Public Capital Stocks in Developing Countries

The paper begins by highlighting a key difference in how public capital is approached in research on OECD versus developing countries. Studies on OECD nations, particularly the United States, frequently employ the concept of 'public capital' or 'public spending.' In contrast, research on developing countries typically emphasizes 'infrastructure.' The authors note that while some studies using 'infrastructure' in developing countries' research might use definitions equivalent to public capital, many rely on past public investment flows or physical infrastructure measures. This distinction is crucial because it shapes the data used and the methods employed to assess public capital's productivity and efficiency. The study aims to specifically evaluate public capital stocks in developing countries using a methodology comparable to that typically used in OECD countries to allow for international comparisons, acknowledging that a simple stock level is insufficient for policy development, as other factors such as investment type, institutional arrangements, and private capital availability heavily influence policy choices. A homogenous definition of public capital stocks is identified as a key element required for a valid comparative study of the social rates of return of public investments.

The paper delves into the definition of capital stocks, aligning with the System of National Accounts (SNA, 1993), which includes both tangible and intangible assets used repeatedly in production for over a year. This contrasts with the 1968 SNA. The authors emphasize that this definition of public capital stocks differs from the concept of 'infrastructure.' They cite Calderon, Easterly, and Serven (2004) to emphasize that not all public sector spending on infrastructure is necessarily productive. The paper explores the use of the Perpetual Inventory Method (PIM) to estimate capital stocks, accumulating historical investment series and deducting retired assets. However, the limitations of the PIM in developing countries are highlighted, as research by Pritchett (1996) demonstrates that the cost of public investment doesn't always equal the value of resulting capital stocks; in some countries, a dollar invested might create less than a dollar's worth of public capital. Therefore, using the PIM directly can overvalue the available public capital, especially relevant given the ongoing discussion about the efficiency of public investment in developing countries (World Bank, 1994). The authors propose using two kinds of public capital stock estimates: PIM-based estimates as benchmarks and revised estimates incorporating efficiency adjustments. The methodology requires assumptions regarding depreciation patterns and initial stocks; a geometric pattern with time-varying depreciation rates is adopted, as the structure of public investments in developing countries has evolved significantly over the last decades. The paper points out that while the accumulated investment flows might serve as a good proxy for capital stocks in some cases, the same methodology applied to reference developing countries reveals a significant discrepancy between investment amounts and the increase in stocks.

The study employs data from the World Bank's World Development Indicators (WDI) on central government capital expenditure, encompassing fixed capital assets, land, intangible assets, and capital grants. These data are deflated using the GDP implicit price deflator to express investments in constant prices. The research considers 26 developing countries selected for data availability and relative political stability, with Mexico and Turkey included despite their OECD membership because their capital stocks weren't estimated in a prior study by Kamps (2004). The authors analyze the average public investment-to-GDP ratio, noting its decline, comparable to the decline observed in OECD countries in the 1970s, but with a delay in developing countries. The average ratio decreased from 5.15% (1974-1984) to 3.80% (post-1985) in the sample, while the decline in OECD countries occurred mainly in the 1970s, reaching 3.70% at the end of the 1990s, higher than the average for OECD countries. The variability in public investment ratios across the 26 countries is analyzed. Using Kamps's database (2004), the standard deviation for OECD countries (1960-2001) ranged from 1.00% to 1.80%, while for the developing countries (1974-1997) it fluctuated from 1.75% to 4.28%, indicating much higher volatility. The choice of a time-varying depreciation profile, rather than a constant one, for the PIM is discussed, which is justified by the substantial changes in the structure of public investments over the decades. The authors explain their use of a geometric depreciation pattern, similar to that used in previous studies. The research also incorporates assumptions about initial capital stock, using an artificial investment series going back to 1880 to construct an initial value, to address data limitations. This initial value and the chosen depreciation pattern are then used in the PIM calculation.

The section directly addresses the core issue: the inefficiency of public investment in creating productive capital, especially within the context of developing countries. The authors point out that the Perpetual Inventory Method (PIM), while a standard approach, suffers from significant limitations when applied to developing economies. The problem isn't with the PIM itself, but with its reliance on total monetary flows of gross investments. Research by Pritchett (1996) is cited to highlight this, showing that in many developing countries, a substantial portion of public investment doesn't translate into productive capital. Pritchett's estimate suggests that often, less than half of the money invested in projects genuinely contributes to increasing public capital stocks. This inefficiency leads to an overvaluation of public capital stocks when using the PIM alone. This problem is linked to the broader debate regarding the effectiveness of public investment in developing countries, a point emphasized by the World Bank in their 1994 report. This inefficiency underlines the need for a revised approach to accurately measure public capital stocks, going beyond simply using the monetary value of investments.

The authors propose estimating an efficiency function to capture the relationship between public investment and the actually productive component of that investment. This function, denoted f(I_t), represents the productive component of investments at time t. The difference between f(I_t) and I_t directly reflects the inefficiency. The greater this difference, the less reliable the PIM becomes. While knowing this function would allow for a more accurate estimation of public capital stocks, there's no universally accepted functional form. The paper explores two methods: a simplified approach assuming a linear and constant efficiency across investment amounts, and a more nuanced non-parametric approach. The linear simplification, though easy to implement, assumes a constant relative efficiency regardless of investment size, which is a potential limitation. The non-parametric method aims to estimate the efficiency function for a subgroup of countries and use that estimated functional form as a benchmark for the whole sample, rather than the efficiency values themselves. This requires three inputs: public investment series, depreciation rates (or their time profile), and a series of effectively available public capital net stocks in the reference countries. This last point presents significant difficulties because accurately estimating public capital stock in developing countries inherently requires knowing the efficiency function.

To address the difficulty of obtaining direct measures of effective public capital, the study proposes using physical measures of infrastructure as proxies for the actual public capital stock. This approach builds upon the work of Calderon, Easterly, and Serven (2004), who used a similar method for Latin American countries. The researchers employ a sectoral decomposition, comparing investment flows with changes in physical infrastructure measures. To ensure that private investment is negligible, they limit their analysis to periods where private investment in the selected sectors does not exceed 15%. Colombia and Mexico are selected as reference countries due to the availability of suitable data. They use a four-sector decomposition (electricity, telecommunications, roads, railways), analyzing the relationship between investment flows and changes in physical infrastructure measures. The efficiency function is estimated using local LOESS regression, which allows for flexibility in capturing the relationship between investment and stock increases. The analysis shows a stark difference between the results for the U.S., where the estimated efficiency function is close to a 45-degree line (indicating high efficiency), and the developing country cases (Colombia and Mexico), where the function is significantly below the 45-degree line, demonstrating considerable inefficiency. This difference highlights the challenges in applying standard methods directly to developing economies and underscores the importance of accounting for efficiency variations when estimating public capital stocks.

The study first generates benchmark estimates of net public capital stocks using the Perpetual Inventory Method (PIM). This method, while common for OECD countries, presents challenges when applied to developing nations, primarily because it assumes a direct link between monetary investment and the creation of productive capital, an assumption often not valid in developing contexts. The PIM requires assumptions about depreciation patterns and the treatment of initial stocks. The authors adopt a geometric depreciation pattern with time-varying depreciation rates, acknowledging that the structure of public investments has evolved considerably in developing countries. The results from the PIM provide a basis for comparison with results obtained from OECD countries using the same methodology, and to highlight the differences that may emerge between the two sets of results. The authors note that while in some cases the cumulative investment flows can act as a reasonable proxy for capital stocks, this is not generally the case for developing economies. A comparison of the results for the two reference countries with this method demonstrates a significant discrepancy between the quantity of investments and the resulting increase in capital stocks. This section lays the groundwork by establishing benchmark estimates for comparison with the adjusted estimates, which will account for the noted inefficiencies.

Recognizing the limitations of the PIM in capturing the true value of public capital stocks in developing economies, this section introduces an adjustment to account for the inefficiency of public investment. The study observes that, for their reference developing countries, the relationship between investment and stock increase appears nearly linear, implying that only a fraction of total investment is truly productive. This observation leads to a modified estimation approach, whereby new series of net public capital stocks are estimated by using only a fraction of the total investment. This approach incorporates sensitivity analysis, creating three estimates for each country corresponding to different values of this 'productive investment' fraction (a sort of efficiency parameter). The methodology is described which assesses the productive component of public investment by comparing the increase in the monetary value of the stocks to the current monetary value of public investments. It provides an estimate of the part of public investments that are genuinely productive in creating capital. The study is carried out for 26 developing countries (including Mexico and Turkey, two OECD members), selected for their data availability and political stability. Data on public investment is sourced from the World Bank's World Development Indicators (WDI), representing central government capital expenditure. This approach marks a crucial shift from simply relying on the total investment figures to a more realistic assessment that incorporates the realities of inefficient public spending in many developing economies.

This section presents the results of the public capital stock estimations and compares them to findings for OECD countries. The study shows the evolution of the public net capital stock to GDP ratio for the 26 developing countries, finding results similar to those concerning public investments in earlier sections. Benchmark results are compared with Kamps (2004a)'s data for OECD countries. A key difference emerges: while Kamps observes a decline in the OECD public capital-to-GDP ratio, the study finds an increase in the average ratio for the developing countries, rising from 58.6% in 1980 to 63.1% in 1990 (and 59.2% to 64.3% for a balanced panel of 19 countries). During the same period, the OECD ratio decreased from 57.8% to 55.3%. This contrasting trend is partly attributed to the differing depreciation patterns and a delay in the decline of public investments in developing countries compared to the 1970s decline in OECD nations. The study emphasizes that the observed increase in the ratio for developing countries does not necessarily imply a lack of shortages, given the inherent drawbacks of using the PIM in these contexts. The key takeaway is the identification of a deceleration in the growth of the public capital stock ratio, occurring in the 1990s, a significant delay compared to the deceleration observed in the United States. At the end of the 1990s, the average ratio in developing countries was similar to, or even greater than, that of OECD countries. This finding underscores the importance of accounting for the limitations of the PIM in drawing conclusions about public investment levels in relation to GDP.

The core of this comparative analysis centers on the differing trends observed in the ratio of public capital to GDP between developing countries and OECD nations. The study's findings for a panel of 26 developing countries contrast sharply with those of Kamps (2004a) for OECD countries. While Kamps reports a decline in the public capital-to-GDP ratio for OECD nations, the research reveals an increase in the average ratio for the developing countries examined. Specifically, the average ratio increased from 58.6% in 1980 to 63.1% in 1990, a contrasting trend to the decrease observed in OECD countries over a similar period. The authors offer several explanations for this divergence. They point out that the decline in the public investment-to-GDP ratio in developing countries began later, in the mid-1980s, whereas the decline in OECD countries primarily occurred in the 1970s. This timing difference explains the delayed effect on the stock ratios. The study acknowledges the heterogeneous nature of the data, acknowledging that the average ratios conceal significant variations across individual countries. The study also notes that even at the end of the 1990s, the average capital-to-GDP ratio in the developing countries remained comparable to, or even exceeded, that of OECD countries.

The paper explores potential reasons for the differing trends in public capital-to-GDP ratios between developing and OECD countries. Two major factors are highlighted: differences in depreciation patterns and the timing of fiscal adjustments. The observed differences in the trends in public capital to GDP ratios, between the developing and the OECD countries, are partly due to differences in the depreciation patterns. The delay in the decline of public investments in developing countries is also identified as an important factor contributing to the different trends in public capital to GDP ratios, between the two groups of countries. The average public investment-to-GDP ratio in the sample of developing countries declined only after the middle of the 1980s, while in most OECD countries this decline occurred in the 1970s. This time lag causes a corresponding delay in the reduction of the stock ratios in the developing countries compared to the OECD countries. The study points out that the decline in public investment in developing countries is comparable to that observed in the 1970s in OECD countries, citing the examples of the United States and France. The decline in developing countries is attributed to fiscal austerity measures implemented in the 1980s and 1990s, in contrast to the earlier decline experienced in OECD countries. The paper cautions against interpreting the similar or higher capital-to-GDP ratios in developing countries at the end of the 1990s as a sign of sufficient public investment, emphasizing the need to consider the significant drawbacks of using the PIM in developing country contexts.

This section further explores the robustness of the comparative analysis by addressing potential sources of bias. The paper acknowledges that its conclusions, especially regarding the comparison between developing and OECD countries, are dependent on various methodological assumptions. The influence of depreciation patterns on the estimates is particularly examined. The study emphasizes that the estimated aggregated depreciation rates are influenced by the composition of public investments, leading to a slight decrease in the time profile of depreciation rates for the developing countries in the sample. A sensitivity analysis is performed, comparing benchmark estimates to those obtained using alternative assumptions about depreciation rates. For instance, using a constant annual depreciation rate of 4% for the period 1970-2000 results in estimates that are consistently lower than the benchmark estimates. The authors note that the dynamics of the estimates remain similar in many aspects, regardless of the variations in the assumptions regarding the depreciation rates. This indicates a reasonable degree of robustness. The choice of a geometric depreciation pattern is also justified, citing previous research supporting its statistical approximation of the true depreciation rate when different capital goods are aggregated. The paper concludes this section by re-emphasizing the value of the benchmark estimates which are based on the same methodology and homogeneous investment data across countries, thereby strengthening the validity of the cross-country comparisons.

This section introduces a novel approach to improve the accuracy of public capital stock estimates, particularly for developing countries. The authors acknowledge that using monetary flows of investment alone (as in the PIM) can be misleading, especially in contexts where the efficiency of public investment varies. The inherent difficulty lies in the lack of reliable data on the actual public capital stock effectively utilized in the economy. To overcome this data constraint, the study proposes a method that leverages physical measures of infrastructure as proxies for the actual public capital stock. This strategy follows similar work by Calderon, Easterly, and Serven (2004) but with modifications to enhance precision. One key challenge in using a sectoral approach is the potential influence of private investment alongside public investment within certain infrastructure sectors. This is particularly true in some Latin American countries where the balance between public and private investment shifted considerably during the relevant period. To address this issue, the study focuses on periods where private investment is insignificant (less than 15% of total investment). This approach helps to isolate the impact of public investments and obtain more accurate estimates by minimizing the effects of private sector involvement.

The methodological approach focuses on a sectoral decomposition of public investments. The study uses the available databases on infrastructure from Canning to identify the physical measures that correspond to the public investment categories. These physical measures vary by sector. For electricity, it's electricity-generating capacity (in megawatts); for telecommunications, it is the number of telephone main lines; and for roads, it is the number of road kilometers (paved or total, depending on data availability). To ensure comparability, the study selects two reference Latin American countries, Colombia and Mexico, because of the detailed decomposition of public investments available in the Calderon, Easterly, and Serven (2004) database for these countries and to establish a correspondence between those investments and physical measures in the Canning's database. These choices are justified by the need to match investment categories with available physical infrastructure measures. The analysis is limited to periods where private investment does not exceed 15% of total investment, minimizing the interference from the private sector's investment. The selection of these two countries and time periods reduces the influence of factors unrelated to public capital creation and thus improves the analysis. The researchers analyze these sectors over specific time spans, which are chosen to exclude periods where private investment became significant. This is particularly important to ensure that the physical measures used truly reflect the impact of public sector investment.

The results of the sectoral analysis demonstrate a significant difference between developed and developing economies regarding the efficiency of public investment. For the United States, using the same methodology on roads and highways (1950-1992), the estimated efficiency function is quite close to the identity function. This suggests that the PIM is a reasonable approximation of public capital stocks for the U.S. The picture is notably different for Colombia and Mexico. The efficiency functions for the electricity, road, and telecommunications sectors are far below the 45-degree line, indicating significant underperformance. The telecommunications sector shows the highest relative efficiency, but even there, the estimated efficiency is low. The short sample periods used for the sectoral analysis (due to data limitations), lead to relatively imprecise estimates. To overcome this challenge, the authors aggregate data across sectors to generate a global efficiency function for each country, which reveals a strikingly similar linear function for both Colombia and Mexico, and shows that the productive component of public investments is much lower than what the PIM indicates. This finding supports Pritchett's (1996) conclusion that public investment often has substantially less impact in developing countries than its monetary value would suggest. The linearity of the estimated efficiency function enables the generalization of this finding to the broader panel of countries, allowing for adjusted public capital stock estimates that incorporate the estimated efficiency parameter. The authors recommend using this method to complement other approaches (investment flows or physical infrastructure measures), acknowledging the strengths and weaknesses of each.