## _wp13187 - 1. Selected Health and Social Indicators

## Source details

**Canonical URL:** [_wp13187 - 1. Selected Health and Social Indicators](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf.json)

---

### Introduction and purpose
- Objective: measure public expenditure inefficiency in the health sector for a sample of 80 emerging and developing economies over the 2001–10 period using a stochastic frontier model (SFA) that controls for socioeconomic determinants of health-sector performance.
- Motivation: previous literature finds significant inefficiencies in health spending (Herrera and Pang, 2005; Gupta and others, 2007; Verhoeven, Gunnarsson, and Carcillo, 2007; Afonso, Schuknecht, and Tanzi, 2010; Joumard, André, and Nicq, 2010; Grigoli and Ley, 2012) but many studies used non-parametric methods that do not control for multiple determinants of health outcomes.
- Key empirical contrast:
  - Bottom quartile could increase life expectancy up to almost five years at current spending levels.
  - A 10 percent increase in public health spending would increase life expectancy by only two months.

### Stylized facts (averages 2001–10)
- Public spending and expenditures:
  - Public spending on health: emerging and developing 3.2 percent of GDP; advanced 6.0 percent.
  - Real public expenditure per capita (PPP, int. dollars): advanced 1,731; emerging and developing 221.
  - Real private expenditure per capita (PPP, int. dollars): advanced 768; emerging and developing 159.
- Health outcomes:
  - HALE (in years at birth): advanced 72.2; emerging and developing 57.2.
  - Life expectancy (in years at birth): advanced 79.1; emerging and developing 65.1.
  - Mortality rate under 5 years (per 1,000 live births): advanced 5.2; emerging and developing 59.5.
  - Infant mortality rate (per 1,000 live births): advanced 3.9; emerging and developing 38.4.
  - Maternal mortality rate (per 100,000 live births): advanced 9.1; emerging and developing 269.3.
  - TB treatment success rate (in percent of cases): advanced 74.7; emerging and developing 77.7.
  - DPT immunization rate (in percent): advanced 94.5; emerging and developing 84.0.
  - Measles immunization rate (in percent): advanced 91.8; emerging and developing 83.2.
  - Polio immunization rate (in percent): advanced 94.3; emerging and developing 84.8.
- Socioeconomic indicators:
  - Real GDP per capita (PPP, int. dollars): advanced 31,086; emerging and developing 7,485.
  - Poverty (below int. dollars 2 per day, percent): advanced 0.8; emerging and developing 27.2.
  - Years of schooling (percent of population over 25): advanced 10.8; emerging and developing 6.7.
  - Gini coefficient: advanced 0.30; emerging and developing 0.43.
  - Alcohol consumption (liters per capita among adults): advanced 9.7; emerging and developing 4.3.
  - Sanitation facilities (percent of population with access): advanced 99.7; emerging and developing 63.8.
  - TB diffusion (per 100,000 people): advanced 14.6; emerging and developing 169.1.
  - HIV diffusion (percent of 15-49 years population): advanced 0.2; emerging and developing 2.6.
  - Adult literacy rate (percent of population over 15): advanced 97.3; emerging and developing 79.8.
  - Water source (percent with access): advanced 99.8; emerging and developing 81.9.
- Governance indicators (WGI, -2.5 to 2.5):
  - Political stability: advanced 0.82; emerging and developing -0.31.
  - Voice and accountability: advanced 1.23; emerging and developing -0.37.
  - Government effectiveness: advanced 1.51; emerging and developing -0.42.
- Statistical relationships:
  - Public health expenditure positively and significantly correlated with HALE (five-year lag structure used).
  - Public health expenditure negatively and significantly correlated with mortality rates.
  - Immunization rates positively and significantly correlated with public health expenditure.
  - TB treatment success/diffusion insignificantly related to public health spending.

### Efficiency measurement: SFA specification and inputs
- Model: averaged cross-section SFA — log(HALE 2006–10) regressed on logs of inputs and covariates averaged 2001–05.
- Inputs/covariates included:
  - Real public expenditure per capita (PPP)
  - Real private expenditure per capita (PPP)
  - Years of schooling (population over 25)
  - Population density per square kilometer of land area
  - Alcohol consumption (liters per adult)
  - Access to sanitation facilities
  - Access to clean water
  - TB diffusion
  - HIV diffusion
- Error term: composite of normally distributed disturbance and a one-sided inefficiency term.
- Inefficiency distributions examined: half-normal and exponential used in sensitivity checks; truncated-normal and gamma did not converge in some specifications.
- Estimation: two-step maximum likelihood with inefficiency estimated as the mean of the conditional distribution.

### Key regression results (selected coefficients)
- Ln public expenditure pc PPP:
  - Column 1: 0.035*** (standard error 0.008)
  - Preferred specification (Column 8, half-normal): 0.025** (standard error 0.012)
  - Exponential inefficiency (Column 9): 0.024** (standard error 0.012)
  - Interpretation: coefficient around 0.03 implies a 10 percent increase in public health spending per capita would raise HALE by only two months (calculation based on an average HALE of 57 years).
- Ln private expenditure pc PPP:
  - Column 1: 0.033*** (standard error 0.008)
  - Preferred specification (Column 8): 0.016 (standard error 0.015)
- Ln years of schooling:
  - Preferred specification (Column 8): 0.050** (standard error 0.024)
- Ln TB diffusion:
  - Preferred specification (Column 8): -0.025** (standard error 0.010)
- Ln HIV diffusion:
  - Preferred specification (Column 8): -0.040*** (standard error 0.006)
- Ln sanitation facility:
  - Significant in some specifications (e.g., Column 3 coefficient 0.076***, standard error 0.020) but becomes insignificant when TB and HIV diffusion are included.

### Efficiency scores and potential gains
- Sample: 80 emerging and developing economies (analysis period 2001–10; SFA dependent HALE 2006–10; inputs 2001–05).
- Highest possible SFA score: 1.0.
- Average SFA efficiency score reported: 0.94 (comparison: Greene (2004) 0.81–0.85; Greene (2005b) 0.87–0.91).
- Regional patterns:
  - African economies have the lowest efficiency on average.
  - Western Hemisphere and Asian economies dominate top positions.
- Selected country SFA efficiency scores (Appendix I point estimates):
  - Papua New Guinea 0.980; Peru 0.979; Vietnam 0.979; Haiti 0.978; Thailand 0.976; Honduras 0.975; Namibia 0.974; Togo 0.974; Panama 0.973; Colombia 0.973; Dominican Republic 0.971; Ecuador 0.969; Paraguay 0.968; Nicaragua 0.967; Cambodia 0.967; Mozambique 0.967; Liberia 0.967; Belize 0.966; Chile 0.966; Nepal 0.963; Guatemala 0.963; Philippines 0.963; Argentina 0.962; Morocco 0.961; Kenya 0.961; Costa Rica 0.961; Uruguay 0.960; Mexico 0.959; India 0.959; Jamaica 0.958; Latvia 0.956; Romania 0.954; Gabon 0.953; Malaysia 0.952; Ukraine 0.952; Tanzania 0.952; Côte d'Ivoire 0.951; El Salvador 0.949; Congo, Republic of 0.948; Mongolia 0.948; Benin 0.948; Sudan 0.946; Gambia, The 0.946; Croatia 0.946; Tunisia 0.946; Mauritius 0.945; Laos 0.945; Botswana 0.944; Malawi 0.944; Algeria 0.943; Turkey 0.943; Bolivia 0.942; Tajikistan 0.942; Sri Lanka 0.940; Bulgaria 0.940; Guyana 0.938; Senegal 0.938; Central African Rep. 0.937; Trinidad and Tobago 0.936; Ghana 0.936; Armenia 0.936; Barbados 0.934; Fiji 0.930; Serbia 0.929; South Africa 0.929; Kyrgyz Republic 0.929; Lithuania 0.924; Burundi 0.923; Qatar 0.921; Cameroon 0.918; Hungary 0.916; Uganda 0.913; Rwanda 0.911; Swaziland 0.903; Egypt 0.902; Kazakhstan 0.899; Mali 0.898; Zambia 0.892; Lesotho 0.892; Sierra Leone 0.810.
- Extreme potential HALE gains from eliminating inefficiency (Figure 2 highlights):
  - Sierra Leone: 8.2 years.
  - Lesotho, Mali, Zambia: just below four years.
  - Barbados: up to 4.7 years.
  - Haiti: 1.2 years.
  - Fiji: 4.7 years.
  - Papua New Guinea: 1.2 years.
  - Egypt: 6.5 years.
  - Morocco: 2.5 years.
  - Hungary: 6.1 years.
  - Latvia: 3 years.
- Quartile means (Table 5):
  - 1st quartile (most efficient): Observations 20; HALE 59.2; Public Exp. Pc PPP 125.3; SFA Efficiency Score 0.97; HALE Increase if moved to frontier 1.7.
  - 2nd quartile: Observations 20; HALE 59.3; Public Exp. Pc PPP 85.2; SFA Efficiency Score 0.95; HALE Increase 2.7.
  - 3rd quartile: Observations 20; HALE 56.3; Public Exp. Pc PPP 82.6; SFA Efficiency Score 0.94; HALE Increase 3.5.
  - 4th quartile (least efficient): Observations 20; HALE 52.2; Public Exp. Pc PPP 78.4; SFA Efficiency Score 0.91; HALE Increase 5.1.

### Comparative magnitude: efficiency gains vs spending increases
- Efficiency effect: bottom quartile could gain up to 5.1 HALE years by eliminating inefficiency.
- Spending effect: a 10 percent increase in public health spending per capita would raise HALE by only two months (based on coefficient ~0.03 and average HALE of 57 years).

### Decomposition of potential gains (average gains from reaching the regional average — Table 4)
- Africa:
  - Public Exp. Pc PPP gain 1.2 HALE years
  - Years of Schooling 0.1
  - TB Diffusion 6.5
  - HIV Diffusion 0.4
  - Efficiency 1.5
- Asia and Pacific:
  - Public Exp. Pc PPP 0.9
  - Years of Schooling 0.1
  - TB Diffusion 2.4
  - HIV Diffusion 0.0
  - Efficiency 1.0
- Europe:
  - Public Exp. Pc PPP 3.0
  - Years of Schooling 0.1
  - TB Diffusion 1.0
  - HIV Diffusion 0.0
  - Efficiency 0.9
- Middle East and Central Asia:
  - Public Exp. Pc PPP 4.1
  - Years of Schooling 0.1
  - TB Diffusion 1.3
  - HIV Diffusion 0.0
  - Efficiency 1.3
- Western Hemisphere:
  - Public Exp. Pc PPP 3.3
  - Years of Schooling 0.1
  - TB Diffusion 1.9
  - HIV Diffusion 0.0
  - Efficiency 0.8
- Interpretation: increasing public health expenditure and reducing TB diffusion generally produce the largest average HALE gains across regions; years of schooling and reductions in HIV diffusion have more modest average effects.

### Country-level potential gains (Appendix II highlights; values in HALE years)
- Selected country entries (table contains many entries; source lists partial values and ellipses where values are omitted):
  - Benin: Public Exp. Pc PPP 1.1; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion 1.2
  - Botswana: Public Exp. Pc PPP ...; Years of Schooling 1.0; TB Diffusion 4.3; HIV Diffusion 0.7; Efficiency 1.1
  - Burundi: Public Exp. Pc PPP 1.5; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion 0.4; Total 2.1
  - Lesotho: Public Exp. Pc PPP 0.8; Years of Schooling 5.7; TB Diffusion 0.6; HIV Diffusion 1.8; Efficiency 8.9
  - Mozambique: Public Exp. Pc PPP 1.2; Years of Schooling 0.2; TB Diffusion 3.1; HIV Diffusion 0.1; Efficiency 4.7
  - Namibia: Public Exp. Pc PPP ..; Years of Schooling 17.3; TB Diffusion 0.3; HIV Diffusion 1.7; Efficiency 7.6
  - Sierra Leone: Public Exp. Pc PPP 1.6; Years of Schooling 0.1; TB Diffusion 1.5; HIV Diffusion 5.3; Efficiency 8.4
  - South Africa: Public Exp. Pc PPP ..; Years of Schooling 1.0; TB Diffusion 0.9; HIV Diffusion 0.4; Efficiency 11.5
  - Swaziland: Public Exp. Pc PPP ..; Years of Schooling 16.4; TB Diffusion 0.7; HIV Diffusion 1.4; Efficiency 18.5
  - Papua New Guinea: Public Exp. Pc PPP 0.1; Years of Schooling 0.1; TB Diffusion 2.8; HIV Diffusion 0.0; Efficiency 0.3; Total 3.0
  - Cambodia: Public Exp. Pc PPP 1.1; Years of Schooling 0.0; TB Diffusion 7.0; HIV Diffusion 0.0; Efficiency 0.8; Total 2.2
  - Haiti: Public Exp. Pc PPP 5.9; Years of Schooling 0.2; TB Diffusion 5.2; HIV Diffusion 0.1; Efficiency 11.3
  - Peru: Public Exp. Pc PPP 2.4; Years of Schooling 0.2; TB Diffusion 0.2; HIV Diffusion ...; Total 4.6
  - Belize: Public Exp. Pc PPP 3.1; Years of Schooling ..; TB Diffusion 0.1; HIV Diffusion 0.3; Efficiency 3.2
  - Bolivia: Public Exp. Pc PPP 2.9; Years of Schooling 0.2; TB Diffusion 0.5; HIV Diffusion 0.1; Efficiency 2.6; Total 6.6
  - Guyana: Public Exp. Pc PPP 3.0; Years of Schooling 0.1; TB Diffusion 1.1; HIV Diffusion 0.0; Efficiency 1.3; Total 5.5
  - Other country entries include ellipses where the source omits values.
- Note on calculation: potential gains from each variable are calculated by multiplying the SFA coefficient by the increase needed to reach the regional average. Above regional average observations are excluded from the potential gain calculations.

### Robustness checks and sensitivity
- Lag structures: three-, four-, and five-year lag specifications show Spearman and Kendall rank correlations of 0.99*** and 0.99*** with preferred specification.
- Alternative dependent variables: mortality rate under 5 years and infant mortality rate yield Spearman correlations of 0.78*** and 0.77*** with preferred SFA ranking.
- Alternative inefficiency distributions: exponential inefficiency yields Spearman 0.99***.
- DEA comparisons:
  - DEA (one input: public health expenditure per capita PPP; one output: HALE): Spearman correlation with SFA scores 0.46*** (Kendall 0.33***).
  - DEA with two inputs (public health expenditure per capita PPP and real GDP per capita PPP) and output HALE: Spearman 0.48***.
  - DEA with public health expenditure per capita PPP plus principal components from SFA regressors: Spearman 0.46***.
- Conclusion: rankings robust to lag structure, measurement in percent of GDP vs per capita PPP, and inefficiency distribution assumptions; correlations with DEA are significant but modest.

### Caveats and limitations
- Data limitations: HALE time series limited (only observation for HALE over 2006–10 is 2007); health outputs/outcomes data constrain analysis.
- Efficiency interpretation: a score of 0.95 is relative to the estimated frontier and does not imply only a 5 percent feasible improvement in absolute terms.
- Possible omitted variable bias: residual omitted variables could affect inefficiency estimates.
- Second-stage analysis: because SFA controls for exogenous factors, efficiency scores are likely uncorrelated with SFA regressors; governance variables did not show systematic relationships with efficiency in second-step tests.

### Policy implications and recommendations
- Priority: improving the efficiency of public health spending should be a core element of reform strategies in emerging and developing economies given larger HALE gains from efficiency improvements than from comparable increases in spending.
- Composition matters: shifting public health outlays toward efforts that reduce TB diffusion is associated with higher efficiency and should be prioritized.
- Complementary policies: improving educational attainment is associated with higher HALE for the same level of spending, implying cross-sectoral investments (education and water/sanitation) can improve health spending efficiency.
- Targeted reforms: country-specific potential gains (Appendix II) can guide prioritization where efficiency gains would translate into the largest HALE improvements.

*Source: _wp13187 - 1. Selected Health and Social Indicators (PDF); canonical URL https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf*

### 1. Selected Health and Social Indicators ...............................................................................

### _wp13187 - 1. Selected Health and Social Indicators ...............................................................................

### Major Sections
- 1. Selected Health and Social Indicators ....................................................................................8
- 2. Multicollinearity Diagnostics...............................................................................................13
- 3. SFA Regressions ..................................................................................................................14
- 4. Average Potential Gain from Reaching the Regional Average ...........................................17
- 5. Means by Quartile of Technical Efficiency .........................................................................17
- 6. Robustness Checks...............................................................................................................19

### Figures
- Figure 1. Public Health Expenditure and Outputs/Outcomes ...............................................................9
- Figure 2. Potential Gains from Eliminating Inefficiency ....................................................................16

### Appendices
- Appendix I. SFA Efficiency Scores..........................................................................................................21
- Appendix II. Potential Gains from Reaching the Regional Average ........................................................22

### Appendix Tables
- Appendix Table 1. Point Estimates.....................................................................................................................21
- Appendix Table 2. Country Estimates ................................................................................................................22

*Source: _wp13187 - 1. Selected Health and Social Indicators (PDF).*

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

### _wp13187 - References

### Introduction and purpose
- Objective: measure public expenditure inefficiency in the health sector for a sample of 80 emerging and developing economies over the 2001–10 period using a stochastic frontier model (SFA) that controls for socioeconomic determinants of health-sector performance.
- Motivation: previous literature finds significant inefficiencies in health spending (Herrera and Pang, 2005; Gupta and others, 2007; Verhoeven, Gunnarsson, and Carcillo, 2007; Afonso, Schuknecht, and Tanzi, 2010; Joumard, André, and Nicq, 2010; Grigoli and Ley, 2012) but many studies used non-parametric methods that do not control for multiple determinants of health outcomes.
- Key empirical contrast: the bottom quartile could increase life expectancy up to almost five years at current spending levels, whereas a 10 percent increase in public health spending would increase life expectancy by only two months.

### Literature and methodological context
- Non-parametric methods reviewed: FDH and DEA. Findings in cited studies: African economies often found inefficient; inefficiency positively correlated with level of government expenditure in some studies.
- Parametric methods reviewed: SFA and related econometric approaches (Evans and others, 2000; Jayasuriya and Wodon, 2003; Greene, 2004/2005). Advantages of SFA emphasized:
  - Can accommodate randomness and measurement error and separate them from inefficiency.
  - Can statistically control for many covariates that influence health outcomes.
- Caveats: SFA requires assumptions on functional form and on the distribution of the inefficiency term (half-normal, exponential, truncated-normal, gamma). No a priori reason to prefer a distribution; sensitivity checks are necessary.

### Stylized facts (averages 2001–10)
- Public spending and outputs:
  - Public spending on health averages 3.2 percent of GDP in emerging and developing economies versus 6.0 percent in advanced economies.
  - Real public expenditure per capita (PPP, int. dollars): advanced 1,731; emerging and developing 221.
  - Real private expenditure per capita (PPP, int. dollars): advanced 768; emerging and developing 159.
- Health outcomes:
  - HALE (in years at birth): advanced 72.2; emerging and developing 57.2.
  - Life expectancy (in years at birth): advanced 79.1; emerging and developing 65.1.
  - Mortality rate under 5 years (per 1,000 live births): advanced 5.2; emerging and developing 59.5.
  - Infant mortality rate (per 1,000 live births): advanced 3.9; emerging and developing 38.4.
  - Maternal mortality rate (per 100,000 live births): advanced 9.1; emerging and developing 269.3.
  - TB treatment success rate (in percent of cases): advanced 74.7; emerging and developing 77.7.
  - DPT immunization rate (in percent): advanced 94.5; emerging and developing 84.0.
  - Measles immunization rate (in percent): advanced 91.8; emerging and developing 83.2.
  - Polio immunization rate (in percent): advanced 94.3; emerging and developing 84.8.
- Socioeconomic indicators:
  - Real GDP per capita (PPP, int. dollars): advanced 31,086; emerging and developing 7,485.
  - Poverty (below int. dollars 2 per day, percent): advanced 0.8; emerging and developing 27.2.
  - Years of schooling (percent of population over 25): advanced 10.8; emerging and developing 6.7.
  - Gini coefficient: advanced 0.30; emerging and developing 0.43.
  - Alcohol consumption (liters per capita among adults): advanced 9.7; emerging and developing 4.3.
  - Sanitation facilities (percent of population with access): advanced 99.7; emerging and developing 63.8.
  - TB diffusion (per 100,000 people): advanced 14.6; emerging and developing 169.1.
  - HIV diffusion (percent of 15-49 years population): advanced 0.2; emerging and developing 2.6.
  - Adult literacy rate (percent of population over 15): advanced 97.3; emerging and developing 79.8.
  - Water source (percent with access): advanced 99.8; emerging and developing 81.9.
- Governance indicators (WGI, -2.5 to 2.5):
  - Political stability: advanced 0.82; emerging and developing -0.31.
  - Voice and accountability: advanced 1.23; emerging and developing -0.37.
  - Government effectiveness: advanced 1.51; emerging and developing -0.42.
- Statistical relationships noted:
  - Positive and significant correlation between public health expenditure and HALE (five-year lag structure used).
  - Negative and significant correlation between public health expenditure and mortality rates.
  - Immunization rates positively and significantly correlated with public health expenditure.
  - Insignificant relationship between TB treatment success/diffusion and public health spending, possibly due to disease-specific dynamics.

### Efficiency measurement: SFA specification and inputs
- Model (averaged cross-section SFA): log(HALE 2006–10) regressed on logs of inputs and covariates averaged 2001–05. Inputs/covariates include:
  - Real public expenditure per capita (PPP)
  - Real private expenditure per capita (PPP)
  - Years of schooling (population over 25)
  - Population density per square kilometer of land area
  - Alcohol consumption (liters per adult)
  - Access to sanitation facilities
  - Access to clean water
  - TB diffusion
  - HIV diffusion
- Error term: composite of normally distributed disturbance and a one-sided inefficiency term. Sensitivity checks use half-normal and exponential inefficiency distributions; truncated-normal and gamma did not converge in some specifications.
- Estimation: two-step maximum likelihood with inefficiency estimated as the mean of the conditional distribution.

### Key regression results (selected coefficients from Table 3)
- Ln public expenditure pc PPP:
  - Column 1: 0.035*** (standard error 0.008)
  - Preferred specification (Column 8, half-normal): 0.025** (standard error 0.012)
  - Exponential inefficiency (Column 9): 0.024** (standard error 0.012)
  - Interpretation: coefficient around 0.03 implies a 10 percent increase in public health spending per capita would raise HALE by only two months (calculation based on an average HALE of 57 years).
- Ln private expenditure pc PPP:
  - Column 1: 0.033*** (standard error 0.008)
  - Preferred specification (Column 8): 0.016 (standard error 0.015)
- Ln years of schooling:
  - Preferred specification (Column 8): 0.050** (standard error 0.024)
- Ln TB diffusion:
  - Preferred specification (Column 8): -0.025** (standard error 0.010)
- Ln HIV diffusion:
  - Preferred specification (Column 8): -0.040*** (standard error 0.006)
- Other relevant results:
  - Ln sanitation facility significant in some specifications (e.g., Column 3 coefficient 0.076***, standard error 0.020) but becomes insignificant when TB and HIV diffusion are included.

### Efficiency scores and potential gains
- Sample and scoring:
  - Sample: 80 emerging and developing economies (analysis period 2001–10; SFA dependent HALE 2006–10; inputs 2001–05).
  - Highest possible SFA score: 1.0.
  - Average SFA efficiency score reported: 0.94 (comparison: Greene (2004) 0.81–0.85; Greene (2005b) 0.87–0.91).
- Regional patterns:
  - African economies have the lowest efficiency on average.
  - Western Hemisphere and Asian economies dominate top positions.
  - Specific extremes for potential HALE gains from eliminating inefficiency (Figure 2 highlights):
    - Sierra Leone: 8.2 years.
    - Lesotho, Mali, Zambia: just below four years.
    - Barbados: up to 4.7 years.
    - Haiti: 1.2 years.
    - Fiji: 4.7 years.
    - Papua New Guinea: 1.2 years.
    - Egypt: 6.5 years.
    - Morocco: 2.5 years.
    - Hungary: 6.1 years.
    - Latvia: 3 years.
- Quartile means (Table 5):
  - 1st quartile (most efficient): Observations 20; HALE 59.2; Public Exp. Pc PPP 125.3; SFA Efficiency Score 0.97; HALE Increase if moved to frontier 1.7.
  - 2nd quartile: Observations 20; HALE 59.3; Public Exp. Pc PPP 85.2; SFA Efficiency Score 0.95; HALE Increase 2.7.
  - 3rd quartile: Observations 20; HALE 56.3; Public Exp. Pc PPP 82.6; SFA Efficiency Score 0.94; HALE Increase 3.5.
  - 4th quartile (least efficient): Observations 20; HALE 52.2; Public Exp. Pc PPP 78.4; SFA Efficiency Score 0.91; HALE Increase 5.1.

### Comparative magnitude: efficiency gains vs spending increases
- Efficiency effect: bottom quartile could gain up to 5.1 HALE years by eliminating inefficiency.
- Spending effect: a 10 percent increase in public health spending per capita would raise HALE by only two months (based on coefficient ~0.03 and average HALE of 57 years).

### Decomposition of potential gains (Table 4, average potential gains from reaching regional average)
- Potential HALE gains from bringing countries below regional mean up to the regional average (selected regional averages reported):
  - Africa: Public Exp. Pc PPP gain 1.2 HALE years; Years of Schooling 0.1; TB Diffusion 6.5; HIV Diffusion 0.4; Efficiency 1.5.
  - Asia and Pacific: Public Exp. Pc PPP 0.9; Years of Schooling 0.1; TB Diffusion 2.4; HIV Diffusion 0.0; Efficiency 1.0.
  - Europe: Public Exp. Pc PPP 3.0; Years of Schooling 0.1; TB Diffusion 1.0; HIV Diffusion 0.0; Efficiency 0.9.
  - Middle East and Central Asia: Public Exp. Pc PPP 4.1; Years of Schooling 0.1; TB Diffusion 1.3; HIV Diffusion 0.0; Efficiency 1.3.
  - Western Hemisphere: Public Exp. Pc PPP 3.3; Years of Schooling 0.1; TB Diffusion 1.9; HIV Diffusion 0.0; Efficiency 0.8.
- Interpretation: increasing public health expenditure and reducing TB diffusion generally produce the largest average HALE gains across regions; years of schooling and reductions in HIV diffusion have more modest average effects.

### Robustness checks and sensitivity
- Lag structures: three-, four-, and five-year lag specifications show Spearman and Kendall rank correlations of 0.99*** and 0.99*** with preferred specification, indicating high robustness.
- Alternative dependent variables: using mortality rate under 5 years and infant mortality rate yields Spearman correlations of 0.78*** and 0.77*** with preferred SFA ranking.
- Alternative inefficiency distributions: exponential inefficiency yields Spearman 0.99***.
- DEA comparisons:
  - DEA with one input (public health expenditure per capita PPP) and one output (HALE): Spearman correlation with SFA scores 0.46*** (Kendall 0.33***).
  - DEA with two inputs (public health expenditure per capita PPP and real GDP per capita PPP) and output HALE: Spearman 0.48***.
  - DEA with public health expenditure per capita PPP plus principal components from SFA regressors: Spearman 0.46***.
- Conclusion from robustness: rankings are robust to lag structure, measurement in percent of GDP vs per capita PPP, and inefficiency distribution assumptions; correlations with DEA are significant but modest, reflecting methodological differences.

### Caveats and limitations
- Data availability limits: health outputs/outcomes data constrain analysis; HALE time series limited (only observation for HALE over 2006–10 is 2007).
- Efficiency interpretation: efficiency is measured relative to estimated frontier; a score of 0.95 does not mean only a 5 percent feasible improvement in absolute terms—scores depend on the sample and frontier.
- Possible omitted variable bias: specification includes many controls but residual omitted variables could affect inefficiency estimates.
- Second-stage analysis: because SFA controls for exogenous factors, efficiency scores are likely uncorrelated with SFA regressors; governance variables (political stability, voice and accountability, government effectiveness) did not show systematic relationships with efficiency in second-step tests.

### Policy implications and recommendations
- Priority: improving the efficiency of public health spending should be a core element of reform strategies in emerging and developing economies given the potentially larger HALE gains from efficiency improvements than from comparable increases in spending.
- Composition matters: shifting public health outlays toward efforts that reduce TB diffusion is associated with higher efficiency and should be prioritized.
- Complementary policies: improving educational attainment is associated with higher HALE for the same level of spending, implying that cross-sectoral investments (education and water/sanitation) can improve health spending efficiency.
- Targeted reforms: country-specific potential gains (Appendix II) can guide prioritization where efficiency gains would translate into the largest HALE improvements.

*Source: _wp13187 - References; canonical URL https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf*

### Appendix I. SFA Efficiency Scores

### Appendix I. SFA Efficiency Scores

### Point estimates — SFA Efficiency Score by country
- Papua New Guinea 0.980
- Peru 0.979
- Vietnam 0.979
- Haiti 0.978
- Thailand 0.976
- Honduras 0.975
- Namibia 0.974
- Togo 0.974
- Panama 0.973
- Colombia 0.973
- Dominican Republic 0.971
- Ecuador 0.969
- Paraguay 0.968
- Nicaragua 0.967
- Cambodia 0.967
- Mozambique 0.967
- Liberia 0.967
- Belize 0.966
- Chile 0.966
- Nepal 0.963
- Guatemala 0.963
- Philippines 0.963
- Argentina 0.962
- Morocco 0.961
- Kenya 0.961
- Costa Rica 0.961
- Uruguay 0.960
- Mexico 0.959
- India 0.959
- Jamaica 0.958
- Latvia 0.956
- Romania 0.954
- Gabon 0.953
- Malaysia 0.952
- Ukraine 0.952
- Tanzania 0.952
- Côte d'Ivoire 0.951
- El Salvador 0.949
- Congo, Republic of 0.948
- Mongolia 0.948
- Benin 0.948
- Sudan 0.946
- Gambia, The 0.946
- Croatia 0.946
- Tunisia 0.946
- Mauritius 0.945
- Laos 0.945
- Botswana 0.944
- Malawi 0.944
- Algeria 0.943
- Turkey 0.943
- Bolivia 0.942
- Tajikistan 0.942
- Sri Lanka 0.940
- Bulgaria 0.940
- Guyana 0.938
- Senegal 0.938
- Central African Rep. 0.937
- Trinidad and Tobago 0.936
- Ghana 0.936
- Armenia 0.936
- Barbados 0.934
- Fiji 0.930
- Serbia 0.929
- South Africa 0.929
- Kyrgyz Republic 0.929
- Lithuania 0.924
- Burundi 0.923
- Qatar 0.921
- Cameroon 0.918
- Hungary 0.916
- Uganda 0.913
- Rwanda 0.911
- Swaziland 0.903
- Egypt 0.902
- Kazakhstan 0.899
- Mali 0.898
- Zambia 0.892
- Lesotho 0.892
- Sierra Leone 0.810

### Appendix II. Potential Gains from Reaching the Regional Average

### Note on calculation
- Potential gains from each variable are calculated by multiplying the SFA coefficient by the increase needed to reach the regional average. Above regional average observations are excluded from the potential gain calculations.

### Country estimates (In HALE years) — variables: Public Exp. Pc PPP; Years of Schooling; TB Diffusion; HIV Diffusion; Efficiency; Total
- Benin: Public Exp. Pc PPP 1.1; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion 1.2
- Botswana: Public Exp. Pc PPP ...; Years of Schooling 1.0; TB Diffusion 4.3; HIV Diffusion 0.7; Efficiency 1.1
- Burundi: Public Exp. Pc PPP 1.5; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion 0.4; Total 2.1
- Cameroon: Public Exp. Pc PPP 1.2; Years of Schooling ...; TB Diffusion ...; HIV Diffusion 0.7; Total 1.9
- Central African Rep.: Public Exp. Pc PPP 1.6; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion ...; Total 1.6
- Congo, Republic of: Public Exp. Pc PPP 0.8; Years of Schooling 0.0; TB Diffusion 0.5; HIV Diffusion ...; Total 1.3
- Côte d'Ivoire: Public Exp. Pc PPP 1.4; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion ...; Total 1.5
- Gabon: Public Exp. Pc PPP .....; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion ...; Total 0.0
- Gambia, The: Public Exp. Pc PPP 0.7; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion ...; Total 0.8
- Ghana: Public Exp. Pc PPP 1.1; Years of Schooling ....; TB Diffusion ...; HIV Diffusion ...; Total 1.1
- Kenya: Public Exp. Pc PPP 1.3; Years of Schooling ....; TB Diffusion ...; HIV Diffusion ...; Total 1.3
- Lesotho: Public Exp. Pc PPP 0.8; Years of Schooling 5.7; TB Diffusion 0.6; HIV Diffusion 1.8; Efficiency 8.9
- Liberia: Public Exp. Pc PPP 1.7; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion ...; Total 1.8
- Malawi: Public Exp. Pc PPP 1.2; Years of Schooling 0.1; TB Diffusion 2.2; HIV Diffusion 0.2; Efficiency 1.6
- Mali: Public Exp. Pc PPP 1.3; Years of Schooling 0.2; TB Diffusion ...; HIV Diffusion 1.6; Total 3.0
- Mauritius: Public Exp. Pc PPP .....; Years of Schooling ...; TB Diffusion ...; HIV Diffusion ...; Total 0.0
- Mozambique: Public Exp. Pc PPP 1.2; Years of Schooling 0.2; TB Diffusion 3.1; HIV Diffusion 0.1; Efficiency 4.7
- Namibia: Public Exp. Pc PPP ..; Years of Schooling 17.3; TB Diffusion 0.3; HIV Diffusion 1.7; Efficiency 7.6
- Rwanda: Public Exp. Pc PPP 1.2; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion ...; Total 2.3
- Senegal: Public Exp. Pc PPP 1.0; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion ...; Total 1.0
- Sierra Leone: Public Exp. Pc PPP 1.6; Years of Schooling 0.1; TB Diffusion 1.5; HIV Diffusion 5.3; Efficiency 8.4
- South Africa: Public Exp. Pc PPP ..; Years of Schooling 1.0; TB Diffusion 0.9; HIV Diffusion 0.4; Efficiency 11.5
- Swaziland: Public Exp. Pc PPP ..; Years of Schooling 16.4; TB Diffusion 0.7; HIV Diffusion 1.4; Efficiency 18.5
- Tanzania: Public Exp. Pc PPP 1.4; Years of Schooling ....; TB Diffusion ...; HIV Diffusion ...; Total 1.4
- Togo: Public Exp. Pc PPP 1.6; Years of Schooling 0.0; TB Diffusion 0.1; HIV Diffusion ...; Total 1.7
- Uganda: Public Exp. Pc PPP 1.4; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion 0.9; Efficiency 2.3
- Zambia: Public Exp. Pc PPP 0.8; Years of Schooling 5.7; TB Diffusion 0.2; HIV Diffusion 1.8; Efficiency 8.5
- Chile: Public Exp. Pc PPP .....; Years of Schooling ...; TB Diffusion ...; HIV Diffusion ...; Total 0.0
- Colombia: Public Exp. Pc PPP .0; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion ...; Total 0.0
- Cambodia: Public Exp. Pc PPP 1.1; Years of Schooling 0.0; TB Diffusion 7.0; HIV Diffusion 0.0; Efficiency 0.8; Total 2.2
- Fiji: Public Exp. Pc PPP ....; Years of Schooling 1.9; TB Diffusion 1.9; HIV Diffusion ...; Total 1.9
- India: Public Exp. Pc PPP 1.2; Years of Schooling 0.1; TB Diffusion 0.5; HIV Diffusion ...; Total 1.8
- Laos: Public Exp. Pc PPP 1.3; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion 0.8; Efficiency 2.2
- Malaysia: Public Exp. Pc PPP ....; Years of Schooling 0.4; TB Diffusion 0.4; HIV Diffusion ...; Total 0.4
- Mongolia: Public Exp. Pc PPP ..; Years of Schooling 1.0; TB Diffusion 0.0; HIV Diffusion 0.7; Efficiency 1.7
- Nepal: Public Exp. Pc PPP 1.4; Years of Schooling 0.2; TB Diffusion 0.0; HIV Diffusion 0.0; Efficiency 1.6
- Papua New Guinea: Public Exp. Pc PPP 0.1; Years of Schooling 0.1; TB Diffusion 2.8; HIV Diffusion 0.0; Efficiency 0.3; Total 3.0
- Philippines: Public Exp. Pc PPP 0.8; Years of Schooling 0.3; TB Diffusion 0.0; HIV Diffusion ...; Total 3.8
- Sri Lanka: Public Exp. Pc PPP 0.3; Years of Schooling ...; TB Diffusion ...; HIV Diffusion 1.3; Efficiency 1.6
- Thailand: Public Exp. Pc PPP .0; Years of Schooling 0.0; TB Diffusion 0.0; HIV Diffusion 0.0; Efficiency 0.1
- Vietnam: Public Exp. Pc PPP 1.0; Years of Schooling 0.1; TB Diffusion 0.3; HIV Diffusion ...; Efficiency 1.3
- Paraguay: Public Exp. Pc PPP 3.6; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion ...; Total 3.7
- Peru: Public Exp. Pc PPP 2.4; Years of Schooling 0.2; TB Diffusion 0.2; HIV Diffusion ...; Total 4.6
- Trinidad and Tobago: Public Exp. Pc PPP ...; Years of Schooling 0.0; TB Diffusion 1.7; HIV Diffusion 1.7; Total 1.7
- Uruguay: Public Exp. Pc PPP ....; Years of Schooling 0.1; TB Diffusion 0.1; HIV Diffusion ...; Total 0.1
- Argentina: Public Exp. Pc PPP .....; Years of Schooling ...; TB Diffusion ...; HIV Diffusion ...; Total 0.0
- Barbados: Public Exp. Pc PPP ....; Years of Schooling ...; TB Diffusion 2.0; HIV Diffusion 2.0; Total 2.0
- Belize: Public Exp. Pc PPP 3.1; Years of Schooling ..; TB Diffusion 0.1; HIV Diffusion 0.3; Efficiency 3.2
- Bolivia: Public Exp. Pc PPP 2.9; Years of Schooling 0.2; TB Diffusion 0.5; HIV Diffusion 0.1; Efficiency 2.6; Total 6.6
- Costa Rica: Public Exp. Pc PPP ....; Years of Schooling 0.0; TB Diffusion 0.0; HIV Diffusion 0.0; Total 0.0
- Dominican Republic: Public Exp. Pc PPP 3.5; Years of Schooling 0.0; TB Diffusion 0.0; HIV Diffusion 0.5; Efficiency 0.0; Total 4.1
- Ecuador: Public Exp. Pc PPP 3.4; Years of Schooling 0.0; TB Diffusion 0.0; HIV Diffusion 0.6; Efficiency ...; Total 4.0
- El Salvador: Public Exp. Pc PPP 1.5; Years of Schooling 0.0; TB Diffusion ...; HIV Diffusion 0.8; Total 2.3
- Guatemala: Public Exp. Pc PPP 4.0; Years of Schooling 0.0; TB Diffusion 0.2; HIV Diffusion ...; Total 4.2
- Guyana: Public Exp. Pc PPP 3.0; Years of Schooling 0.1; TB Diffusion 1.1; HIV Diffusion 0.0; Efficiency 1.3; Total 5.5
- Haiti: Public Exp. Pc PPP 5.9; Years of Schooling 0.2; TB Diffusion 5.2; HIV Diffusion 0.1; Efficiency 11.3
- Honduras: Public Exp. Pc PPP 3.4; Years of Schooling 0.1; TB Diffusion 0.8; HIV Diffusion 0.0; Total 4.4
- Jamaica: Public Exp. Pc PPP 2.1; Years of Schooling ..; TB Diffusion 0.0; HIV Diffusion 0.0; Efficiency 2.3
- Mexico: Public Exp. Pc PPP ....; Years of Schooling 0.2; TB Diffusion 0.2; HIV Diffusion ...; Total 0.2
- Nicaragua: Public Exp. Pc PPP 3.9; Years of Schooling 0.1; TB Diffusion ...; HIV Diffusion ...; Total 4.0
- Panama: Public Exp. Pc PPP ...; Years of Schooling 0.0; TB Diffusion 0.0; HIV Diffusion 0.0; Total 0.0
- Barbados: (see above) ....2.0
- Europe and Africa; Asia and Pacific; Middle East and Central Asia; Western Hemisphere — region headings included in source.

*Source: Authors’ calculations.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13187.pdf_