## _wp14148

## Source details

**Canonical URL:** [_wp14148](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14148.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14148.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14148.pdf.json)

---

### Research question and scope
- Empirical examination of whether public investment drives (investment booms) over the past four decades triggered higher economic growth, productivity and long term development.
- Compares growth after episodes of large public investment rises with growth before those episodes in the same countries, and with countries that never had such episodes.
- Considers alternative definitions and measures of booms, World Bank project evidence, and five country case studies (Bolivia, Mexico, Philippines, Korea, Taiwan province of China).
- Sample: 124 lower and middle income countries with WEO data for at least one year during 1960-2011; average of 30 observations per country.

### Key data and measurement features
- Two primary public investment measures used:
  - public investment ratio (I/Y), preferred for timing of booms because it is a direct measure of spending effort; and
  - growth of the public capital stock (constructed from investment flows).
- Public capital stock construction begins with an initial capital stock chosen to avoid implausible discrepancies between capital growth and the investment ratio.
- Boom-dating (preferred method):
  - compute unweighted 5-year forward and backward moving averages of I/Y;
  - identify years where the forward-backward difference exceeded the 80th percentile for three or more years and exceeded 4 percentage points of GDP;
  - start = year of first substantial rise; end = year when I/Y fell back to at least 120 percent of its pre-boom level (if never fell back, classified as continued long boom);
  - refined by graph inspection and corroborated with capital growth series.
- Alternative (robustness) method:
  - boom = investment ratio above the 70th percentile for three or more years; boom dummy = 1 during those years.
- Final sample details:
  - Of 3,591 data points, 335 are rated as boom years.
  - Boom variable (I/Y – I/Y(0)): median 4.90; maximum 20.03.
  - Boom variable based on public capital growth: median 4.36; maximum 27.34.
  - 21 countries determined to have experienced booms according to the preferred method.

### Conceptual framework and identification logic
- Production function with public capital stock as input: Y = A Z^{εψ} K^{β} L^{α}.
- Public capital accumulation includes inefficiency parameter s (0<s<1) and effective elasticity reduced by ε (0<ε<1).
- Identification strategies:
  1. Treat very large, sustained increases in I/Y as deliberate policy shifts (average difference across all 24 episodes between minimum and maximum investment ratio was just over 13 percentage points of GDP; mean duration just over 14 years).
  2. Use timing/lags: production-function effects of large public investments should appear with multi-year lags while contemporaneous I/Y–GDP associations are more consistent with reverse causality.
- Key empirical prediction: production-function-driven growth should show little contemporaneous I/Y–GDP association but positive lagged effects.

### Principal empirical findings — summary
- Econometric evidence:
  - small positive and instantaneous (contemporaneous) associations between public investment booms and economic growth;
  - little or no long-run impact of booms on growth.
- Diagnostics:
  - Most positive association occurs immediately; F-tests fail to reject that the long run impact (sum of coefficients on lagged investment booms) is zero.
  - Estimated associations are small and not of the magnitudes suggested by big push models.
- Country patterns:
  - Clear-cut examples of booms followed by sustained acceleration in GDP growth are difficult to find; several booms were followed by slumps.
  - Ethiopia may be an exception in recent data but causality is difficult to establish.
  - Korea and Taiwan: rapid growth began before their public investment drives; their public investment drives were modest relative to economy size.
  - China and Vietnam growth revivals were driven by abolition of price controls in Agriculture, not major public investment drives.
- World Bank project evidence and case studies highlight pervasive problems: incentive and agency problems, weak project selection, and fragmented micro-data.

### Short-run empirical results (exact reported coefficients and diagnostics)
- Estimated contemporaneous association (Table 2, Boom VarI No Lag): Boom VarI = 0.14** (t = (2.29)).
- Short-run lag pattern (Table 2 examples):
  - Boom VarI (-3) = -0.05 (−0.80).
  - Boom VarI (-4) = -0.12** (−2.02).
- Regression fit: R-squared ≈ 0.14 (R-squared = 0.14, 0.14, 0.14, 0.15, 0.17 across columns).
- Control variables (examples from Tables 2 and 3):
  - Black Market Exchange Rate Premium: negative and highly significant (e.g., -0.27*** with t = (-4.08)).
  - Life Expectancy at birth (years): 0.03** (t ≈ 2.33–2.43).
  - Financial Depth (Deposit Bank assets % of GDP): negative (e.g., -0.01* with t ≈ (-1.92)).
  - Inflation: -0.00** (t ≈ -2.33).
  - Real Export growth: 0.09***–0.10*** (t ≈ 11.22–11.58).
- Alternative measure: public capital stock growth (Table 3) strengthens short-run coefficients:
  - Boom VarK No Lag: 0.24*** (t = (5.01)).
  - Boom VarK (-1): 0.16*** (t = (3.10)).
  - Boom VarK (-2): 0.10* (t = (1.71)).
  - Longer lags not significant.
- Robustness checks:
  - Excluding natural resource intensive countries left contemporaneous coefficient virtually unchanged.
  - Using only positive values of the boom variable increased estimated effect to 0.38 (se=0.058) for the strongest specification.

### Long-run impact tests (preserving reported sums and p-values)
- Procedure: regressions with contemporaneous and five lags; long-run impact = sum of coefficients.
- Using I/Y to measure booms (Table 4): estimated long-run impact across five specifications = -0.03, -0.02, -0.01, 0.00, -0.03 (average -0.02); average p-value = 0.71. Individual F-stat/p-values for null that sum=0: 0.33 (p=0.56), 0.15 (p=0.70), 0.08 (p=0.78), 0.00 (p=0.96), 0.39 (p=0.56).
- Using Public Capital Growth to measure booms (Table 5): estimated long-run impact across five specifications = 0.10, 0.10, 0.12, 0.10, 0.07 (average 0.10); corresponding F-stat for null that sum=0 = 1.65, 2.47, 2.50, 3.03, 1.92 with p-values 0.20, 0.12, 0.11, 0.08, 0.17.
- Conclusion: no robust evidence that investment booms exerted a positive long-term impact on the level of GDP even under favorable causality assumptions.

### Post-1990 sample results and sensitivity (exact coefficients)
- Restricting sample to post-1990 (Table 6) — Boom VarI results:
  - Boom VarI No Lag: 0.31*** (t=3.14).
  - Boom VarI (-1): 0.33*** (t=3.42).
  - Boom VarI (-2): 0.34*** (t=3.76).
  - Boom VarI (-3): 0.30*** (t=3.17).
  - Boom VarI (-4): 0.24** (t=2.40).
  - Observations: 1,180; 1,174; 1,169; 1,161; 1,151. R-squared = 0.15–0.16.
- Using public capital stock growth in post-1990 (Table 7):
  - Boom VarK No Lag: 0.31*** (t=2.76).
  - Boom VarK (-4): 0.17** (t=2.07); other lags weaker.
  - Observations: 1,171; 1,164; 1,156; 1,146; 1,135. R-squared = 0.15–0.16.
- Sensitivity:
  - Positive post-1990 result in Table 6 is driven by Ethiopian observations; removing Ethiopia changes coefficient from .30778141 (se = .09798922, t = 3.14) to .20123576 (se = .11155869, t = 1.8).
  - Uganda and Panama create sensitivity depending on measurement and dating choices.
- Interpretation: post-1990 evidence base is small and empirically fragile.

### Robustness of boom-year identification (probabilities and indicator regressions)
- Probability statistics:
  - Probability an observation is classified as a boom in Set 2 given Set 1 = 41 percent.
  - Probability for Set 3 given Set 1 = 49 percent.
  - Probability of Set 3 given Set 2 = 24 percent.
- Indicator regressions (Appendix I) — selected exact coefficients:
  - Table IA (full sample, BoomI indicator): BoomI (No Lag) = 0.69 (1.45); BoomI (-4) = -0.79* (-1.68); Observations: 3,143 down to 3,109; R-squared = 0.15.
  - Table IB (full sample, BoomK indicator): BoomK (No Lag) = 1.25*** (2.77); BoomK (-1) = 1.42*** (2.99); Observations: 3,119 down to 3,079; R-squared = 0.15.
  - Table IC (post-1990, BoomI indicator): BoomI (No Lag) = 2.32*** (2.95); lags up to -4 are significant (coefficients 2.32*** to 1.65***); Observations ≈ 1,743; R-squared = 0.20.
  - Table ID (post-1990, BoomK indicator): BoomK (No Lag) = 1.61*** (2.63); lags up to -4 are significant; Observations ≈ 1,742; R-squared = 0.20.
- Main message: alternative boom definitions yield broadly similar full-sample patterns; post-1990 results are highly sensitive to individual countries.

### Crowding in or crowding out? (private investment responses with exact reported coefficients)
- Evidence indicates public investment booms are associated with lower private investment expenditures as a percent of GDP — consistent with crowding-out.
- Full-sample (Table 8, no lag): Boom VarI = -0.405*** (t = -4.155).
- Full-sample lagged coefficients: Boom VarI (-1) = -0.389***; Boom VarI (-2) = -0.372***; Boom VarI (-3) = -0.350***; Boom VarI (-4) = -0.346***.
- Post-1990 (Table 9, no lag): Boom VarI = -1.135*** (t = -7.166).
- Post-1990 lagged coefficients: Boom VarI (-1) = -1.212***; Boom VarI (-2) = -1.248***; Boom VarI (-3) = -1.230***; Boom VarI (-4) = -1.230***.
- Illustrative magnitudes:
  - A 5 percentage point rise in the public investment ratio associated with a 2 percentage point decline in private investment using whole-sample result (5 * -0.4).
  - A 5.7 percentage point decline using the post-1990 result (5 * -1.135).
- Other controls for private investment:
  - Greater life expectancy, lower black market exchange rate premium, and greater financial depth → higher private investment rates.
  - Inflation and export growth show little association with private investment.

### Pre- vs post-1990 booms and episode statistics (exact values)
- Number of boom episodes: 24.
- Mean size of booms (I/Y - I/Y(0)): 5.07 percentage points of GDP.
- Mean of maximum deviations across countries: 9.26 percentage points of GDP.
- Of 24 countries: 16 had booms predominantly pre-1990; 7 had booms predominantly post-1990; Dominican Republic’s second boom (1985–2003) not classified.
- Typical example: median pre-boom I/Y close to 5 percent of GDP; a typical boom would see a rise from 5.0 to 10.07 percent of GDP.
- Median length of pre-1990 booms: 15 years.

### Five case studies — high-level findings (exact illustrative figures)
- Mexico:
  - Public expenditure rose from 20.5 percent in 1971 to 32 percent in 1976.
  - Capital expenditures by SOEs increased at an annual rate of 29 percent.
  - Fiscal deficit grew from 2.5 percent of GDP to 10 percent in 1975.
  - Public investment rose from 7.2 to 10.8 percent of GDP between 1977 and 1981; major debt crisis in 1982.
- Bolivia:
  - Public sector expected to account for over 70 percent of total national investment in a 1975 plan.
- Korea:
  - Seoul–Busan Expressway: committee constrained by a budget of 33 billion won; final cost 42.9 billion won (30 percent over initial estimate); unit cost per km one-eighth of comparable Japanese cost.
  - Public investment stayed at approximately 5 percent of GDP for several decades—no sharp booms.
- Taiwan province of China:
  - “Ten Major Construction Projects” in the 1970s occurred after rapid growth began; financing via Postal Savings transfers; public investment used to accommodate growth.
- Philippines:
  - Public investment rose from 1.6 to 7.2 percent of GDP between 1972 and 1982.
  - Government investment rose as a share of GDP from 3.4 to 4.3 percent (mid-1970s shift toward energy-related investments).
  - Number of state owned enterprises increased from 30 to 96 between 1972 and 1984.

### World Bank project surge and performance (1970s)
- The World Bank increased the total number of projects approved each year during the 1970s.
- Evaluation lag: projects approved in a given year take on average 7 years to complete and then another year to be given a performance rating.
- Empirical pattern: steady decline in project performance over the 1970’s; project performance inversely associated with amount of lending.
- Reported regression output in source: coef = -.00330766, se = .00031874, t = -10.38.
- Additional numeric values appearing in figures/text: 3.76623, 4.8125, -144.23587.7647, -.436034, .490287.

### Common diagnostic findings from case studies
- No evidence of rational decision processes or consistent project-level analytics.
- Basic data to judge investment quality frequently unavailable.
- Cost-benefit analysis rarely informed implementation; forecasts systematically over-optimistic.
- Presidential leadership mattered in Korea and Taiwan; decentralized coalitions drove decisions in other cases.
- Distorted incentives: actors profited from project execution rather than operation; state enterprises used to reduce scrutiny.
- Financing ladder observed: domestic taxation/tax reform → domestic borrowing → borrowing backed by public guarantees → foreign financing.
- Inertia: programs continued despite adverse shocks, often shifting to alternative financing rather than curtailment.

### Policy-relevant implications and recommendations (diagnosis preserved)
- Large public capital expenditures alone are unlikely to guarantee sustained increases in growth; evidence casts doubt on unconditional claims about powerful spillovers from big push initiatives.
- Success of future public capital drives likely depends on overcoming:
  - incentive and agency problems;
  - weak project selection and due-diligence;
  - data collection and monitoring shortcomings;
  - governance and political-economy distortions that favor rent-seeking.
- Empirical strategy recommendations:
  - Distinguish contemporaneous from lagged associations to infer causality (public investment-driven growth should show lagged effects due to construction delays).
  - Use boom-episode identification (large, sustained shifts in I/Y) and narrative corroboration to treat investment effort as a plausibly exogenous forcing variable.
- Final diagnosis of five major problems with past surges:
  - Incentive problems rewarding execution over social worthiness.
  - Failure to collect key information at appraisal and ex-post and failure to act on evidence.
  - Chronically over-optimistic forecasts and lack of safeguards against self-serving analysis.
  - Shallow or non-existent economic analysis.
  - Use of public enterprises as vehicles for investment, enabling lower transparency and scrutiny.
- Policy implication: current public investment drives are more likely to succeed if governments adopt rigorous analytics, collect and use objective data, and safeguard decision processes against interest-driven distortions.

### Key numerical summary (preserving reported numbers)
- Sample: 124 countries, period 1960-2011, average 30 observations per country.
- Data points: 3,591; boom years: 335.
- Boom variable (I/Y – I/Y(0)): median 4.90; maximum 20.03.
- Public capital growth boom variable: median 4.36; maximum 27.34.
- Average difference across all 24 episodes between minimum and maximum investment ratio: just over 13 percentage points of GDP.
- Mean duration of booms: just over 14 years.
- Number of boom episodes: 24.
- Mean size of booms (I/Y - I/Y(0)): 5.07 percentage points of GDP.
- Mean of maximum deviations across countries: 9.26 percentage points of GDP.

*Source: _wp14148 - Introduction (content unit provided).*

### Introduction

### _wp14148 - Introduction

### Research question and scope
- Empirical examination of whether public investment drives (investment booms) over the past four decades triggered higher economic growth, productivity and long term development.
- Compares growth after episodes of large public investment rises with:
  - growth before those episodes in the same countries, and
  - countries that never had such episodes.
- Considers alternative definitions and measures of booms, World Bank project evidence, and five country case studies (Bolivia, Mexico, Philippines, Korea, Taiwan province of China).
- Sample covers 124 lower and middle income countries with WEO data for at least one year during 1960-2011; average of 30 observations per country.

### Key data and measurement features
- Two primary public investment measures used:
  - public investment ratio (I/Y), preferred for timing of booms because it is a direct measure of spending effort; and
  - growth of the public capital stock (constructed from investment flows).
- Public capital stock construction begins with an initial capital stock chosen to avoid implausible discrepancies between capital growth and the investment ratio.
- Boom-dating (preferred method):
  - compute unweighted 5-year forward and backward moving averages of I/Y;
  - identify years where the forward-backward difference exceeded the 80th percentile for three or more years and exceeded 4 percentage points of GDP;
  - start = year of first substantial rise; end = year when I/Y fell back to at least 120 percent of its pre-boom level (if never fell back, classified as continued long boom);
  - refined by graph inspection and corroborated with capital growth series.
- Alternative (robustness) method:
  - boom = investment ratio above the 70th percentile for three or more years; boom dummy = 1 during those years.
- Final sample: of 3,591 data points, 335 (~a little fewer than 10 percent) are rated as boom years.
- Boom variable (I/Y – I/Y(0)) based on public investment ratio: median 4.90, maximum 20.03.
- Boom variable based on public capital growth: median 4.36, maximum 27.34.
- Selection and exclusions: some countries removed for poor data or contextual reasons (e.g., Eritrea, Iran, Lesotho, Mauritania, Mozambique, Angola); ultimately 21 countries were determined to have experienced booms according to the preferred method (table in source).

### Conceptual framework and identification logic
- Model ingredients:
  - production function with public capital stock as input: Y = A Z^{εψ} K^{β} L^{α} (notation preserved from source);
  - public capital accumulation with inefficiency parameter s (0<s<1) capturing loss through corruption and an effective elasticity reduced by ε (0<ε<1) capturing poor project selection;
  - public finance constraint linking tax revenues and public investment (tax rate τ or investment rate as the policy parameter).
- Two empirical identification strategies to separate production-function effects from simultaneity with public finance:
  1. Identify episodes where I/Y rises as a deliberate policy choice (treat I/Y variation in booms as the forcing variable).
     - Rationale: very large, sustained booms (average difference across all 24 episodes between minimum and maximum investment ratio was just over 13 percentage points of GDP; mean duration just over 14 years) are likely deliberate policy shifts.
     - Narrative corroboration cited for several booms (e.g., Egypt 1972-78 plan; Morocco 1971 King’s speech; Nigeria 1970 reconstruction plan).
  2. Use timing/lags: large public investments (roads, dams) have long construction lags, so production-function effects should appear with multi-year lags; contemporaneous positive I/Y–GDP associations are more consistent with reverse causality (good years → higher I/Y).
- Key empirical prediction used: growth driven by production-function effects of public investment should show little contemporaneous I/Y–GDP association but positive lagged effects; growth driven by productivity or other shocks should show contemporaneous association with no lagged effect.

### Principal empirical findings (summary)
- Econometric evidence reveals:
  - small positive and instantaneous (contemporaneous) associations between public investment booms and economic growth;
  - little or no long-run impact of booms on growth.
- Specific diagnostics and patterns:
  - Most of the positive association occurs immediately (a spending boom tends to be immediately associated with a rise in GDP in the same year), but not in subsequent years.
  - F-tests fail to reject that the long run impact (sum of coefficients on lagged investment booms) is zero.
  - The estimated associations are small and not of the magnitudes suggested by big push models.
- Country-case observations:
  - Clear-cut historical examples of a public investment boom followed by sustained acceleration in GDP growth are difficult to find; several clear-cut booms were followed by slumps rather than booms.
  - Ethiopia in the past decade may be an exception (public investment, GDP, and several other variables rose together), but causality is difficult to establish.
  - Korea and Taiwan: rapid growth began before their public investment drives; their public investment drives were modest relative to economy size.
  - China and Vietnam growth revivals were driven by the abolition of price controls in Agriculture, not preceded by major public investment drives.
- World Bank project evidence and case studies:
  - World Bank project performance ratings and economic rate of return estimates are used where available, but micro-level investment-specific data are often fragmentary.
  - Case studies highlight pervasive problems in past investment drives: deep incentive and agency problems, avoidance of rational analysis, and difficulty obtaining or collecting the critical data to support rational investment choices.
- Heterogeneity over time:
  - Results differ by period; post-1990 data (including Ethiopia’s boom) show stronger results in favor of public investment than the rest of the sample.

### Mechanisms, risks, and alternative hypotheses
- Reasons why booms might have above-average impacts:
  - Big push models: simultaneous public investments can overcome coordination failures, create demand for each other, and raise average social returns above private returns.
- Reasons why booms might have below-average or no positive impacts:
  - Perverse selection: on-the-shelf projects pulled during booms may be those previously rejected for low impact.
  - Government behavior change: when money is loose, incentives to conduct rigorous cost-benefit analysis decline; special interest pressures may intensify.
  - Diminishing returns: additional investments once bottlenecks are resolved yield lower marginal returns.
  - Crowding-out or diversion of resources: financing public capital may crowd out more productive private investment or current government spending.
  - Governance and political economy factors: low-quality governance and limited checks and balances are associated with dramatically higher public investment and rent-seeking; political incentives can favor loss-making “white elephant” projects.
- The simultaneity problem: least squares regressions of output on public capital estimate a combination of production-function effects and public finance equilibrium relations; bi-directional causality between infrastructure and GDP is well documented in related literature.

### Relevant quantitative results and literature benchmarks (preserved as in source)
- Average difference across all 24 episodes between the minimum and maximum investment ratio: just over 13 percentage points of GDP.
- Mean duration of the booms: just over 14 years.
- Sample: 124 lower and middle income countries, period 1960-2011, unbalanced panel, average 30 observations per country.
- Of 3,591 data points, 335 are boom years (a little fewer than 10 percent).
- Boom variable (I/Y – I/Y(0)) statistics: median 4.90; maximum 20.03.
- Public capital growth boom variable: median 4.36; maximum 27.34.
- Example policy announcement: in September 2013 the government of Egypt announced $3.19 billion in additional infrastructure spending (1.2 percent of 2012 GDP) forecast to raise GDP growth from 2.3 percent to 3.5 percent, reduce unemployment from 13.2 to 9 percent and reduce the government’s budget deficit from 14 to 9 percent of GDP (source cited in text).
- Selected literature findings quoted:
  - Aschauer (1989): elasticity of output with respect to public capital approximately 0.3 (U.S. time series).
  - Gramlich (1994): subsequent work showed lower impacts or zero.
  - Calderon and Serven (2010): estimated African countries could boost annual economic growth by approximately 1.5 percentage points per year by cutting in half their infrastructure deficit with respect to other regions.
  - Gupta, Kangar, Papageoriou and Wane (2011): average elasticity of output with respect to public capital approximately 0.15 for lower income countries.
  - Canning and Pedroni (2008): conclude both long run and short run causality is bi-directional; some developing countries may have too much infrastructure.

### Policy-relevant implications and diagnostics
- Large public capital expenditures alone are unlikely to guarantee sustained increases in growth; the evidence casts doubt on unconditional claims about powerful spillovers and externalities from big push initiatives.
- Success of future public capital drives likely depends on overcoming:
  - incentive and agency problems,
  - weak project selection and due-diligence,
  - data collection and monitoring shortcomings,
  - governance and political-economy distortions that favor rent-seeking.
- Empirical strategy recommendations:
  - Distinguish between contemporaneous and lagged associations to infer causality (public investment-driven growth should show lagged effects due to construction delays).
  - Use boom-episode identification (large, sustained shifts in I/Y) and narrative corroboration to treat investment effort as a plausibly exogenous forcing variable.

*Source: _wp14148 - Introduction (content unit provided).*

### 0.70 percentage points (0.14*5=0.70).   The control variables have the anticipated signs except for

### _wp14148 - 0.70 percentage points (0.14*5=0.70).   The control variables have the anticipated signs except for

### Short-run association between public investment booms and growth
- Estimated contemporaneous association (Table 2, Boom VarI No Lag): Boom VarI = 0.14** (t-statistic shown as (2.29)).
- The estimated positive association largely dies out after a lag of one year; lags three and four years show negative coefficients (e.g., Boom VarI (-3) = -0.05 (−0.80); Boom VarI (-4) = -0.12** (−2.02) in Table 2).
- Fit of regressions: R-squared approximately 14 percent in Table 2 (R-squared = 0.14, 0.14, 0.14, 0.15, 0.17 across columns).
- Control variables (signs and significance reported in Tables 2 and 3):
  - Black Market Exchange Rate Premium: negative and highly significant (Table 2 examples: -0.27***, -0.26***, -0.26***, -0.25***, -0.29*** with t-statistics (-4.08), (-3.80), (-3.69), (-3.58), (-4.26)).
  - Life Expectancy at birth (years): positive and significant (0.03** in Table 2 across columns with t-statistics (2.33) to (2.43)).
  - Financial Depth (Deposit Bank assets % of GDP): negative but not always significant in Table 2 (e.g., -0.01* with t-statistics (-1.92), (-1.94), (-1.80), (-1.62), (-1.61)).
  - Inflation: negative and significant (reported as -0.00** with t-statistics around (-2.33) to (-2.34)).
  - Real Export growth (deflated by import price index): strongly positive and highly significant (0.09*** to 0.10*** with t-statistics e.g. (11.36), (11.58), (11.38), (11.55), (11.22)).

### Alternative measure: public capital stock growth (Table 3)
- Using Boom VarK (Kgrowth - Kgrowth(0)) strengthens short-run coefficients:
  - Boom VarK No Lag: 0.24*** (t-statistic (5.01)).
  - Boom VarK (-1): 0.16*** (t-statistic (3.10)).
  - Boom VarK (-2): 0.10* (t-statistic (1.71)).
  - Longer lags (Boom VarK (-3) = 0.05 (0.68); Boom VarK (-4) = -0.02 (−0.32)) are not significant.
- Fit and sample: Observations 1,744; 1,707; 1,671; 1,635; 1,595 across columns; R-squared 0.15, 0.15, 0.15, 0.17, 0.17.
- Control variables show similar signs and magnitudes as in Table 2 (e.g., Black Market Exchange Rate Premium ~ -0.28***; Real Export growth ~ 0.09*** to 0.10***).

### Robustness checks on short-run association
- Excluding natural resource intensive countries (Algeria, Mexico, Libya, Venezuela, Trinidad and Tobago and United Arab Emirates) left the estimated contemporaneous coefficient virtually unchanged for the first regression in Table 2.
- Using only positive values of the boom variable (excluding zero values) increased the estimated effect to 0.38 (se=0.058) for the strongest-specification regression.

### Long-run impact tests (Tables 4 and 5)
- Procedure: regressions with contemporaneous and five lags of the investment boom variable; estimated long-run impact equals sum of coefficients on boom variable and its lags.
- Using I/Y to measure booms (Table 4): estimated long-run impact (sum) across five specifications = -0.03, -0.02, -0.01, 0.00, -0.03 (average -0.02); average p-value = 0.71. Individual F-stat/p-values for null that sum=0: 0.33 (p-value 0.56), 0.15 (p-value 0.70), 0.08 (p-value 0.78), 0.00 (p-value 0.96), 0.39 (p-value 0.56).
- Using Public Capital Growth to measure booms (Table 5): estimated long-run impact (sum) across five specifications = 0.10, 0.10, 0.12, 0.10, 0.07 (average 0.10); corresponding F-stat for null that sum=0 = 1.65, 2.47, 2.50, 3.03, 1.92 with p-values 0.20, 0.12, 0.11, 0.08, 0.17. None are robustly significant at conventional levels.
- Conclusion: no robust evidence that investment booms exerted a positive long-term impact on the level of GDP even under the favorable assumption of causality running from booms to GDP.

### Post-1990 sample results and sensitivity (Tables 6 and 7; figures described)
- Restricting sample to post-1990 (Table 6) produces stronger positive associations for Boom VarI:
  - Boom VarI No Lag: 0.31*** (t=3.14).
  - Boom VarI (-1): 0.33*** (t=3.42).
  - Boom VarI (-2): 0.34*** (t=3.76).
  - Boom VarI (-3): 0.30*** (t=3.17).
  - Boom VarI (-4): 0.24** (t=2.40).
  - Observations: 1,180; 1,174; 1,169; 1,161; 1,151. R-squared across columns = 0.15, 0.15, 0.16, 0.16, 0.16.
  - In this sample financial depth and inflation are not statistically significant.
- Using public capital stock growth in the post-1990 sample (Table 7) yields weaker and less consistent results:
  - Boom VarK No Lag: 0.31*** (t=2.76).
  - Lags generally not significant except Boom VarK (-4) = 0.17** (t=2.07).
  - Observations: 1,171; 1,164; 1,156; 1,146; 1,135. R-squared = 0.15, 0.15, 0.15, 0.15, 0.16.
- Sensitivity to individual countries:
  - The positive post-1990 result in Table 6 is driven by Ethiopian observations; removing Ethiopia makes the estimated contemporaneous association no longer significant (figures discussed show estimated coefficients: with Ethiopia coef = .30778141, se = .09798922, t = 3.14; without Ethiopia coef = .20123576, se = .11155869, t = 1.8).
  - Uganda’s role differs depending on whether the boom is measured by investment ratio or capital growth; results are sensitive to measurement choice.
- Interpretation: post-1990 evidence base is small (few post-1990 booms) and empirically fragile; results invite further research on country-specific cases (Ethiopia, Uganda) rather than implying a general post-1990 pattern.

### Sensitivity to alternative boom-year selection method
- Alternative selection method: define a boom as when the variable exceeds its 70th percentile for a specific country for at least three consecutive years; applied to both investment ratio and public capital growth yielding indicator boom variables (value 1 during boom years, 0 otherwise).
- The three sets of boom years (the paper’s preferred set called “Set 1”, the percentile-based for investment ratio “Set 2”, and the percentile-based for capital growth “Set 3”) are reported as quite different; the text notes there are 363 observations defined as booms in Set [sentence truncated in source].

*Italic source attribution: IMF working paper content from the provided PDF excerpt.*

### 1.  The probability of an observation being classified as a boom in Set 2 given that it was so classified

### 1.  The probability of an observation being classified as a boom in Set 2 given that it was so classified in Set 1

### Robustness of boom-year identification
- Probability statistics:
  - The probability an observation is classified as a boom in Set 2 given it was so classified in Set 1 is 41 percent.
  - The probability for Set 3 given Set 1 is 49 percent.
  - The probability of being defined as a boom in Set 3 given that an observation was so defined in Set 2 is 24 percent.
- Regression coverage:
  - Regressions use boom years from Sets 2 and 3 and a 0/1 indicator for booms; results reported in Appendix 1 for four combinations: Sets 2 and 3, and full and post-1990 samples.
- Main patterns:
  - For the whole sample period (tables 1A (Set 2) and 1B (Set 3)), major results are similar to those using Set 1.
  - The investment boom variable is sometimes not statistically significant; when significant, significance is for the contemporaneous term or the first few lags, never for longer lags.
  - Other RHS variables show similar results across methods.
- Sensitivities and country-specific pitfalls:
  - Post-1990 regressions (tables 1C and 1D) show huge apparent impacts of booms on growth; results are highly sensitive to individual countries, especially Ethiopia, Nigeria, and Panama.
  - Ethiopia’s inclusion drives statistical significance of contemporaneous associations; exclusion eliminates that significance.
  - Nigeria: ambiguous dating of boom periods; alternative methods (e.g., treating back-to-back booms 1970–1988 and 1988–2004 or using a 70th percentile cutoff) can misclassify long contiguous periods as booms despite low mid-period investment; authors omit Nigeria from empirical analysis as an inherently difficult case.
  - Panama: Set 3 (70th percentile-based, using smoothed series) misclassifies an extended 2000–2011 boom because of a pre-boom plateau slightly above the 70th percentile; Set 1 (identifying boom years as 2005–2011) preferred.
- Conclusion on methods:
  - Using all data, alternative methods deliver similar results to the preferred Set 1.
  - Post-1990 differences reflect small sample size and sensitivity to particular countries; authors view tables IC and ID (post-1990 results based on Sets 2 and 3 for Nigeria/Panama) as inferior data and place low weight on those regressions.

### Crowding in or crowding out?
- Concept:
  - Reduced-form estimates capture both direct productivity impacts of public capital and productivity improvements via stimulation of private investment.
  - Evidence indicates public investment booms are associated with lower private investment expenditures as a percent of GDP — consistent with crowding-out rather than crowding-in.
- Regression evidence (summary of Tables 6–9):
  - The public investment boom variable has a consistent negative association with private investment rates; the negative association persists with lags up to four years.
  - Other control variable associations:
    - Greater life expectancy, lower black market exchange rate premium, and greater financial depth are associated with higher private investment rates.
    - Financial depth has a stronger association with private investment than with overall economic growth.
    - Inflation and export growth show little association with private investment.
- Magnitudes and sample differences:
  - Comparing full-sample and post-1990 sample estimates:
    - Full-sample coefficient (Table 8, no lag): Boom VarI = -0.405*** (t-statistic -4.155).
    - Full-sample lagged coefficients: Boom VarI (-1) = -0.389***; Boom VarI (-2) = -0.372***; Boom VarI (-3) = -0.350***; Boom VarI (-4) = -0.346***.
    - Post-1990 sample (Table 9, no lag): Boom VarI = -1.135*** (t-statistic -7.166).
    - Post-1990 lagged coefficients: Boom VarI (-1) = -1.212***; Boom VarI (-2) = -1.248***; Boom VarI (-3) = -1.230***; Boom VarI (-4) = -1.230***.
  - Illustrative magnitude calculations (authors’ presentation):
    - Other things constant, a 5 percentage point rise in the public investment ratio was associated with:
      - a 2 percentage point decline in private investment using the whole sample results in table 9 (5 * -0.4).
      - a 5.7 percentage point decline using the post-90 results (5 * -1.135).
- Interpretation:
  - Evidence suggests crowding out of private capital formation was an important reason for muted macroeconomic impacts of public investment booms.
  - This contrasts with Cobb-Douglas model presumptions that public capital stimulates private capital formation.

### Pre- vs post-1990 booms: episodes and growth performance
- Episode inventory and summary statistics (Table 10):
  - Number of boom episodes: 24.
  - Mean size of booms (I/Y - I/Y(0)): 5.07 percentage points of GDP.
  - Mean of maximum deviations across countries: 9.26 percentage points of GDP.
  - Of 24 countries:
    - 16 had booms predominantly pre-1990.
    - 7 had booms predominantly post-1990.
    - Dominican Republic’s second boom (1985–2003) is not classified into either group.
  - Example: since the median public investment ratio prior to the booms was close to 5 percent of GDP, a typical boom would see a rise from 5.0 to 10.07 percent of GDP.
  - Median length of pre-1990 booms (as illustrated) was 15 years.
- Comparative dynamics (Figures 12–16 summary):
  - Public investment medians:
    - Both pre-1990 (16 countries) and post-1990 (7 countries) groups show strong rises in median public investment ratio after t=0 (start of each boom).
  - Public capital growth:
    - Post-90 boom countries had a sharper rise at the beginning of the boom when measured by capital growth.
  - GDP growth patterns:
    - Pre-1990 group: GDP growth rose during the first 5 years of the 15-year boom and fell sharply in later years of the boom.
    - Post-1990 group: data are limited (seven countries, six years of GDP growth after booms commenced); median growth rose before booms and remained high in the early years of booms. It is too early to conclude whether a later slump (as in pre-1990 booms) will occur.
  - Cross-group comparison:
    - For years immediately after the start of booms, mean economic growth in post-1990 countries is not hugely different from pre-1990 countries.

### Five case studies (high-level findings)
- Mexico:
  - Major upsurge in public investment in the 1970’s.
  - Public expenditure rose from 20.5 percent in 1971 to 32 percent in 1976.
  - Capital expenditures by SOEs increased at an annual rate of 29 percent.
  - Fiscal deficit grew from 2.5 percent of GDP to 10 percent in 1975.
  - Public investment rose from 7.2 to 10.8 percent of GDP between 1977 and 1981; major debt crisis and crash came in 1982.
  - Many public sector investments later evaluated as yielding very low social returns; lack of hard performance data and secrecy around SOE investments were noted.
- Bolivia:
  - 1970s public investment spree tied to a state-led development strategy; public sector expected to account for over 70 percent of total national investment in a 1975 plan.
  - Petroleum export revenues underperformed; foreign debt of public sector institutions rose and use of public-sector guarantees increased implicit liabilities.
  - National System of Projects (1974) intended project screening but was circumvented; project execution and evaluation were poor and cross-subsidies obscured performance data.
- Korea:
  - Three major episodes: post-war reconstruction; reaction to bottlenecks from rapid growth; and a large-scale highway construction “big push”.
  - Reconstruction financed in part by foreign aid and Korean government funds; did not put government heavily in debt.
  - Seoul–Busan Expressway: committee constrained by a budget of 33 billion won; final cost 42.9 billion won (30 percent over initial estimate) but unit cost per km was one-eighth of comparable Japanese cost; construction unusually rapid.
  - Financing strategy prudent: early sections financed with tax revenue, later sections increasingly financed by tolls on previously constructed sections; public investment stayed at approximately 5 percent of GDP for several decades—Korea did not experience sharp public investment booms.
  - Strong implementation discipline (competitive bids, cost control) and alignment with market-driven urban-centered growth.
- Taiwan province of China:
  - Public investment not presented as a leading development driver; growth began before the 1970s public investment upsurge.
  - Export processing zones (1966–1969) provided infrastructure for export industries.
  - Financing mechanism: Postal Savings transfers to public utilities via the Central Bank helped scale infrastructure as household saving rose.
  - “Ten Major Construction Projects” in the 1970s occurred after rapid growth had begun and financed projects that largely supported exports and broad industry benefits.
  - Policy emphasis: Pragmatism and adherence to market signals; public investment used to accommodate growth rather than to lead it.

*Source: _wp14148 - 1.  The probability of an observation being classified as a boom in Set 2 given that it was so classified*

### 1973.  Both were reactions to perceived inadequacies after growth had occurred for many years.

### _wp14148 - 1973.  Both were reactions to perceived inadequacies after growth had occurred for many years.

### Philippines: public investment surge under Marcos
- Public investment spending rose from 1.6 to 7.2 percent of GDP between 1972 and 1982.
- Government investment rose as a share of GDP from 3.4 to 4.3 percent (mid-1970s shift toward energy-related investments).
- The number of state owned enterprises increased from 30 to 96 between 1972 and 1984; many of the 96 were actually holding companies overseeing approximately 149 subsidiaries.
- The military budget grew by a factor of four.
- Institutional and implementation features:
  - National Economic and Development Authority (NEDA) was the planning agency.
  - Public investment was used to consolidate political authority and foster crony-capitalists and public employees loyal to the administration.
  - Methods of rent extraction included kickbacks, overpricing, high consultant fees, and diversion of funds.
  - Road building was transferred to the military at inflated prices.
- Financing and inertia:
  - External borrowing rose sharply to finance the program.
  - Despite the 1975 collapse in international commodity prices, the investment program continued, indicating programmatic inertia and vested interests favoring execution over evaluation.
  - A second terms of trade shock in 1980 prompted further expansion financed partly through foreign loans after anticipated foreign equity inflows did not materialize.
- Implementation incentives:
  - “The result was that profits in many of the investment projects undertaken during the Marcos years were made at the investment and construction stage and not from profitable operation of the facilities constructed.” P. 471.
  - Critical factor: interests rewarded for executing projects (construction phase) rather than for profitable or socially beneficial operation.

### Common findings from case studies (Bolivia, Mexico, Philippines; Taiwan province of China, Korea)
- No evidence of rational decision process; non-existent analytics and data.
- Basic data to judge investment quality frequently not available (for example, how much was spent, on what and by whom).
- No evidence of rational public decision-making based on comparison of alternative investments using project analysis.
- Cost-benefit analysis was mentioned once (Bolivia) but results were ignored in implementation.
- Bolivia: complex cross-financing and cross-subsidization schemes obscured basic facts.
- Micro-evidence on the impact of investments ex-post is never cited.
- Presidential leadership mattered in Korea and Taiwan province of China; in other cases decentralised coalitions of interests drove decisions.
- Distorted incentives: many implementing groups stood to profit from doing the program, not from whether it had positive economic consequences.
- Interest groups favoring investment programs included politicians, regional government officials, private firms and sub-contractors, professional consultants, and laborers.
- State enterprises were used to implement investments to exploit less stringent reporting, ability to borrow on own collateral, and for political patronage.
- Inertia: programs continued despite problems.
- Obstacles prompted shifts to alternative financing rather than curtailment (shift to foreign borrowing or borrowing backed by public guarantees in Bolivia, Mexico, Philippines).
- Mexico example: Finance Minister fired for opposing the program; program continued under direct presidential control.
- Ladder of financing options observed: domestic taxation/tax reform → domestic borrowing → borrowing backed by public guarantees → foreign financing.
- Initial justifications consistently based on forecasts of benefits that were too high; forecasts were overoptimistic in every example cited.

### World Bank project surge and performance (1970s)
- The World Bank increased the total number of projects approved each year during the 1970s under Robert McNamara.
- Project performance ratings: 1-6 scale (6 = highly satisfactory; 1 = highly unsatisfactory).
- Evaluation lag: on average, projects approved in a given year take 7 years to complete and then another year to be given a performance rating; full verdict on a cohort may be available by about 1990 for a 1979 cohort. It may take as much as ten years for sufficient completion and evaluation.
- Empirical pattern:
  - There was a steady decline in project performance over the decade of the 1970’s.
  - Project performance was inversely associated with the amount of lending during the 1970’s.
  - Regression output reported in source: coef = -.00330766, se = .00031874, t = -10.38
  - Additional numeric values appearing in the figures/text: 3.76623, 4.8125, -144.23587.7647, -.436034, .490287 (reported as printed in source figures).

### Summary and conclusions on public investment booms
- Primary research question: Have major public investment drives in practice accelerated national economic growth?
- Main conclusion: “The answer appears to be ‘probably very little’.” The conclusion pertains to big increases in public capital spending (booms), not routine levels of public investment.
- Timing and effects:
  - Public investments take several years to finish; distinguishing short-term demand-driven output effects from long-term supply effects of completed capital is essential.
  - Empirical evidence using the full sample fails to support the hypothesis that public capital booms causally increase long-term GDP growth via lagged effects. Lag terms are mostly not statistically significant.
  - Contemporaneous positive associations often reflect governments spending more during boom times or short-term Keynesian demand stimulation, not long-term supply improvements.
- Sensitivity and exceptions:
  - Some post-1990 regression results initially appeared contrary but were driven by Ethiopia (and sensitive to Uganda data definitions and Panama measurement). These do not indicate a generalized post-1990 effect.
- Public capital booms tended to crowd-out private capital accumulation (observed in whole sample and post-1990 booms), providing a mechanism for limited growth impact.
- Case-study contrasts:
  - Bolivia, Mexico, Philippines: narratives show lack of rational selection, overoptimistic forecasts, low or negative returns, inertia, vulnerability to abuse, and use of state enterprises to circumvent scrutiny.
  - Korea and Taiwan province of China: rapid growth preceded targeted investment programs that addressed bottlenecks after congestion emerged; programs were more focused and disciplined (Korea’s public investment in percent of GDP remained roughly constant at five percent; Taiwan’s drive commenced in 1973). Korea limited cost overrun to 30 percent.
- The World Bank cohort evidence is consistent with diminishing returns to project lending when the number of projects expands sharply.
- Policy-relevant diagnosis of five major problems with past public investment surges:
  - Incentive problems: actors benefit from executing investments rather than ensuring social worthiness.
  - Failure to collect key information at appraisal and ex-post and failure to act on evidence.
  - Chronically over-optimistic forecasts and lack of safeguards against self-serving analysis.
  - Shallow or non-existent economic analysis.
  - Use of public enterprises as vehicles for investment, enabling lower transparency and scrutiny.
- Relevance for future booms:
  - Success of future public investment drives depends on whether the five problems above have been sufficiently overcome.
  - The use of public enterprises as vehicles for investments has diminished over the past thirty years, but incentives of politicians and construction firm leaders and the prevalence of over-optimistic forecasts and weak evidence-based decision-making persist.
  - Recommendation implication: current public investment drives are more likely to succeed if governments adopt rigorous analytics, collect and use objective data, and safeguard decision processes against interests that distort spending decisions.

*Source: _wp14148 - 1973.  Both were reactions to perceived inadequacies after growth had occurred for many years.*

### References

### _wp14148 - References

### References list
- Aschauer, D. (1989) “Is Public Expenditure Productive?” Journal of Monetary Economics, 23: 127-2000.
- Berg, Andrew, Jonathan D. Ostry and Jeromin Zettelmeyer, (2012) “What makes Growth Sustained?” Journal of Development Economic 98 149-166.
- Berg, Elliot and Andrew Berg (1983) “Absorptive Capacity in the Sahel Countries” Elliot Berg Associates: Arlington VA.
- Buffie, Edward E., Andrew Berg, Catherine Pattillo, Rafael Portillo, Luis-Felipe Zanna, (2012), “Public Investment, Growth and Debt sustainability: Putting Together the Pieces”, IMF Working Paper 12/144.
- Canning, David and Peter Pedroni (2008), “Infrastructure, Long-run Economic Growth and Causality Tests for Cointegrated Panels”, The Manchester School, Vol. 76, No. 5. Special Issue 2008, pp. 504-527.
- Calderon, Cesar, and Luis Serven, (2010), Infrastructure and Economic Development in Sub-Saharan Africa, Journal of African Economies, Vol. 19, pp. i13-i87.
- Estache, A., (2006) “Infrastructure: A survey of recent and upcoming issues, World Bank mimeo.
- Gupta, Sanjeev, Alvar Kangur, Chris Papageorgiou, and Abdoul Wane, (2011) “Efficiency Adjusted Public Capital and Growth” IMF working paper /11/217.
- Gramlich, Edward M., (1994) “Infrastructure Investment: a Review Essay”, Journal of Economic Literature, Vol. 32, No. 3, pp. 1176-1196.
- Harberger, Arnold C., “Observing the Ethiopian Economy” USAID, November 2010.
- Hulten, C. (1996) “Infrastructure Capital and Economic Growth: How well you use it may be more important than how much you have” NBER Working Paper 5847.
- Hulten C. R. and Schwab, R.M., (1991) “Is there too little Public Capital?” American Enterprise Institute Conference on Infrastructure needs.
- International Monetary Fund, “The Federal Republic of Ethiopia: 2012 Article IV Consultation” IMF Country Report no. 12/287, October 2012.
- Li, Kuo-Ting, (1995) “Evolution of Policy behind Taiwan’s Development Success” 2nd Edn. : Singapore: World Scientific.
- Robert S. Doher and Ponciano Intal, Jr. “The Marcos Legacy: Economic Policy and Foreign Debt in the Philippines.
- Chung-yum, Kim (2011) “From despair to hope: Economic Policy making in Korea”, Seoul: Korea Development Institute.
- Devarajan, Shantayanan, Vinaya Swaroop, and Heng-fu Zou. 1996. "The Composition of Public Expenditure and Economic Growth." Journal of Monetary Economics 37, no. 2: 313-344.
- Patinkin, Don, (1993) “Israel's Stabilization Program of 1985, Or Some Simple Truths of Monetary Theory” The Journal of Economic Perspectives, Vol. 7, No. 2 (Spring, 1993) pp. 103-128.
- Fernald, John G., (1999) “Roads to Prosperity? Assessing the Link between Public Capital and Prosperity” The American Economic Review, Volume 89, No. 3, pp. 619-638.
- Murphy, Kevin M., Andrei Shleifer and Robert W. Vishny, “Industrialization and the Bid Push” Journal of Political Economy, Vol. 97, No. 5 (Oct., 1989) pp.1003-1026.
- Thompson-Reuters news service “Africa Development Bank emphasizes infrastructure as a growth spur”. July 8th 2013.
- Romp, Ward and Jakob de Haan “Public Capital and Economic Growth: A Critical Survey” Perspektiven der Wirtschaftspolitik, 2007, v. 8, pp. 6-52.
- Robinson James A. and Ragnar Torvik, (2005) White Elephants” Journal of Public Economics 89, pp. 197-210.
- Rostow, Walt W. The Stages of Economic Growth: A Non-Communist Manifesto. Cambridge: Cambridge University Press, 1960.
- Rosenstein_Rodin, Paul N., “Notes on the Theory of the Big Push” in Economic Development for Latin America, Edited by Howard S. Ellis and Henry C. Wallich. New York: St. Martin’s, 1961.
- Holtz-Eakin, D. (1994) Public Sector Capital and the Productivity Puzzle, Review of Economics and Statistics 76, pp. 12-21.
- Keefer, Philip and Stephen Knack (2007), Boondoggles, Rent-Seeking, and Political Checks and Balances: Public Investment under Unaccountable Governments, The Review of Economics and Statistics, Vol. 89, No. 3, pp. 566-572.
- World Bank (1972) “Current Economic Position and Prospects of the Arab Republic of Egypt” December 1972 Report No. EMA-56a.
- World Bank (1974) “Current Economic Position and Prospects of Morocco” February 1974 Report No. 329-MOR.
- World Bank (1971) “Current Economic Position and Prospects of Nigeria” January 1971 Report No. AW-22a.

### Appendix I — Regression results (alternative boom selection, indicator variable for booms)
- Summary: Regressions of Annual Growth in Real GDP per-person on alternative public investment boom variables and other regressors. Two alternative boom definitions are used:
  - BoomI: indicator = 1 whenever I/Y exceeds its 70th percentile for three or more years.
  - BoomK: indicator = 1 whenever public capital growth exceeds its 70th percentile for three or more years.
- Common control variable coefficients (reported across tables):
  - Black Market Exchange Rate Premium: -0.20*** (Table IA No Lag) with t-statistic (-3.61); similar -0.20*** to -0.21*** across specifications and tables with t-statistics reported.
  - Life Expectancy at birth (years): coefficients reported as 0.00, -0.00, or 0.02 depending on sample; t-statistics reported (e.g., (0.10), (-0.01), (1.60)).
  - Financial Depth (Deposit Bank assets % of GDP): -0.02*** with t-statistics around (-7.85) to (-8.16).
  - Inflation: -0.00** or -0.00 with t-statistics around (-2.45) to (-1.37) depending on sample.
  - Real Export growth (deflated by import price index): 0.10*** with t-statistics around (13.40) down to (9.19) in post-1990 samples.
- Table IA (full sample, BoomI indicator): key coefficients and stats
  - BoomI (No Lag): 0.69 (1.45)
  - BoomI (-1): 0.12 (0.25)
  - BoomI (-2): 0.03 (0.07)
  - BoomI (-3): -0.32 (-0.66)
  - BoomI (-4): -0.79* (-1.68)
  - Observations: 3,143 (No Lag) down to 3,109 (Lag 4)
  - R-squared: 0.15 across specifications
- Table IB (full sample, BoomK indicator): key coefficients and stats
  - BoomK (No Lag): 1.25*** (2.77)
  - BoomK (-1): 1.42*** (2.99)
  - BoomK (-2): 1.10** (2.16)
  - BoomK (-3): 0.70 (1.32)
  - BoomK (-4): -0.02 (-0.04)
  - Observations: 3,119 (No Lag) down to 3,079 (Lag 4)
  - R-squared: 0.15 across specifications
- Table IC (post-1990 sample, BoomI indicator): key coefficients and stats
  - BoomI (No Lag): 2.32*** (2.95)
  - BoomI (-1): 2.86*** (3.64)
  - BoomI (-2): 2.96*** (4.02)
  - BoomI (-3): 3.16*** (4.39)
  - BoomI (-4): 1.65*** (2.89)
  - Observations: 1,743 (No Lag) down to 1,738 (Lag 4)
  - R-squared: 0.20 across specifications
- Table ID (post-1990 sample, BoomK indicator): key coefficients and stats
  - BoomK (No Lag): 1.61*** (2.63)
  - BoomK (-1): 2.32*** (3.61)
  - BoomK (-2): 3.06*** (4.75)
  - BoomK (-3): 2.92*** (4.16)
  - BoomK (-4): 2.28*** (4.43)
  - Observations: 1,742 (No Lag) down to 1,736 (Lag 4)
  - R-squared: 0.20 across specifications
- Significance notation used in tables:
  - *** p<0.01, ** p<0.05, * p<0.1
- Additional table notes:
  - Robust t-statistics are reported in parentheses for all coefficient estimates.

### Appendix II — 18 selected examples of public investment booms (selected country-year statistics)
- Uruguay
  - year range: 1969-2011; Gov I/GDP 5-yr Moving Average: 2.93 and 14.69
- Venezuela
  - year range: 1968-2011; Gov I/GDP 5-yr moving average: 5.74 and 23.66
- Morocco
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 0.33 and 12.63
- Egypt
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 3.02 and 18.92
- Algeria
  - year range: 1990-2011; Gov I/GDP 5-yr moving average: 6.03 and 19.00
- Iran
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 4.90 and 24.83
- Nigeria
  - year range: 1963-2011; Gov I/GDP 5-yr moving average: -0.39 and 17.27
- South Africa
  - year range: 1966-2011; Gov I/GDP 5-yr moving average: 0.19 and 10.07
- Uganda
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 0.03 and 9.61
- Mexico
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 0.55 and 10.52
- Bolivia
  - year range: 1969-2011; Gov I/GDP 5-yr moving average: 2.22 and 10.69
- Philippines
  - year range: 1970-2011; Gov I/GDP 5-yr moving average: 0.00 and 5.00
- Cote D’Ivoire
  - year range: 1969-2011; Gov I/GDP 5-yr Moving Average: 0.97 and 14.31
- Dominican Republic
  - year range: 1969-2011; Gov I/GDP 5-yr Moving Average: 3.34 and 27.67
- Libya
  - year range: 1976-2009; Gov I/GDP 5-yr Moving Average: 4.77 and 29.83
- Ethiopia
  - year range: 1987-2011; Gov I/GDP 5-yr moving average: 3.25 and 18.13
- Trinidad and Tobago
  - year range: 1971-2011; Gov I/GDP 5-yr Moving Average: 0.20 and 15.00
- Jordan
  - year range: 1963-2011; Gov I/GDP 5-yr Moving Average: 0.06 and 23.29

*Source: _wp14148 - References (Appendices I and II).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14148.pdf_