit of output and adds to the expected return on investment.
- π may be negative; a negative π implies the asset’s forward price is negative (high required return).

### Net demand for financial assets and equilibrium asset prices
- Country j’s net demand for financial assets:
  - Y'_jj = -γ h q B π   (expression as given: ' . jj jj YY π =− − γ hq B)
- World net demand must clear: 0 = Y'_w w = -γ h q B π   (expression as given: 0', ww ww YY π ==−− γ hq B)
- Equilibrium asset prices:
  - π' w = -γ q / Y_w   (expression as given: '. w w Y π = − γ q B)
- Economic interpretation:
  - The forward price of an asset is high if it is negatively correlated with an industry that has a high share in world production (demand for insurance high → asset expensive).
  - If the asset is positively correlated with a large industry, the price can be negative (investors require a low price / high return).

### First-order condition substitution and Theorem 1 result
- Using BB' = Ω − Σ and substituting asset prices into the first-order condition yields (as in source):
  - 1111 0 1 (). j ww j jww R YYY −−−− −= −− γ q q q Σ1PA1ΣΩ
- Straightforward algebra shows (as in source):
  - 111 0 11111 00 1 (), 1 (), w w w j w wj jw R Y RR YY −−− −−− −− =− γ −=− γ q Ω1PA1 q q ΣPA1PA1
- Introducing notation:
  - 11 0 11 0 jj w R R −− −− =− =− m1 PA1 m1 PA1
- These relations give the result of Theorem 1 (as stated in the source).

### Country j’s demand for financial assets and portfolio composition
- Country j’s demand for financial assets (as expressed in the source):
  - ' . jj w jwj YYY ( ) = − γ hq q B   (expression shown: '. jj w jwj YYY  =−    hq q B)
- Portfolio implications:
  - The country will hold long positions (h_j,n > 0) in assets that are positively correlated with industries that are small relative to the world average.
  - The country will hold short positions in assets positively correlated with its large industries.

### Appendix II — Data definitions and examples
- Table A1. Definition of Sectors (examples as listed):
  - Food products: ISIC (rev. 2) 311; BEA 1, 4
  - Beverages: ISIC 313; BEA 2
  - Tobacco: ISIC 314; BEA 3
  - *Textiles + apparel: ISIC 321, 322; BEA 5
  - *Leather products + footwear: ISIC 323, 324; BEA 6
  - Paper and products: ISIC 341; BEA 7, 8
  - Transport equipment: ISIC 384; BEA 28, 29
  - Note: Industries denoted by * contain multiple ISIC categories.
- Table A2. Examples of Financial Openness Indices (selected country time series, 1982–1994; 0 = closed, 3 = fully liberalized capital account; source: Abiad and Mody (2003)):
  - ARG: 1982 0, 1983 0, 1984 0, 1985 0, 1986 0, 1987 0, 1988 0, 1989 0, 1990 0, 1991 3, 1992 3, 1993 3, 1994 3
  - AUS: 1982 0, 1983 0, 1984 3, 1985 3, 1986 3, 1987 3, 1988 3, 1989 3, 1990 3, 1991 3, 1992 3, 1993 3, 1994 3
  - CAN: 1982 3, 1983 3, 1984 3, 1985 3, 1986 3, 1987 3, 1988 3, 1989 3, 1990 3, 1991 3, 1992 3, 1993 3, 1994 3
  - DEU: 1982 3, 1983 3, 1984 3, 1985 3, 1986 3, 1987 3, 1988 3, 1989 3, 1990 3, 1991 3, 1992 3, 1993 3, 1994 3
  - USA: 1982 3, 1983 3, 1984 3, 1985 3, 1986 3, 1987 3, 1988 3, 1989 3, 1990 3, 1991 3, 1992 3, 1993 3, 1994 3
  - IND: 1982 0, 1983 0, 1984 0, 1985 0, 1986 0, 1987 0, 1988 0, 1989 0, 1990 0, 1991 1, 1992 1, 1993 1, 1994 1
  - PER: 1982 0, 1983 0, 1984 0, 1985 0, 1986 0, 1987 0, 1988 0, 1989 0, 1990 0, 1991 3, 1992 3, 1993 3, 1994 3
  - Notes: Table lists many additional country-year index values for 1982–1994.
- Table A3. Riskiness of SIC (1987 version) Sectors in the United States (σ(dTFP) by 2-digit SIC; source: Author’s calculations from Bartelsman, Becker and Gray (2000)):
  - SIC 20 Food and Kindred Products: 0.0132
  - SIC 21 Tobacco Products: 0.0563
  - SIC 22 Textile Mill Products: 0.0192
  - SIC 23 Apparel and Other Textile Products: 0.0154
  - SIC 24 Lumber and Wood Products: 0.0250
  - SIC 25 Furniture and Fixtures: 0.0241
  - SIC 26 Paper and Allied Products: 0.0238
  - SIC 27 Printing and Publishing: 0.0211
  - SIC 28 Chemicals and Allied Products: 0.0271
  - SIC 29 Petroleum and Coal Products: 0.0360
  - SIC 30 Rubber and Miscellaneous Plastics Products: 0.0238
  - SIC 31 Leather and Leather Products: 0.0277
  - SIC 32 Stone, Clay and Glass Products: 0.0225
  - SIC 33 Primary Metal Industries: 0.0307
  - SIC 34 Fabricated Metal Products: 0.0234
  - SIC 35 Industrial Machinery and Equipment: 0.0332
  - SIC 36 Electronic and Other Electric Equipment: 0.0512
  - SIC 37 Transportation Equipment: 0.0192
  - SIC 38 Instruments and Related Products: 0.0207
  - SIC 39 Miscellaneous Manufacturing Industries: 0.0283

*Source: _wp03233 - 1. From the definition of x (IMF Working Paper excerpt).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03233.pdf_