## 8. Comparison of Quantile Regression and OLS Coefficient Estimates

## Source details

**Canonical URL:** [8. Comparison of Quantile Regression and OLS Coefficient Estimates](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13176.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13176.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13176.pdf.json)

---

### Introduction
- Structural transformation: reallocation from low productivity to high productivity activities; canonical shifts from agriculture → manufacturing → services.
- Sample and period: cross-section of advanced and developing economies over the 1970–2010 period; predicted real value added shares obtained from cross-country OLS and quantile regressions controlling for fundamentals (GDP per capita, demographic structure, population, among others).
- Main contributions:
  - Statistical benchmarking to evaluate structure of economies via deviations of actual sectoral shares from predicted levels.
  - Comprehensive empirical analysis of policy and institutional drivers of structural transformation across a wide cross-section of countries.

### Stylized facts
- Sectoral patterns (UN Statistics Division data, 1970–2010, sample of 168 countries):
  - Agriculture: negative relationship with real GDP per capita (log).
  - Manufacturing: inverted U-shaped relationship with real GDP per capita; manufacturing share peaks at GDP per capita (log) of 8.5 (around US$ 5,000).
  - Services: positive relationship with real GDP per capita; acceleration of services share when log GDP per capita reaches about 8.5.
- Regional patterns (weighted averages by real GDP pc):
  - Sub-Saharan Africa highest agriculture share (roughly constant early 1980s to mid-1990s, declined since then).
  - Asia: rapid decline in agriculture and relatively high manufacturing share.
  - Latin America: high and increasing services share (slightly less than 50 percent in early 1970s).
  - Central and Eastern Europe: shows typical AE pattern.
- Country cases:
  - China: agriculture share fell from >45 percent in 1970 to <5 percent by 2010; services share in 2010 slightly below 30 percent despite per capita income around US$7,000 (median services share for upper middle-income countries was 50 percent in 2010).
  - Korea: atypical high-income case where decline in manufacturing and increase in services seen in other AEs has yet to occur.
  - India: relatively high services share.
  - Mexico: relatively high and stable services and mining shares.
- Heterogeneity: large variation within regions and across countries at similar development levels; country fundamentals explain a significant proportion of cross-sectional variation in agriculture and services shares, less so in manufacturing.

### Methodology and data
- Econometric approach:
  - Panel OLS and quantile regressions (quantiles at 10th, 25th, 50th, 75th, 90th percentiles).
  - Quantile regressions allow coefficients to vary across the dependent-variable distribution and avoid sample selection bias from stratified OLS.
- Dependent variables: real value added share in agriculture, manufacturing, services (UN Statistics Division).
- Baseline controls (fundamentals): land area (log), population (log), arable land (percent of total land area), dummy for island economies with populations < 1 million, age dependency ratios (young, old), GDP per capita (log) and its square, mining output share, dummy for transition economies; time dummies included.
- Policy/institutional variables (lagged by one year):
  - Product market reform indices: Agriculture Index and Networks Index (electricity and telecommunications), rescaled 0–1 (higher = greater liberalization).
  - Trade openness: ratio of real exports plus imports to real GDP.
  - Gross FDI inflows in the non-resources sector.
  - Domestic credit to private sector (ratio to GDP).
  - Tertiary education enrollment (share).
- Sample details:
  - Baseline panel: annual data 1970–2010, 168 countries, unbalanced panel.
  - Policy regressions sample reduced due to reform variable availability (e.g., Number of countries = 93; observational counts reported in tables).

### Empirical results

- A. Baseline OLS regressions (Table 1)
  - Agriculture (OLS highlights):
    - Land area [Log]: coefficient 1.083 (robust SE [0.081], significance ***).
    - Population [log]: coefficient -1.893 (SE [0.103], ***).
    - GDPPC [log]: coefficient -34.648 (SE [0.714], ***); Square of (GDPPC [log]) 1.54 (SE [0.045], ***).
    - Mining output share: -0.052 (SE [0.007], ***).
    - R-squared: 0.795; Observations: 5,341; Number of countries: 168.
  - Manufacturing (OLS highlights):
    - Population [log]: 0.983 (SE [0.060], ***).
    - GDPPC [log]: 7.788 (SE [0.414], ***); Square -0.366 (SE [0.026], ***).
    - Mining output share: -0.192 (SE [0.004], ***).
    - R-squared: 0.516; Observations: 5,341.
  - Services (OLS highlights):
    - Age dependency ratios positive for services; GDPPC [log] 14.33 (SE [0.895], ***); Square -0.522 (SE [0.056], ***).
    - R-squared: 0.661; Observations: 5,341.
  - Overall explanatory power: fundamentals explain close to 70 to 80 percent of cross-country variation in services and agricultural shares; about 52 percent for manufacturing.

- B. Quantile regression vs OLS (Table 2; Figure 8)
  - Quantile estimates differ substantially from OLS: coefficients vary across the 10th, 25th, 50th, 75th, 90th quantiles.
  - Agriculture:
    - Importance of land area and arable land strengthens as agriculture share rises.
    - Mining output share negative and significant across quantiles (examples: agriculture quantile 10 coefficient -0.041 [SE 0.009]***; quantile 90 coefficient -0.017 [SE 0.012]).
  - Manufacturing:
    - Population coefficient increases along the distribution (evidence of increasing returns to scale in manufacturing).
    - Trade openness positive and highly significant across manufacturing distribution in quantile regressions.
  - Services:
    - Age dependency ratios positive and significant, but effects weaken along the distribution except at the top.
    - Tertiary education and domestic credit show increasing importance along services distribution.
  - Key numerical examples from quantile results:
    - Agricultural share quantiles (pseudo R-squares): 10th 0.422; 25th 0.527; 50th 0.604; 75th 0.639; 90th 0.648 (Observations 5,341).
    - Manufacturing quantiles (pseudo R-squares): 10th 0.327; 25th 0.372; 50th 0.364; 75th 0.299; 90th 0.259.
    - Services quantiles (pseudo R-squares): 10th 0.435; 25th 0.463; 50th 0.451; 75th 0.45; 90th 0.448.

- C. Actual versus predicted shares (quantile-based benchmarking)
  - Using median (50th percentile) quantile regressions to compute fitted values, group averages:
    - Agriculture: EMEs average sectoral shares consistently below predicted levels; LICs the opposite.
    - Manufacturing: AEs on average below predicted levels until the “great moderation”; EMEs positive gaps since 1980s; LICs large negative gaps since 1990s.
    - Services: across country groups, levels higher than predicted by fundamentals particularly since the 2000s.
  - Cross-country heterogeneity (2010, Figure 7):
    - LICs: many Sub-Saharan African countries with real per capita incomes < US$300 have agricultural shares above predicted levels.
    - Manufacturing outliers with large positive gaps: China, Thailand, Indonesia, Korea, Malaysia; also El Salvador, Argentina, Trinidad and Tobago.
    - Manufacturing below-predicted: resource-rich countries (e.g., Nigeria, Sierra Leone, Botswana, Kazakhstan), some Middle East countries (Algeria, Syria).
    - Advanced economies: agriculture share very low (~2–3 percent); services >60 percent in most AEs; manufacturing positive outliers include Ireland, Japan, Germany, Finland; US, UK, France, Greece, Luxembourg have manufacturing shares well below predicted levels.

- D. Role of policy and institutional variables (Tables 3–5; Figure 8)
  - OLS (Table 3) key coefficients (policy variables, lagged):
    - Tertiary education [lag]: Agriculture -4.061 (SE [1.257], ***); Manufacturing -0.587 (SE [0.838], n.s.); Services 2.69 (SE [1.710], n.s.).
    - Agriculture Index [lag]: Agriculture -2.442 (SE [0.335], ***); Manufacturing 0.718 (SE [0.223], ***); Services 0.759 (SE [0.456], *).
    - Networks Index [lag]: Agriculture -2.227 (SE [0.742], ***); Manufacturing -0.444 (SE [0.495], n.s.); Services 3.473 (SE [1.010], ***).
    - Trade openness [lag]: Agriculture -1.413 (SE [0.468], ***); Manufacturing 1.631 (SE [0.312], ***).
    - FDI & Non-resource countries [lag]: Agriculture 0.02 (SE [0.033], n.s.); Manufacturing 0.052 (SE [0.022], **); Services -0.126 (SE [0.045], ***).
    - Domestic credit to private [lag]: positive association with services (coefficient 0.75, SE [0.661], n.s. in OLS but significance varies in quantiles).
  - Quantile regression findings (Table 4; Figure 8):
    - Agricultural policy reforms: negative and statistically significant for agriculture; positively associated with manufacturing and services, especially at upper ends of their distributions.
      - Example: agriculture reforms coefficient in manufacturing regressions increases from -2 at 10th percentile to over 1 for above-median observations.
    - Networks reforms: positively and significantly associated with manufacturing and services at low levels (below-median), importance declines along distribution.
    - Tertiary education: becomes positive and significant for above-median manufacturing shares; strongly increasing along services share distribution.
    - Domestic credit: positive and significant for lower end of manufacturing distribution (10th percentile); private credit strongly significant at median and 75th percentile of services distribution.
    - Trade openness: positive and highly significant for manufacturing across all quantiles; negative and significant for countries with below-median services shares, positive at 90th percentile for services.
    - FDI: significant mainly for low and above-median manufacturing shares; negative and significant for services.
  - Time evolution (Table 5: pre-1992 vs post-1992):
    - Pre-1992: network reforms positively associated with above-median agriculture shares; FDI more important for agriculture and manufacturing in pre-globalization period.
    - Post-1992 (globalization period): network reforms strongly positively associated with below-median manufacturing shares; trade openness becomes more important across manufacturing distribution; FDI positively correlated with services share during globalization period.

### Robustness
- Sample restrictions: restricting to EMEs and LICs (Table 6) leaves main findings intact; explanatory power roughly the same.
- Alternative variables (Table 7):
  - Replaced FDI with capital account openness (Chinn and Ito index), trade openness with trade liberalization (average tariff rates), domestic credit with financial sector reforms index, tertiary with secondary enrollment.
  - Financial reform index: strong explanatory power for services share (significant across distribution); financial sector reforms negatively associated with manufacturing and agriculture shares.
  - Capital account openness: strongly positive for agriculture and manufacturing shares at various quantiles.
  - Low average tariff rates relevant across agriculture distribution, negatively associated with manufacturing share.
- Additional controls (Tables 8.1–8.3):
  - Capital stock (log real capital stock): negatively and significantly associated with agriculture and services shares; positively and significantly associated with manufacturing share. Importance decreases along manufacturing distribution.
  - Polity IV (quality of political institutions): significant determinant of agriculture share across distribution; more important for services at lower end; negatively and significantly associated with manufacturing share.
  - Real effective exchange rate (REER, log): appreciation associated with higher agriculture shares and above-median services shares; depreciation associated with higher manufacturing share at lower end of distribution (price-sensitivity).
  - Labor regulation (index of labor market flexibility): greater flexibility strongly associated with higher agriculture shares and with services shares at median and 75th percentiles; little evidence that labor flexibility explains variation in manufacturing shares (coefficients negative and significant across most of manufacturing distribution).
  - Globalization index (Dreher et al., 2008): greater globalization strongly negatively associated with agriculture shares; positive and significant for manufacturing below median and positive across services distribution.
  - Aggregate TFP: little evidence of significant association with sectoral value added shares (consistent with Nickell et al., 2008).

### Policy findings and recommendations
- Product market reforms:
  - Liberalizing and enhancing productivity in the agricultural sector in EMEs and LICs can release resources to other sectors and raise incomes (agriculture reforms associated with lower agriculture shares and higher manufacturing/services shares).
  - Network reforms (electricity and telecommunications) are positively associated with services shares and important for manufacturing at low-manufacturing-share countries.
- Trade and exchange rate policy:
  - Trade openness strongly associated with higher manufacturing shares; attention needed to promote price competitiveness and diversified trade.
  - Real exchange rate level matters for manufacturing at low levels of manufacturing share (depreciation associated with higher manufacturing share at lower quantiles).
- Human and physical capital:
  - Tertiary education important for higher manufacturing shares (above-median) and strongly for services distribution; physical capital accumulation (infrastructure such as power and transport) supports reallocation to manufacturing.
- Finance and labor market:
  - Adequate financing (credit, FDI) and financial sector reforms facilitate resource reallocation and services growth.
  - Greater labor market flexibility associated with higher services shares (median and upper quantiles) and higher agriculture shares; evidence for manufacturing mixed.
- Heterogeneity emphasis:
  - Determinants of sectoral shares vary across points in the distribution and over time; policy prescriptions should account for country-specific position in sectoral-share distributions and stages of globalization.

### Key statistics and model diagnostics
- Baseline OLS (Table 1) observations: 5,341; Number of countries: 168.
- OLS R-squared: Agriculture 0.795; Manufacturing 0.516; Services 0.661.
- Quantile regressions (full sample, Table 2) pseudo R-square examples:
  - Agriculture quantile 50: 0.604 (Observations 5,341).
  - Manufacturing quantile 50: 0.364.
  - Services quantile 50: 0.451.
- Policy OLS regressions (Table 3) sample: Observations 1,684; Number of countries 93; (pseudo) R-square: Agriculture 0.862; Manufacturing 0.505; Services 0.708.
- Time-split quantile regressions (Table 5) sample sizes:
  - Pre-1992 observations: 951; Number of countries: 68.
  - Post-1992 observations: 733; Number of countries: 91.
- Robustness alternative-variable regressions (Table 7) observations: 1,261; Number of countries: 71.

### Conclusion
- Country fundamentals (endowments, population, dependency ratios) explain a large proportion of variation in agriculture and services sector shares; less explanatory power for manufacturing.
- Structural reforms, globalization, human and physical capital, financial development, and institutional factors matter for observed patterns of structural change; their relevance varies across the distribution of sectoral shares and over time.
- Policy implications:
  - Liberalize and raise productivity in agriculture to free resources for higher-productivity sectors in EMEs and LICs.
  - Promote trade openness and price competitiveness to support manufacturing, particularly for countries with low manufacturing shares.
  - Invest in physical infrastructure and financial deepening to enable reallocation.
  - For countries with high manufacturing shares, emphasize innovation and deeper human capital; for services, reform distribution sectors (telecommunications and electricity), strengthen human capital, and enhance labor market flexibility.
- Caveats: potential endogeneity of policy and institutional variables; models explain agricultural and services shares well but less so manufacturing; intra-sectoral productivity dynamics require further analysis.

*Source: _wp13176 - 8. Comparison of Quantile Regression and OLS Coefficient Estimates (IMF Staff working paper), pages 3–39.*

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

### _wp13176 - References

### Tables
- 1. Role of Country Fundamentals: OLS Regressions ..............................................................29
- 2. Role of Country Fundamentals (Full Sample): Quantile Regressions .................................30
- 3. Role of Policy and Institutional Factors: OLS Regressions.................................................31
- 4. Role of Policy and Institutional Factors: Quantile Regressions ..........................................32
- 5. Drivers of Diversification over Different Time Horizons: Quantile Regressions ...............33
- 6. Role of Country Fundamentals (EMEs and LICs): Quantile Regressions ..........................34
- 7. Alternative Independent Variables: Quantile Regressions ..................................................35
- 8.1 Additional Controls-Agriculture: Quantile Regressions ....................................................36
- 8.2 Additional Controls-Manufacturing: Quantile Regressions ..............................................37
- 8.3 Additional Controls- Services: Quantile Regressions........................................................38

### Figures
- 1. Real Value Added Shares and Real GDP per capita, 1970-2010 ........................................21
- 2. Shares by Sector of Economic Activity in EMs and LICs by Region .................................22
- 3. Mining vs. Manufacturing and Services Shares in Total Value Added ...............................23
- 4. Real Value Added Shares by Sector of Economic Activity in Selected LICs .....................24
- 5. Real Value Added Shares by Sector of Economic Activity in Selected EMEs ...................25
- 6. Actual vs. Predicted Sectoral Shares Across Income Groups, 1970-2010 ..........................26
- 7. Actual vs. Predicted (Median) Shares by Countries, 2010 ..................................................27

*Source: _wp13176 - References.*

### 8. Comparison of Quantile Regression and OLS Coefficient Estimates .................................28

### 8. Comparison of Quantile Regression and OLS Coefficient Estimates

### Introduction
- Structural transformation: reallocation from low productivity to high productivity activities; canonical shifts from agriculture → manufacturing → services.
- Sample and period: cross-section of advanced and developing economies over the 1970–2010 period; predicted real value added shares obtained from cross-country OLS and quantile regressions controlling for fundamentals (GDP per capita, demographic structure, population, among others).
- Main contributions:
  - Statistical benchmarking to evaluate structure of economies via deviations of actual sectoral shares from predicted levels.
  - Comprehensive empirical analysis of policy and institutional drivers of structural transformation across a wide cross-section of countries.

### Stylized facts
- Sectoral patterns documented (UN Statistics Division data, 1970–2010, sample of 168 countries):
  - Agriculture: negative relationship with real GDP per capita (log).
  - Manufacturing: inverted U-shaped relationship with real GDP per capita; manufacturing share peaks at GDP per capita (log) of 8.5 (around US$ 5,000).
  - Services: positive relationship with real GDP per capita; acceleration of services share when log GDP per capita reaches about 8.5.
- Regional patterns (weighted averages by real GDP pc): sub-Saharan Africa highest agriculture share (roughly constant early 1980s to mid-1990s, declined since then); Asia rapid decline in agriculture and relatively high manufacturing share; Latin America high and increasing services share (slightly less than 50 percent in early 1970s); Central and Eastern Europe shows typical AE pattern.
- Country cases:
  - China: agriculture share fell from >45 percent in 1970 to <5 percent by 2010; services share in 2010 slightly below 30 percent despite per capita income around US$7,000 (median services share for upper middle-income countries was 50 percent in 2010).
  - Korea: atypical high-income case where decline in manufacturing and increase in services seen in other AEs has yet to occur.
  - India: relatively high services share.
  - Mexico: relatively high and stable services and mining shares.
- Heterogeneity: large variation within regions and across countries at similar development levels; country fundamentals explain a significant proportion of cross-sectional variation in agriculture and services shares, less so in manufacturing.

### Methodology and data
- Econometric approach:
  - Panel OLS and quantile regressions (quantiles at 10th, 25th, 50th, 75th, 90th percentiles).
  - Quantile regressions allow coefficients to vary across the dependent-variable distribution and avoid sample selection bias from stratified OLS.
- Dependent variables: real value added share in agriculture, manufacturing, services (UN Statistics Division).
- Baseline controls (fundamentals): land area (log), population (log), arable land (percent of total land area), dummy for island economies with populations < 1 million, age dependency ratios (young, old), GDP per capita (log) and its square, mining output share, dummy for transition economies; time dummies included.
- Policy/institutional variables (lagged by one year):
  - Product market reform indices: Agriculture Index and Networks Index (electricity and telecommunications), rescaled 0–1 (higher = greater liberalization).
  - Trade openness: ratio of real exports plus imports to real GDP.
  - Gross FDI inflows in the non-resources sector.
  - Domestic credit to private sector (ratio to GDP).
  - Tertiary education enrollment (share).
- Sample details:
  - Baseline panel: annual data 1970–2010, 168 countries, unbalanced panel.
  - Policy regressions sample reduced due to reform variable availability (e.g., Number of countries = 93; observational counts reported in tables).

### Empirical results

#### A. Baseline OLS regressions (Table 1)
- Agriculture (OLS highlights):
  - Land area [Log]: coefficient 1.083 (robust SE [0.081], significance ***).
  - Population [log]: coefficient -1.893 (SE [0.103], ***).
  - GDPPC [log]: coefficient -34.648 (SE [0.714], ***); Square of (GDPPC [log]) 1.54 (SE [0.045], ***).
  - Mining output share: -0.052 (SE [0.007], ***).
  - R-squared: 0.795; Observations: 5,341; Number of countries: 168.
- Manufacturing (OLS highlights):
  - Population [log]: 0.983 (SE [0.060], ***).
  - GDPPC [log]: 7.788 (SE [0.414], ***); Square -0.366 (SE [0.026], ***).
  - Mining output share: -0.192 (SE [0.004], ***).
  - R-squared: 0.516; Observations: 5,341.
- Services (OLS highlights):
  - Age dependency ratios positive for services; GDPPC [log] 14.33 (SE [0.895], ***); Square -0.522 (SE [0.056], ***).
  - R-squared: 0.661; Observations: 5,341.
- Overall explanatory power: fundamentals explain close to 70 to 80 percent of cross-country variation in services and agricultural shares; about 52 percent for manufacturing.

#### B. Quantile regression vs OLS (Table 2; Figure 8)
- Quantile estimates differ substantially from OLS: coefficients vary across the 10th, 25th, 50th, 75th, 90th quantiles.
- Agriculture:
  - Importance of land area and arable land strengthens as agriculture share rises.
  - Mining output share negative and significant across quantiles (examples: agriculture quantile 10 coefficient -0.041 [SE 0.009]***; quantile 90 coefficient -0.017 [SE 0.012]).
- Manufacturing:
  - Population coefficient increases along the distribution (evidence of increasing returns to scale in manufacturing).
  - Trade openness positive and highly significant across manufacturing distribution in quantile regressions.
- Services:
  - Age dependency ratios positive and significant, but effects weaken along the distribution except at the top.
  - Tertiary education and domestic credit show increasing importance along services distribution.

Key numerical examples from quantile results:
- Agricultural share quantiles (pseudo R-squares): 10th 0.422; 25th 0.527; 50th 0.604; 75th 0.639; 90th 0.648 (Observations 5,341).
- Manufacturing quantiles (pseudo R-squares): 10th 0.327; 25th 0.372; 50th 0.364; 75th 0.299; 90th 0.259.
- Services quantiles (pseudo R-squares): 10th 0.435; 25th 0.463; 50th 0.451; 75th 0.45; 90th 0.448.

#### C. Actual versus predicted shares (quantile-based benchmarking)
- Using median (50th percentile) quantile regressions to compute fitted values, group averages:
  - Agriculture: EMEs average sectoral shares consistently below predicted levels; LICs the opposite.
  - Manufacturing: AEs on average below predicted levels until the “great moderation”; EMEs positive gaps since 1980s; LICs large negative gaps since 1990s.
  - Services: across country groups, levels higher than predicted by fundamentals particularly since the 2000s.
- Cross-country heterogeneity (2010, Figure 7):
  - LICs: many Sub-Saharan African countries with real per capita incomes < US$300 have agricultural shares above predicted levels.
  - Manufacturing outliers with large positive gaps: China, Thailand, Indonesia, Korea, Malaysia; also El Salvador, Argentina, Trinidad and Tobago.
  - Manufacturing below-predicted: resource-rich countries (e.g., Nigeria, Sierra Leone, Botswana, Kazakhstan), some Middle East countries (Algeria, Syria).
  - Advanced economies: agriculture share very low (~2–3 percent); services >60 percent in most AEs; manufacturing positive outliers include Ireland, Japan, Germany, Finland; US, UK, France, Greece, Luxembourg have manufacturing shares well below predicted levels.

#### D. Role of policy and institutional variables (Tables 3–5; Figure 8)
- OLS (Table 3) key coefficients (policy variables, lagged):
  - Tertiary education [lag]: Agriculture -4.061 (SE [1.257], ***); Manufacturing -0.587 (SE [0.838], n.s.); Services 2.69 (SE [1.710], n.s.).
  - Agriculture Index [lag]: Agriculture -2.442 (SE [0.335], ***); Manufacturing 0.718 (SE [0.223], ***); Services 0.759 (SE [0.456], *).
  - Networks Index [lag]: Agriculture -2.227 (SE [0.742], ***); Manufacturing -0.444 (SE [0.495], n.s.); Services 3.473 (SE [1.010], ***).
  - Trade openness [lag]: Agriculture -1.413 (SE [0.468], ***); Manufacturing 1.631 (SE [0.312], ***).
  - FDI & Non-resource countries [lag]: Agriculture 0.02 (SE [0.033], n.s.); Manufacturing 0.052 (SE [0.022], **); Services -0.126 (SE [0.045], ***).
  - Domestic credit to private [lag]: positive association with services (coefficient 0.75, SE [0.661], n.s. in OLS but significance varies in quantiles).
- Quantile regression findings (Table 4; Figure 8):
  - Agricultural policy reforms: negative and statistically significant for agriculture; positively associated with manufacturing and services, especially at upper ends of their distributions.
    - Example: agriculture reforms coefficient in manufacturing regressions increases from -2 at 10th percentile to over 1 for above-median observations.
  - Networks reforms: positively and significantly associated with manufacturing and services at low levels (below-median), importance declines along distribution.
  - Tertiary education: becomes positive and significant for above-median manufacturing shares; strongly increasing along services share distribution.
  - Domestic credit: positive and significant for lower end of manufacturing distribution (10th percentile); private credit strongly significant at median and 75th percentile of services distribution.
  - Trade openness: positive and highly significant for manufacturing across all quantiles; negative and significant for countries with below-median services shares, positive at 90th percentile for services.
  - FDI: significant mainly for low and above-median manufacturing shares; negative and significant for services.
- Time evolution (Table 5: pre-1992 vs post-1992):
  - Pre-1992: network reforms positively associated with above-median agriculture shares; FDI more important for agriculture and manufacturing in pre-globalization period.
  - Post-1992 (globalization period): network reforms strongly positively associated with below-median manufacturing shares; trade openness becomes more important across manufacturing distribution; FDI positively correlated with services share during globalization period.

### Robustness
- Sample restrictions: restricting to EMEs and LICs (Table 6) leaves main findings intact; explanatory power roughly the same.
- Alternative variables (Table 7):
  - Replaced FDI with capital account openness (Chinn and Ito index), trade openness with trade liberalization (average tariff rates), domestic credit with financial sector reforms index, tertiary with secondary enrollment.
  - Financial reform index: strong explanatory power for services share (significant across distribution); financial sector reforms negatively associated with manufacturing and agriculture shares.
  - Capital account openness: strongly positive for agriculture and manufacturing shares at various quantiles.
  - Low average tariff rates relevant across agriculture distribution, negatively associated with manufacturing share.
- Additional controls (Tables 8.1–8.3):
  - Capital stock (log real capital stock): negatively and significantly associated with agriculture and services shares; positively and significantly associated with manufacturing share. Importance decreases along manufacturing distribution.
  - Quality of political institutions (Polity IV): significant determinant of agriculture share across distribution; more important for services at lower end; negatively and significantly associated with manufacturing share.
  - Real effective exchange rate (REER, log): appreciation associated with higher agriculture shares and above-median services shares; depreciation associated with higher manufacturing share at lower end of distribution (price-sensitivity).
  - Labor regulation (index of labor market flexibility): greater flexibility strongly associated with higher agriculture shares and with services shares at median and 75th percentiles; little evidence that labor flexibility explains variation in manufacturing shares (coefficients negative and significant across most of manufacturing distribution).
  - Globalization index (Dreher et al., 2008): greater globalization strongly negatively associated with agriculture shares; positive and significant for manufacturing below median and positive across services distribution.
  - Aggregate TFP: little evidence of significant association with sectoral value added shares (consistent with Nickell et al., 2008).

### Policy findings and recommendations (as presented in the paper)
- Product market reforms:
  - Liberalizing and enhancing productivity in the agricultural sector in EMEs and LICs can release resources to other sectors and raise incomes (agriculture reforms associated with lower agriculture shares and higher manufacturing/services shares).
  - Network reforms (electricity and telecommunications) are positively associated with services shares and important for manufacturing at low-manufacturing-share countries.
- Trade and exchange rate policy:
  - Trade openness strongly associated with higher manufacturing shares; attention needed to promote price competitiveness and diversified trade.
  - Real exchange rate level matters for manufacturing at low levels of manufacturing share (depreciation associated with higher manufacturing share at lower quantiles).
- Human and physical capital:
  - Tertiary education important for higher manufacturing shares (above-median) and strongly for services distribution; physical capital accumulation (infrastructure such as power and transport) supports reallocation to manufacturing.
- Finance and labor market:
  - Adequate financing (credit, FDI) and financial sector reforms facilitate resource reallocation and services growth.
  - Greater labor market flexibility associated with higher services shares (median and upper quantiles) and higher agriculture shares; evidence for manufacturing mixed.
- Heterogeneity emphasis:
  - Determinants of sectoral shares vary across points in the distribution and over time; policy prescriptions should account for country-specific position in sectoral-share distributions and stages of globalization.

### Key statistics and model diagnostics (selected exact figures)
- Baseline OLS (Table 1) observations: 5,341; Number of countries: 168.
- OLS R-squared: Agriculture 0.795; Manufacturing 0.516; Services 0.661.
- Quantile regressions (full sample, Table 2) pseudo R-square examples:
  - Agriculture quantile 50: 0.604 (Observations 5,341).
  - Manufacturing quantile 50: 0.364.
  - Services quantile 50: 0.451.
- Policy OLS regressions (Table 3) sample: Observations 1,684; Number of countries 93; (pseudo) R-square: Agriculture 0.862; Manufacturing 0.505; Services 0.708.
- Time-split quantile regressions (Table 5) sample sizes:
  - Pre-1992 observations: 951; Number of countries: 68.
  - Post-1992 observations: 733; Number of countries: 91.
- Robustness alternative-variable regressions (Table 7) observations: 1,261; Number of countries: 71.

### Conclusion
- Country fundamentals (endowments, population, dependency ratios) explain a large proportion of variation in agriculture and services sector shares; less explanatory power for manufacturing.
- Structural reforms, globalization, human and physical capital, financial development, and institutional factors matter for observed patterns of structural change; their relevance varies across the distribution of sectoral shares and over time.
- Policy implications:
  - Liberalize and raise productivity in agriculture to free resources for higher-productivity sectors in EMEs and LICs.
  - Promote trade openness and price competitiveness to support manufacturing, particularly for countries with low manufacturing shares.
  - Invest in physical infrastructure and financial deepening to enable reallocation.
  - For countries with high manufacturing shares, emphasize innovation and deeper human capital; for services, reform distribution sectors (telecommunications and electricity), strengthen human capital, and enhance labor market flexibility.
- Caveats: potential endogeneity of policy and institutional variables; models explain agricultural and services shares well but less so manufacturing; intra-sectoral productivity dynamics require further analysis.

*Source: _wp13176 - 8. Comparison of Quantile Regression and OLS Coefficient Estimates (IMF Staff working paper), pages 3–39.*

### References

### _wp13176 - References

### References
- Acemoglu, Daron and Veronica Guerrieri (2008), “Capital deepening and non-balanced economic growth,” Journal of Political Economy, Vol. 116, No.3, June 2008, pp. 467-498.
- Bah, El-hadj (2011), “Structural Transformation Paths Across Countries,” Emerging Markets Finance and Trade, May-June 2011, Vol. 47, Supplement 3, pp. 5-19.
- Bernstein, Jeffrey R., and David E. Weinstein (2002), “Do endowments predict the location of production? Evidence from national and international data,” Journal of International Economics Vol. 56, pp. 55-76.
- Buera, Francisco J., and Joseph P. Kaboski (2012a), “The Rise of the Service Economy,” American Economic Review, Vol. 102, Issue 6, pp. 2540-2569.
- Buera, Francisco J., and Joseph P. Kaboski (2012b), “Scale and the Origins of Structural Change,” Journal of Economic Theory, Vol. 147, Issue 2, pp. 684-712.
- Chenery, Hollis B. (1960), “Patterns of Industrial Growth,” American Economic Review, Vol. 50, No. 3, September 1960, pp. 624–54.
- Dreher, Axel, Noel Gaston, and Pim Martens, 2008, Measuring Globalization - Gauging its Consequences, New York: Springer.
- Duarte, Margarida, and Diego Restuccia (2010), “The Role of the Structural Transformation in Aggregate Productivity,” Quarterly Journal of Economics, Vol. 125, No. 1, February 2010, pp. 129-173.
- Echevarria, Cristina (1997), “Changes in Sectoral Composition Associated with Economic Growth,” International Economic Review, Vol. 38, No. 2, May 1997, pp. 431-452.
- Echevarria, Cristina (2000), “Nonhomothetic Preferences and Growth,” Journal of International Trade and Economic Development, Vol. 9, No. 2, June 2000, pp. 151-172.
- Herrendorf, Berthold, Richard Rogerson, and Akos Valentinyi (2013), “Growth and Structural Transformation,” NBER Working Paper No. 18996, Cambridge, MA: National Bureau of Economic Research, April 2013.
- International Monetary Fund (2013), “Jobs and Growth: Analytical and Operational Considerations for the Fund,” International Monetary Fund: Washington, DC.
- Kaldor, N., Causes of the slow rate of growth of the United Kingdom, Cambridge, Cambridge University Press, 1966.
- Kongsamut, Piyabha, Sergio Rebelo, and Danyang Xie (2001), “Beyond Balanced Growth,” Review of Economic Studies, Vol. 68, October 2001, pp. 869-82.
- Kuznets, Simon (1966), Modern Economic Growth, New Haven, CT: Yale University Press.
- Kuznets, Simon (1973), “Modern Economic Growth: Findings and Reflections,” Nobel Lecture, American Economic Review, Vol. 63, June 1973, pp. 247-258.
- Kuznets, Simon (1971), Economic Growth of Nations: Total Output and Production Structure, Cambridge, Mass.: Harvard University Press.
- Jaumotte, Florence, and Nikola Spatafora (2007), “Asia Rising: A Sectoral Perspective,” IMF Working Paper WP/07/130, Washington DC: International Monetary Fund.
- Kochhar, Kalpana, Utsav Kumar, Raghuram Rajan, Arvind Subramanian, and Ioannis Tokatlidis (2006), “India’s Pattern of Development: What Happened, What Follows?” IMF Working Paper WP/02/22, Washington DC: International Monetary Fund.
- Lee, Donghoon, and Kenneth I. Wolpin (2006), “Intersectoral Labor Mobility and the Growth of the Service Sector,” Econometrica, Vol. 74, Issue 1, pp. 1-46.
- Matsuyama, Kiminori (2008), “Structural Change,” in The New Palgrave Dictionary of Economics, Second Edition, Steven N. Durlauf and Lawrence E. Blume, editors, New York, NY: Palgrave Macmillan.
- McMillan, Margaret, and Dani Rodrik (2011), “Globalization, structural change and productivity growth,” Chapter 2 in Making Globalization Socially Sustainable, Marc Bacchetta and Marion Jansen, eds. Geneva, Switzerland: International Labour Organization and World Trade Organization.
- Ngai, Rachel, and Christopher Pissarides (2007), “Structural Change in a Multisector Model of Growth,” American Economic Review, Vol. 97, No. 1, March 2007, pp. 429-443.
- Nickell, Stephen, Stephen Redding, and Joanna Swaffield (2008), “The Uneven Pace of Deindustrialisation in the OECD,” The World Economy, Vol. 31, Issue 9, pp. 1154-1184.
- OECD (2012), Economic Policy Reforms 2012: Going for Growth (Paris: Organisation for Economic Cooperation and Development).
- Pratti, Alessandro, Massimiliano Onorato, and Chris Papageorgiou “Which Reforms Work and under What Institutional Environment: Evidence from a New Dataset on Structural Reforms,” forthcoming, Review of Economics and Statistics.
- Schott, Peter K. (2003) “One Size Fits All? Heckscher-Ohlin Specialization in Global Production,” The American Economic Review, Vol. 93(3), pp. 686-708.

### Appendix I: List of Countries in the Sample
- Note: LICs: IMF PRGT-eligible countries. Advanced and Emerging: IMF World Economic Outlook classification
- Countries with a star are included in the OLS and Quantile augmented regressions (Table 3 and 4)

- Advanced (selection as presented in source):
  - United States of America *AlbaniaPhilippines *AfghanistanLesotho
  - United Kingdom *AlgeriaPoland *Armenia *Liberia
  - Austria *AngolaQatarBangladesh *Madagascar *
  - Belgium *Argentina *RomaniaBenin *Malawi *
  - Denmark *Azerbaijan *RussiaBhutanMaldives
  - France *Bahamas *Saudi Arabia *Bolivia *Mali
  - Germany *BahrainSerbiaBurkina Faso *Moldova *
  - Italy *BarbadosSouth Africa *BurundiMongolia *
  - LuxembourgBelarus *Sri Lanka *CambodiaMozambique *
  - Netherlands *Belize *SurinameCameroon *Myanmar
  - Norway *Bosnia and HerzegovinaSwazilandCape VerdeNepal *
  - Sweden *BotswanaSyriaCentral African Republic *Nicaragua
  - Switzerland  *Brazil  *Thailand  *Chad  *Niger
  - Canada *BruneiTrinidad *ComorosNigeria *
  - Japan *Bulgaria *Tunisia *Congo, DR - ZairePapua New Guinea
  - Finland *Chile *Turkey *Congo, Rep.Rwanda
  - Greece *ChinaTurkmenistanDjiboutiSaint Lucia *
  - Iceland *Colombia *Ukraine *EritreaSaint Vincent and the Grenadines *
  - Ireland *Costa RicaUnited Arab EmiratesEthiopia *Senegal
  - MaltaCroatiaUruguay *Gambia *Sierra Leone *
  - Portugal *Dominican RepublicVenezuela *Georgia *Solomon Islands
  - Spain *Ecuador *ZimbabweGhana *Sudan
  - Australia *Egypt *GrenadaTajikistan
  - New Zealand *El SalvadorGuineaTanzania *
  - CyprusEquatorial GuineaGuinea-BissauTogo *
  - IsraelFijiGuyana *Tonga
  - SingaporeGabonHaitiUganda *
  - Czech Republic *Guatemala *Honduras *Uzbekistan
  - Slovakia *Hungary *Kenya *Vanuatu
  - EstoniaIndia *Kyrgyz Republic *Western Samoa
  - SloveniaIndonesia *LaosZambia *
  - Iran
  - Iraq
  - Jamaica *
  - Jordan *
  - Kazakhstan
  - Korea *
  - Kuwait
  - Latvia *
  - Libya
  - Lithuania *
  - Macedonia
  - Malaysia *
  - Mauritius
  - Mexico *
  - Montenegro
  - Morocco *
  - Namibia *
  - Oman *
  - Pakistan *
  - Panama
  - Paraguay
  - Peru *

- Emerging/LICs: (labels and country listings continue as in source; see note on starred entries above)

### Appendix II: Data Description

H3 Dependent Variables
- Agriculture, manufacturing, services
  - The real value added share of agriculture, manufacturing and services sectors in total value added, constructed from the real value added by economic activity series.
  - Source: UN data
- Structural Variables
  - Transition [D]: Binary variable equals to 1 if the country is transition economy
  - Island [D]: Binary variable equals to 1 if the country is island economy with population less than 1 million
- Mining, % of total value added
  - Sectoral real value added share of mining, in percent
  - Source: UNdata
- Land Area [log]
  - Natural log of land area, in sq km
  - Source: World Bank: World Development Indicators
- Population [log]
  - Natural log of total population
  - Source: World Bank: World Development Indicators
- Arable land, % of total
  - Percentage of arable land of total land area, in percent
  - Source: World Bank: World Development Indicators
- Age dependency ratio - Young
  - Age dependency ratio, young, as a percentage of working-age population
  - Source: World Bank: World Development Indicators
- Age dependency ratio - Old
  - Age dependency ratio, old as a percentage of working-age population
  - Source: World Bank: World Development Indicators
- GDPPC [log]
  - Natural log of GDP per capita, in constant PPP U.S. dollar
  - Source: IMF: World Economic Outlook
- Square of (GDPPC, [log])
  - Square term of the natural log of GDP per capita, in constant PPP U.S. dollar
  - Source: IMF: World Economic Outlook

H3 Policy Variables
- Tertiary Education
  - Total enrollment in tertiary ducation, regardless of age, expressed as a percentage of the total population of the five-year age group following on from secondary school leaving.
  - Source: World Bank: World Development Indicators
- Agriculture Index
  - The index captures intervention in the market for the main agricultural export commodity in each country. The index can take four values: 1) zero - public monopoly or monopsony in production, transportation, or marketing; 2) 1/3 - administered prices; 3) 2/3 - public ownership of relevant producers or concession requirements; 4) 1 - no public intervention. All values are normalized to lie between 0 to 1.
  - Source: IMF: Giuliano, Mishra and Spilimbergo (2010)
- Network Index
  - This index is constructed by summing the two underlying indicators, electricity and telecommunication provisions and rescaling between 0 to 1. The electricity indicators capture 1) the degree of unbundling of generation, transmission, and distribution; 2) whether a regulator other than government has been established; 3) whether the wholesale market has been liberalized; 4) privatization. The telecommunication indicator captures 1) the degree of competition in local services; 2) whether a regulator other than government has been established; 3) the degree of liberalization of interconnection charges; 4) privatization.
  - Source: IMF: Giuliano, Mishra and Spilimbergo (2010)
- Domestic Credit
  - Domestic credit to private banks and other financial institutions as a share of GDP, in percent
  - Source: World Bank: World Development Indicators
- Trade Openness
  - Total real trade (exports plus imports) as a share of real GDP (2000 constant price), in percent
  - Source: Penn World Table 7.2
- FDI
  - Foreign direct investment (inflow) as a share of GDP
  - Source: World Bank: World Development Indicators

H3 Policy Variables (contd.)
- Financial reform index
  - The index measures financial liberalization, constructed from 8 underlying dimensions: 1) credit controls and reserve requirements; 2) aggregate credit ceilings; 3) interest rate liberalization; 4) banking sector entry; 5) capital account transactions; 6) privatization; 7) securities market; 8) banking sector supervision. The index is normalized between 0 to 1, higher value indicating more financial liberalization.
  - Source: Abiad, Detragiache and Tressel (2008)
- Capital account openness
  - The index measures a country's degree of capital account openness, based on the binary dummy variables that codify the tabulation of restrictions on cross-border financial transactions reported in the IMF's Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER).
  - Source: Chinn and Ito (2011)
- Labor regulation index
  - Index constructed using principal components analysis on four labor market flexibility indicators: labor tax wedge, the ratio of minimum to mean wage, unemployment benefits coverage, and severance pay after 9 months.
  - Source: Aleksynska and Schindler (2011).
- Polity index
  - Constraint on the executive. Values ranges from 1 (denoting no regulations on the executive’s power) to 7 (denoting countries where political bodies, such as legislatures, have equal or even larger authority than the executive)
  - Source: Polity IV
- Capital stock
  - Natural log of real capital stock constructed using perpetual inventory method. Gupta et al (2011)
- Trade liberalization index
  - Average tariff rates, with missing values extrapolated using implicit weighted tariff rates. Index normalized to be between zero and unity: zero means the tariff rates are 60 percent or higher, while unity means the tariff rates are zero.
  - Source: Pratti et al. (2013)
- Index of economic globalization
  - The index is the economic dimension of the KOF globalization index, constructed on actual economic flows and proxies for restrictions to trade and capital. The index lie between 1 to 100, higher values indicating higher degree of economic globalization.
  - Source: Dreher et al (2008)
- REER
  - Real effective exchange rate (2005=100)
  - Source: IMF: International Financial Statistics

*Content derived from the cited References, Appendix I, and Appendix II of the source PDF.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13176.pdf_