The Price of Capital Goods: A Driver of Investment Under Threat

Abstract

Over the past three decades, the price of machinery and equipment fell dramatically relative to other prices in advanced and emerging market and developing economies. Using cross-country and sectoral data, we show that the decline in the relative price of tangible tradable capital goods provided a significant impetus to the capital deepening that took place during the same time period. The broad-based decline in the relative price of machinery and equipment, in turn, was driven by the faster productivity growth in the capital goods producing sectors relative to the rest of the economy, and deeper trade integration, which induced domestic producers to lower prices and increase their efficiency. Our findings suggest an additional channel through which rising trade tensions and sluggish productivity could threaten real investment growth going forward.

1 Introduction

For decades, real investment in machinery and equipment has outpaced real GDP growth in many emerging market and developing economies. Since 1995, the real investment rate in machinery and equipment has doubled, rising from about 3–6 percent of real GDP (Fig. 1). This capital deepening coincided with a steep decline in the price of capital goods relative to the price of consumption. The relative price of tangible tradable capital goods fell by over 50 percent since 1995 for the median emerging market and developing economy.¹ This process seemed to be accompanied by stronger trade integration in the capital goods producing sectors, with the rise in import penetration in this sector exceeding that in other sectors of the economy (Fig. 2).

Fig. 1

Sources: Penn World Table (PWT) 9.1; and authors’ calculations

Evolution of the relative price of machinery and equipment and investment rates. Note: Figure shows the cross-country median for emerging market and developing economies (for country sample see “Appendix 1”) of the real investment in machinery and equipment to real GDP ratio (blue line) and the year fixed effects from a regression of log relative prices on year fixed effects and country fixed effects to account for entry and exit during the sample period and level differences in the price of machinery and equipment relative to the price of consumption (red line). Year fixed effects are normalized to show percent change from the relative investment prices in 1995.

Fig. 2

Sources: World Input-Output Database (WIOD); and authors’ calculations

Relative import penetration. Note: The figures show the cross-country average and median for emerging market and developing economies of the ratio of import penetration for capital goods sector to overall economy. Import penetration is defined as total imports over value added.

Economists have long argued that the relative price of replaceable capital goods, especially machinery and equipment, is one of the key determinants of economic performance.² The fact that, in the cross section, the price of capital goods, relative to the price of consumption, is much higher in poor countries was considered crucial in explaining the lower investment rates, living standards and growth observed in these economies. Yet, there is little consensus on the underlying causes of the cross-country heterogeneity in the relative price of capital goods. Some have argued that it mainly reflects differences in countries’ productivity in the making of machinery and equipment or other tradable goods that could be exchanged for machinery and equipment (Hsieh and Klenow 2007). Others link it to distortionary policy choices, such as higher barriers to trade, taxes on capital goods, monopoly power in the production of machinery and equipment,³ or discriminatory pricing by exporters (Alfaro and Ahmed 2009). While the literature on the cross-country differences in relative capital goods prices blossomed, existing studies have largely neglected to examine the drivers of the changes in the relative price of capital goods over time.

In this paper, we revisit the debate about the key drivers of the relative price of investment and study the macroeconomic implications of the falling relative price of capital goods. First, we use newly available data from the 2011 round of the International Comparison Project (ICP) database to study whether absolute and relative prices of machinery and equipment are higher in countries with higher trade barriers.⁴ Previous studies highlight poor data quality as an important constraint in understanding the drivers of the cross-country dispersion of the relative price of investment goods (see, for example, Hsieh and Klenow 2007; Alfaro and Ahmed 2009). The 2011 ICP round introduces many methodological improvements to address data quality concerns of older rounds (see Feenstra et al. 2015; Deaton and Aten 2017; Alfaro and Ahmed 2009).

Second, we analyze the roles of trade integration and productivity in the decline of the relative price of capital over time. We combine sector-level tariff data constructed by Feenstra and Romalis (2014) and the World Input and Output database, which provides output prices and trade flows at the sector level, to study how trade-policy-induced changes in import penetration affected producer prices.

Third, we examine the effect of relative prices of capital goods on real investment rates over the last 30–60 years. While the theoretical link between the relative price of capital goods and investment is not hard to establish, the empirical evidence on this issue is scant and relies mostly on aggregate cross-sectional data from earlier periods (see, for example, Sarel 1995; Restuccia and Urrutia 2001). Using both country-level and sectoral data, we quantify how much of the increase in the real investment rate in machinery and equipment that occurred since the 1990s can be attributed to the decline in the relative prices of machinery and equipment.

Our analysis shows that the reduction in trade costs, and the associated rise in trade integration, was an important factor in the decline of the relative price of machinery and equipment in the past decades. Two pieces of evidence are consistent with this conclusion. First, according to the latest (2011) ICP data, across countries, those with higher trade costs tend to pay a higher price for a comparable basket of machinery and equipment in both absolute terms and relative to the price of consumption. This contrasts with previous findings by Hsieh and Klenow (2007), who do not find a negative correlation between the absolute price of capital goods and development, concluding that trade costs play no role in the variation in the price of capital goods.⁵ Second, analysis of sectoral producer price data suggests that relative prices are highly responsive to changes in import penetration. We find that rising import penetration lowers domestic producer prices both directly, as producers lower prices due to foreign competition or specialization, and indirectly, by boosting their productivity, which ultimately leads to lower prices. We combine the estimated coefficients with the change in the relative trade exposure in the capital goods sectors to provide an illustrative quantification of how much trade has contributed to the decline in the relative price of machinery and equipment during 2000–2011. We find that, on average, more than two-thirds of the fall in the relative price of tradable investment goods between 2000 and 2011 can be attributed to trade integration.

We also show that the decline in the relative price of capital goods has played a crucial role in increasing real investment rates over the past three decades. Cross-country panel regressions relying on over 60 years of data across 180 economies suggest that real investment in machinery and equipment is highly sensitive to its relative price. These results hold even after controlling for all global shocks, time-invariant country characteristics, and a host of other policies and time-varying factors shown by economic theory and previous studies to shape investment rates. In addition, we confirm our results when instrumenting the relative price of capital goods with import tariffs on capital goods relative to tariffs on consumption. The estimated sensitivities are also very similar if we use sectoral data instead. We analyze sectoral investment rates across 15 broad sectors in 18 economies during 1971–2015 from the EU KLEMS database. The sectoral analysis allows us to properly account for the role of all factors that affect overall investment within a country in a particular year, such as financial conditions, economy-wide growth prospects, quality of regulations, and the like. Overall, the decline in the relative price of capital goods can explain about 40 percent of the increase in real investment rates in the average economy since the 1990s.

These results are important not only to shed light on the academic debate on the underlying drivers of relative prices, but to draw attention to possible emerging risks, which may hamper much-needed capital deepening in low-income countries. Since trade integration has indeed played a key role in driving down the relative price of investment goods, the waning pace of trade liberalization and the slowdown in global trade would limit further declines in the price of capital goods. Even more immediate is the threat from higher trade barriers in some advanced economies.

The rest of the paper is organized as follows. Section 2 describes some key stylized facts on the absolute and relative price of capital goods from a variety of sources. Section 3 analyzes the drivers of the prices of machinery and equipment, while Section 4 presents estimates of the sensitivity of real investment to changes in relative prices. Section 5 concludes summarizing the key results and policy implications.

2 Stylized Facts

Since the 1990s, capital goods prices relative to consumption prices have displayed three key patterns.⁶ First, the relative prices of the four main types of fixed capital assets—structures, machinery and equipment, transportation equipment, and intellectual property products—have evolved quite differently (Fig. 3, panels 1–4). According to data in the Penn World Table version 9.1 across 180 countries, the prices of tradable investment goods, namely machinery and equipment and transportation equipment, have declined very significantly since the early 1990s when compared with the consumption deflator. The price of residential and nonresidential structures, on the other hand, has more closely tracked consumption prices and even increased since the mid-2000s, in relative terms, in advanced economies. Within tangible tradable capital goods, the dramatic decline in the relative prices of computing equipment (such as computer hardware, whose prices fell by 90 percent since 1990) and to a lesser extent communications equipment (whose prices fell by almost 60 percent) (Fig. 3, panels 5–6) supports the hypothesis that advances in information technology have played an important role in driving down the relative price of investment.⁷

Fig. 3

Sources: EU KLEMS; Penn World Table 9.1; World KLEMS; and authors’ calculations

Dynamics of relative prices across types of capital goods and broad country groups. Note: Panels 1–4 use data from the Penn World Table 9.1 capital detail file, while panels 5–7 use data from the EU and World KLEMS databases. The relative price of investment (for each type of capital good) is obtained by dividing the relevant investment deflator by the consumption deflator. The solid line plots year fixed effects from a regression of log relative prices on year fixed effects and country fixed effects to account for entry and exit during the sample period and level differences in relative prices. Year fixed effects are normalized to show percent change from the relative investment prices in 1995. Shaded areas indicate 95 percent confidence intervals. AEs advanced economies; EMDEs emerging market and developing economies.

These patterns also suggest that deepening trade integration might have supported the downward trend in capital goods prices. As depicted in Fig. 2, the rise in import penetration in the capital goods sector exceeded significantly that in other sectors of the economy. Among tangible capital goods, the decline in prices was most pronounced in those that can be more easily traded across borders. The production of machinery and equipment is also strongly embedded in global value chains, as depicted in Fig. 4.

Fig. 4

Sources: Eora MRIO database; and authors’ calculations

Backward participation in GVCs. Note: The figure depicts the median and interquartile range of the sector’s backward global value chain participation (defined as the foreign value added in exports) across all economies in the Eora MRIO database deemed to have sufficient data quality at the sectoral level during 1995–2015. For calculation of backward participation in GVCs, see Aslam et al. (2017).

The second notable pattern is the slowdown in the pace of decline in the relative price of machinery and equipment in recent years. While up-to-date data are not widely available, Fig. 3, panel 2 suggests that the decline in relative prices of capital goods has been less pronounced since the global financial crisis, coinciding with a slowdown in global trade and the process of trade liberalization. Byrne and Pinto (2015) document a similar slowdown in the decline of high-tech equipment price in the case of the USA.

Third, despite the broad-based decline in the relative price of tradable capital goods over time, the prices of these goods vary substantially across countries, especially relative to the price of consumption. According to the latest data from ICP, the absolute price of machinery and equipment in 2011 was inversely related to countries’ development levels, with lower-income countries facing slightly higher prices than advanced economies.⁸ The same basket of machinery and equipment costs about 8 percent more in the median low-income country compared to the median advanced economy (Fig. 5, panel 1).

Fig. 5

Sources: International Comparison Program (ICP) 2011; and authors’ calculations

Absolute and relative prices of machinery and equipment across countries in 2011. Note: The absolute price of machinery and equipment is the price level of machinery and equipment relative to its USA level, derived by the ICP using a similar basket of products across countries. The relative price is the price of machinery and equipment relative to the price of consumption. GDP per capita is expressed relative to its USA level. All variables are in natural logarithms. AEs advanced economies; EMs emerging market economies; LICs low-income countries; PPP purchasing power parity.

As established by earlier studies (see, for example, Hsieh and Klenow 2007), the difference between advanced economies and lower-income countries is particularly striking for the relative price of machinery and equipment (i.e., relative to the countries’ consumption price level). The relative price of machinery and equipment in the median low-income country is 2.7 times the price in the median advanced economy (Fig. 5, panel 2).

3 Drivers of Relative Investment Prices

In this section, we examine empirically the key sources of differences in the relative price of tradable capital goods across countries, and the factors underpinning the dramatic declines in the relative price of machinery and equipment over time.

In theory, the relative price of capital goods is shaped by several factors. Of prime importance is the efficiency with which an economy can produce machinery and equipment (or other tradable goods that it can exchange for investment goods) compared with the efficiency in other sectors.⁹ But that is not the only factor. In countries that import a significant fraction of investment goods (as in most emerging market and developing economies), the relative price of machinery and equipment also reflects prices that international suppliers charge for these goods, as well as other factors that drive a wedge between international and domestic prices. These include various trade costs, such as transportation costs, the efficiency of the domestic distribution sector, import tariffs, customs regulations, and the time and costs associated with the logistics of importing goods.¹⁰ We proceed to shed light on the importance of these factors in explaining both the cross-country heterogeneity in relative capital goods prices and their evolution over time.

3.1 Cross-Country Analysis

Determining which factors explain the observed differences in the absolute and relative prices of tradable capital goods in the 2011 ICP data is a daunting task. Because price levels of capital goods that are comparable across countries are available only at one point in time, it is difficult to disentangle the causal contribution of various potential drivers. We examine each potential source of differences in capital goods prices across countries—namely trade costs, and relative efficiency in the production of tradable goods—and relate these to the relative price of capital goods from the 2011 ICP data.

Trade costs tend to be much lower for advanced economies.¹¹ Despite significant progress in liberalizing the international exchange of goods and services and reducing trade costs, emerging markets, and especially low-income developing countries, still face significantly higher policy-related barriers to trade than advanced economies, in addition to their larger natural trade barriers (Fig. 6). They tend to be located farther from key capital goods exporters, measured as the weighted average of a country’s distance to all other countries, where the weights are the share of capital good exports. They are also less connected to global shipping networks based on the UNCTAD index which captures the number of ships, their container-carrying capacity, maximum vessel size, number of services, and number of companies that deploy container ships in a country’s port.

Fig. 6

Sources: CEPII, GeoDist database; Eora MRIO database; Feenstra and Romalis (2014); Fraser Institute; United Nations Conference on Trade and Development (UNCTAD); World Bank, Doing Business Indicators; and authors’ calculations

Trade costs in 2011. Note: Distance to exporters of machinery and equipment (M&E) is calculated as the weighted average of a country’s distance to all other countries, where the weights are equal to the partner countries’ exports of capital goods as a share of global capital goods exports. The UNCTAD liner shipping connectivity index captures how well countries are connected to global shipping networks based on five components of the maritime transport sector: number of ships, their container-carrying capacity, maximum vessel size, number of services, and number of companies that deploy container ships in a country’s port. The Fraser Institute’s Freedom to Trade Internationally index is based on four different types of trade restrictions: tariffs, quotas, hidden administrative restraints, and controls on exchange rate and the movement of capital. The cost and time indicators measure the cost (excluding tariffs) and time associated with three sets of procedures—documentary compliance, border compliance, and domestic transport—within the overall process of importing a shipment of goods. AEs advanced economies; EMs emerging market economies; LICs low-income countries.

EMDEs also impose significantly higher tariffs on imports of capital goods, and the time and cost associated with the logistics of importing goods—such as documentary and border compliance and domestic transportation—are substantially higher according to the World Bank Doing Business Indicators.

To understand whether countries with higher trade costs also exhibit a larger absolute price of capital goods we estimate the following regression:

$${ \ln }\left( {P_{\text{I}} } \right)_{\text{i}} = \alpha + \beta \cdot { \ln }\left( {{\text{Trade }}\;{\text{Cost}}} \right)_{\text{i}} + \epsilon_{\text{i}}$$

(1)

where $P_{I}$ is the absolute price of capital goods from the 2011 ICP data of country i in 2011 and ${\text{Trade}}\; {\text{Cost }}$ is either the distance to machinery and equipment exporters, the liner shipping connectivity, the freedom to trade internationally, the tariffs on capital goods, or the costs/time to import. In Fig. 7, panel 1 shows the coefficients of the above regression multiplied by one standard deviation increase in the measures of trade costs, for comparability across different types of trade costs. We find that countries with higher trade costs tend to have higher absolute prices of machinery and equipment.

Fig. 7

Source: Authors’ calculations

Trade costs, relative productivity, and the price of capital goods in 2011. Note: Panel 1 depicts the percent change in the 2011 International Comparison Program (ICP) absolute price of machinery and equipment associated with a one standard deviation increase in alternative measures of trade costs. In panel 2, the cross-country variation in the 2011 ICP price of machinery and equipment relative to consumption is decomposed into the share explained by differences in the labor productivity in the tradable goods sectors relative to the non-tradable goods sectors, and alternative measures of trade costs. M&E = machinery and equipment. ***p < 0.01; **p < 0.05; *p < 0.1.

Putting together the two key contending explanations of the cross-country dispersion in relative capital goods prices, namely trade costs and relative productivity differences, we examine their contribution to the cross-country variation in relative prices of capital goods. We estimate the following regression:

$${ \ln }\left( {\frac{{P_{\text{I}} }}{{P_{\text{C}} }}} \right)_{\text{i}} = \alpha + \beta \cdot { \ln }\left( {\frac{{a_{\text{T}} }}{{a_{\text{NT}} }}} \right)_{\text{i}} + \gamma \cdot { \ln }\left( {{\text{Trade }}\;{\text{Cost}}} \right)_{\text{i}} + \epsilon_{\text{i}}$$

(2)

where $\frac{{P_{\text{I}} }}{{P_{\text{C}} }}$ is the relative price of machinery and equipment (using ICP 2011 data), ${\text{Trade Cost}}$ is the different measures of trade costs as described above and $\frac{{a_{\text{T}} }}{{a_{\text{NT}} }}$ is the labor productivity in the tradable goods producing sector relative to the labor productivity in the non-tradable goods producing sectors. Labor productivity is calculated as the real value added in the relevant sector, divided by the number of employees in the sector. We rely on the EORA global database to obtain nominal values of the relevant variables across a wide sample of countries and construct the prices of tradables and non-tradables using the 2011 ICP data. In a second step, we use these regression estimates to decompose the cross-country variation in the log of relative prices into the variance that can be explained by the relative productivity measure versus trade costs. Given the cross-sectional nature of the data, this analysis is purely illustrative. As elaborated in the next section, relative productivity and trade costs are not independent of one another, complicating the interpretation of their estimated contribution to the variation in relative prices. The relative productivity in the tradable goods sector may be affected by trade barriers, as production of tradable goods likely relies on imported inputs. Furthermore, policy-related trade barriers may be erected with the goal of protecting low-productivity tradable goods sectors.

With these caveats in mind, Fig. 7, panel 2 shows that both relative productivity differences and trade costs are important in explaining the cross-country heterogeneity in relative prices. Together, relative productivity differences in the production of tradable goods and trade costs can explain up to 60 percent of the cross-country variation in the relative price of machinery and equipment, depending on which measure of trade cost is used.¹² Interestingly, policy-related trade barriers, such as tariffs and cost and time of importing, are a more powerful predictor of relative prices than natural barriers to trade such as distance and connectivity. While causal interpretation is difficult in the cross-country setting and in light of the likely relationship between relative productivity and trade barriers, these findings are consistent with the idea that the relative prices of capital goods are higher in emerging market and developing economies in part due to their higher trade barriers.

3.2 Time-Series Analysis

Prior studies have primarily focused on the cross-country variation in relative capital goods prices [for example, Hsieh and Klenow (2007), and Sposi (2015)], as they explored the roles of trade and productivity. We aim to shed light on the drivers of the big declines in the relative prices of tradable capital goods over time. We show that differences in the rate of trade integration and relative productivity growth within countries over time can lead to large variations in relative prices.

We use sectoral producer price data across 33 sectors and 40 advanced and emerging market economies during 1995–2011 from the Socioeconomic Accounts of the World Input-Output Database. This allows us to control for all factors that affect prices equally across sectors within a country in a particular year (such as exchange rate fluctuations and policies, commodity price changes, aggregate demand and productivity shocks, and the like) and all time-invariant differences in prices across countries and sectors.¹³

This approach faces two challenges. First, trade integration, in the sense of more market access for foreign producers (as measured by the ratio of imports to domestic sectoral value-added) fosters competition, inducing domestic producers to reduce markups of prices over marginal costs.¹⁴ In practice, the feedback from higher domestic prices to greater ability of foreign producers to gain market share complicates the interpretation of the estimated relationship between the two variables. To overcome this challenge, the analysis uses import tariffs as an instrument for exposure to trade, thus isolating changes in import penetration that were triggered by policy choice, rather than those driven by changes in domestic prices.¹⁵ Second, exposure to foreign competition affects relative domestic prices indirectly, through its impact on sectoral labor productivity as documented in numerous studies (see, for example, Ahn et al. 2019; Amiti and Konings 2007; Topalova and Khandelwal 2011). Thus, simply applying the elasticities estimated in the regression in the first step will understate the contribution of trade to producer price changes. To correct for this, we quantify the changes in labor productivity that can be attributed to changes in import penetration, and, in the second step, distinguish the contribution of trade-related changes in labor productivity from changes in productivity due to other factors (such as sectoral technological advances) to the decline in the relative price of machinery and equipment.

3.2.1 Regression Framework

We estimate two separate regressions to understand the contributions of global integration and productivity growth to the decline in the relative price of machinery and equipment in the past decades.

First, we estimate the sensitivity of relative producer prices at the sector level to changes in relative labor productivity and import penetration, using the following equation:

$$\begin{aligned} \ln \left( {\frac{{P_{i,j,t} }}{{\bar{P}_{i,t} }}} \right) & = \alpha_{i,j} + \mu_{i,t} + \left( {\beta + \phi \times 1_{{\left\{ {j \,{\text{is}}\, {\text{a}}\, {\text{capital}}\, {\text{goods }}\,{\text{producing }}\,{\text{sector}}} \right\} }} } \right) \\ & \quad \times \left[ {\ln \left( {\frac{{M_{i,j,t} }}{{{\text{VA}}_{i,j,t} }}} \right) - \ln \left( {\frac{{\bar{M}_{i,t} }}{{\overline{\text{VA}}_{i,t} }}} \right)} \right] + \gamma \ln \left( {\frac{{{\text{LP}}_{i,j,t} }}{{\overline{\text{LP}}_{i,t} }}} \right) + \varepsilon_{i,j,t} , \\ \end{aligned}$$

(3)

where $\frac{{P_{i,j,t} }}{{\bar{P}_{i,t} }}$ is the relative price of sector j in country i at time t; $\alpha_{i,j}$ denotes country–sector fixed effects; $\mu_{i,t}$ denotes country–year fixed effects; $\ln \left( {\frac{{M_{i,j,t} }}{{{\text{VA}}_{i,j,t} }}} \right) - \ln \left( {\frac{{\bar{M}_{i,t} }}{{\overline{\text{VA}}_{i,t} }}} \right)$ is the relative import penetration (measured as imports divided by value-added); and $\frac{{{\text{LP}}_{i,j,t} }}{{\overline{\text{LP}}_{i,t} }}$ is the relative productivity of labor (measured as real value-added per employee).¹⁶

The relative import penetration, $\ln \left( {\frac{{M_{i,j,t} }}{{{\text{VA}}_{i,j,t} }}} \right) - \ln \left( {\frac{{\bar{M}_{i,t} }}{{\overline{\text{VA}}_{i,t} }}} \right)$ is instrumented by relative import tariff, defined as $\tau_{i,j,t} - \bar{\tau }_{i,t}$, with $\bar{\tau }_{i,t}$ defined as $\frac{{\mathop \sum \nolimits_{j = 1}^{J} \tau_{i,j,t} }}{J}$ and $\tau_{i,j,t}$ is defined as

$$\tau_{i,j,t} = \frac{{\mathop \sum \nolimits_{{l \in \varLambda_{j} }}^{{}} m_{i,k,l,t} \hat{\tau }_{i,k,l,t} }}{{\mathop \sum \nolimits_{{l \in \varLambda_{j} }}^{{}} m_{i,k,l,t} }},$$

in which $m_{i,k,l,t}$ is the import of country $i$ from country $k$ in sector l at time t, and $\hat{\tau }_{i,k,l,t}$ is the tariff imposed on these imports. $\hat{\tau }_{i,k,l,t}$ comes from the SITC 4-digit level bilateral preferential tariff data compiled by Feenstra and Romalis (2014).

Second, we estimate the impact of trade liberalization on relative labor productivity through the following equation:

$$\ln \left( {\frac{{{\text{LP}}_{i,j,t} }}{{\overline{\text{LP}}_{i,t} }}} \right) = \alpha_{i,j}^{\text{LP}} + \mu_{i,t}^{\text{LP}} + \left( {\beta^{\text{LP}} + \phi^{\text{LP}} \times 1_{{\left\{ {j \,{\text{is }}\,{\text{a }}\,{\text{capital}}\, {\text{goods }}\,{\text{producing}}\, {\text{sector}}} \right\} }} } \right) \times \left[ {\ln \left( {\frac{{M_{i,j,t} }}{{{\text{VA}}_{i,j,t} }}} \right) - \ln \left( {\frac{{\bar{M}_{i,t} }}{{\overline{\text{VA}}_{i,t} }}} \right)} \right] + \varepsilon_{i,j,t}^{\text{LP}} ,$$

where $\ln \left( {\frac{{M_{i,j,t} }}{{{\text{VA}}_{i,j,t} }}} \right) - \ln \left( {\frac{{\bar{M}_{i,t} }}{{\overline{\text{VA}}_{i,t} }}} \right)$ is also instrumented by relative import tariff, due to the concern of reverse causality: if a country’s capital goods producing sector becomes more productive, it may import less machinery and equipment from abroad. The estimation results indeed confirm the need to address this endogeneity issue: the OLS coefficient is much smaller than the estimate obtained using the instrumental variable.

Import tariffs are assumed to satisfy the exogeneity conditions:

$$\text{cov} \left( {\tau_{i,j,t} - \bar{\tau }_{i,t} ,\varepsilon_{i,j,t} } \right) = \text{cov} \left( {\tau_{i,j,t} - \bar{\tau }_{i,t} ,\varepsilon_{i,j,t}^{\text{LP}} } \right) = 0 .$$

Table 1 lists the mean, standard deviation, 25th percentile, and 75th percentile of cross-sectional distribution of import tariff across country–sector pairs.¹⁷ The average tariff declined from 2.69 to 0.91 percentage points in advanced economies, and from 6.07 to 3.84 percentage points in emerging market and developing economies between 1995 and 2011. While such changes in import tariff are not large, there is significant variation across sectors as depicted in Appendix Fig. 11.

Table 1 Summary statistics of output tariff by country group and year

In Table 2, column 2, we present the results from the first stage, the relationship between import tariffs and import penetration. As expected, country–sectors which experience larger cuts in import tariffs have higher import penetration. Column (1) of the same table suggests the robustness of the relationship between import tariff and import penetration by dropping relative productivity from the regression. Columns (3) and (4) show the reduced form relationship between the instrument and the dependent variable of interest, namely relative producer prices. There is a strong negative link between import tariffs and producer prices, whether controlling for labor productivity or not.

Table 2 First-stage relationship, effects of import tariff on producer prices.

Tables 3 and 4 report the main estimation results. A general pattern that emerges is that deeper import penetration increases the relative labor productivity of a domestic sector. It also reduces producer prices directly. As robustness tests, we allow $\beta$, $\gamma$ and $\beta^{\text{LP}}$ to differ across advanced economies and emerging market and developing economies, and the results are broadly the same.

Table 3 Labor productivity and trade integration.

Table 4 Relative producer prices, trade integration and relative productivity.

The impact of import tariffs on import penetration and the association between import penetration and producer prices are economically significant. A 1 percent increase in the import ratio, which can be achieved by a 0.7 percentage point cut in tariffs, reduces the sectoral producer price by about 0.6 percent. Changes in labor productivity also have a significant impact on producer prices, with a 1 percent increase in sectoral labor productivity reducing producer prices by about 0.3 percent.

Moreover, labor productivity of the capital goods producing sector is particularly sensitive to deepening trade integration, a finding consistent with the larger reliance on global value chains for the production of these goods.¹⁸

It is worth highlighting that our findings on the role of import penetration remain qualitatively and quantitatively similar if we do not control for relative productivity in the regression analysis, for which finding an appropriate instrument is difficult (see Table 2 and Appendix Table 13). In column (8) of Table 4, we explore this issue by using the lagged value of relative productivity as an instrument for relative productivity and find broadly similar findings.

Figure 8 decomposes the average decline in the relative price of the machinery and equipment producing sectors between 2000 and 2011 into four parts: (1) the direct effect of deepening trade integration; (2) the effect of trade integration through higher labor productivity; (3) the effect of higher labor productivity, which is not due to deepening trade integration; and (4) a residual. Rising trade integration accounts for the bulk of the decline in relative prices of machinery and equipment, both through its direct effect on producer prices and indirectly, through higher labor productivity of domestic capital goods producers. Productivity gains in the capital goods sector, which cannot be directly linked to trade integration, were also a significant factor.^19,20 We also explore whether tariff affects producer prices by changing input costs. Appendix Table 14 suggests that this channel does not play a significant role on average in our sample.

Fig. 8

Source: Authors’ calculations

Contributions to changes in relative producer prices of capital goods: 2000–2011. Note: The figure combines the estimated elasticities of producer prices to trade integration and relative labor productivity, and changes in these factors for the capital goods sector between 2000 and 2011 to compute their contribution to the observed change in the producer price of capital goods relative to the price of consumption. AEs advanced economies; EMDEs emerging market and developing economies.

4 Relative Price of Investment Goods and Real Investment Rate

This section aims to quantify the impact of relative investment prices on real investment rate. In particular, how much does the relative price of capital goods matter for a country’s investment rate? What share of the dramatic increase in machinery and equipment investment over the past 60 years can be attributed to the decline in the relative price of these goods? To answer these questions, we focus on medium-term changes in investment goods prices and its link with real investment rate.

The analysis in this section assumes that the relative price of investment moves due to supply shocks, rather than demand shocks.²¹ We view international trade and technological progress as the key drivers of these supply shocks, as discussed in the previous section. Since our focus here is on explaining long-run changes in real investment, supply shocks are the likely relevant drivers, given well-known evidence from Blanchard and Quah (1989) that supply shocks are more persistent over time than demand shocks. We also provide evidence from an instrumental variable approach in which relative investment prices are instrumented by relative import tariffs.

4.1 Cross-Country Empirical Evidence

The empirical framework used to assess the role of relative investment prices for investment-to-GDP ratios is inspired by the reduced form relationship that can be derived from a number of theoretical papers, such as Restuccia and Urrutia (2001) and Sarel (1995). The general intuition from these models is that a shock that leads to a decline in the relative price of investment, such as productivity increase in the capital goods sector or a decline in capital goods tariffs, would raise the optimal (steady-state) level of capital stock as a share of output. Because a higher level of capital stock needs to be maintained, real investment would rise permanently as a share of real output in order to keep up with capital stock’s depreciation.

The general regression relates the log of the real investment-to-GDP ratio in machinery and equipment and the log of the price of machinery equipment relative to the price of consumption,²² controlling for all time-invariant differences across countries (µ_i) and period fixed effects (θ_t) to capture common global shocks:

$$\ln \left( {\frac{{{\text{Real}}\, {{{\text{M}\&\text{E}}}}\,{\text{Investment}}}}{{{\text{Real }}\,{\text{GDP}}}}} \right)_{i,t} = \beta \cdot \ln \left( {\frac{{P_{{{{\rm M}\&\text{E}}}} }}{{P_{{\rm C}} }}} \right)_{i,t} + {\text{ Controls}}_{i,t} + \mu_{i} + \theta_{t} + \varepsilon_{i,t} .$$

The regressions are estimated on five-year non-overlapping window averaged data. This approach aims to smooth the influence of short-term fluctuations, and to capture the potential medium-run relationship.

Based on empirical studies of the long-run determinants of the aggregate investment rates,²³ the set of additional controls includes lagged level and growth rate of real GDP per capita in purchasing-power-parity terms to account for possible convergence effect, lagged dependent variable to account for persistence in investment rates, availability and cost of finance (proxied by real interest rates and credit-to-GDP ratio). For comparability across columns with and without controls, we impute missing values in the control variables using a mean and include a dummy that takes the value of one when the data is missing for each of the control variables. The choice of control variables is primarily aimed at attenuating potential omitted variable bias. The full list of data sources can be found in “Appendix 1.”

Estimation results, based on OLS and IV regressions, are reported in Table 5. The estimates are robust to alternative specifications, choice of subsamples, and estimation methods. In the instrumental variable (IV) regressions, three alternative instruments are used. First, the relative price is instrumented using its own lag. This strategy allows to minimize the bias (toward finding a negative relationship) stemming from the potential negative correlation in the measurement errors of real investment and its price, under the assumption that measurement error is unlikely to be correlated over time.²⁴ Second, the relative price is instrumented with the average relative price of all other countries except the country’s own. This approach allows to isolate technologically driven changes in the relative price from those that may occur due to changes in demand for investment goods within a country, thus minimizing the measurement error bias as measurement error in a country’s own prices is unlikely to be correlated with measurement error in other countries’ prices. Third, the relative price is instrumented using log of import tariffs on capital goods relative to tariffs on consumption goods. The last two columns of Table 5 are estimated excluding major capital goods exporting countries and excluding commodity exporting countries to account for the possible effect of commodity price booms on investment rates.

Table 5 Real investment rate and relative price of machinery and equipment: country level.

Across specifications, a 1 percent decline in the relative prices of tradable capital goods is associated with a 0.2–0.9 percent increase in the real investment rate over a five-year period. It is important to note that these empirical estimates likely represent an upper bound of the true effect of changes in relative price on real investment rates. As discussed above, relative investment prices are endogenous and reflect many factors, including changes in policies that could have a direct impact on investment rates.

4.2 Sectoral Empirical Evidence

A sectoral perspective can complement the cross-country analysis in an important way. The relative price of capital goods is but one of the considerations that shape investment decisions. While the cross-country analysis attempts to control for many factors, the estimated relationship between real investment rates and prices could be biased due to the omission of factors that may correlate with relative prices but are not properly captured in the estimation. Sectoral analysis allows us to isolate the relationship between real investment rates and the price of investment across different sectors while properly accounting for the role of all factors that affect investment within a country in a particular year. These include financial conditions, economy-wide growth prospects, quality of regulations that affect investment returns, exchange rate fluctuations and policies, international capital flows, availability of complementary public infrastructure, and the like.

The data come from EU KLEMS and World KLEMS, which offers detailed information about the price level of different types of capital goods within Machinery and Equipment: IT (computer hardware), CT (telecommunications equipment), transport equipment, and other machinery and equipment. The price of machinery and equipment, $P_{{{\text{M}\&\text{E}}}}$, is constructed as a weighted average of the prices of each of the four types of capital, as in the equation below.

$$P_{{{\text{M}\&\text{E}}}} = \frac{{I_{\text{IT}} }}{{I_{{{\text{M}\&\text{E}}}} }}P_{\text{IT}} + \frac{{I_{\text{CT}} }}{{I_{{{\text{M}\&\text{E}}}} }}P_{\text{CT}} + \frac{{I_{\text{TraEq}} }}{{I_{{{\text{M}\&\text{E}}}} }}P_{\text{TraEq}} + \frac{{I_{\text{OMach}} }}{{I_{{{\text{M}\&\text{E}}}} }}P_{\text{OMach}} .$$

The sample varies somewhat depending on the specification and data availability for specific variables. Typically, the analysis relies on 18–19 countries, mostly European, with the addition of USA, UK, Brazil and Colombia, and uses 15 broad sectors, covering the period 1971–2015. This is an unbalanced panel.

The baseline specification mirrors that of country-level regressions, using 5-year averaged data, which is common in the literature when looking at long-term, slow-moving factors. In the main specification, the log relative price of investment (expressed relative to the price of consumption) is instrumented with is lagged value.

$$\ln \left( {\frac{{{\text{Real }}\,{\text{M}}\& {\text{E}}\,{\text{Investment}}}}{{{\text{Real }}\,{\text{VA}}}}} \right)_{i,t,s} = \beta \cdot \ln \left( {\frac{{P_{{{\text{M}\&\text{E}}}} }}{{P_{{\rm C}} }}} \right)_{i,t,s} + \gamma \cdot \ln \left( {\frac{{{\text{Real }}\,{\text{M}}\& {\text{E}}\,{\text{Investment}}}}{{{\text{Real }}\,{\text{VA}}}}} \right)_{i,t - 1,s} + \mu_{i,t} + \theta_{i,s} + \varepsilon_{i,t,s}$$

A range of possible estimates using slightly different specifications are presented in Table 6. The estimated elasticity, according to which a 1 percent decline in the relative price of machinery and equipment investment is associated with a 0.2–0.5 percent increase in real investment in these capital goods, is comparable to those uncovered in the cross-country analysis. The baseline specification includes country-period and country–sector fixed effects, where the period refers to five-year non-overlapping periods. However, country-period fixed effects may absorb too much variation, for example, if there is an aggregate effect of the relative price of investment that is common to all sectors within a country–year. For that reason, an alternative specification includes country–sector and period (or year) fixed effects, where this problem is addressed (columns 5–8, Table 6).

Table 6 Sectoral real investment rate and relative prices of machinery and equipment: range of possible estimates.

Table 7 presents the baseline results first with country-period and country–sector fixed effects (columns 1–4), followed by period and country–sector fixed effects (columns 5–8), for each of four dependent variables: the machinery and equipment investment rate, followed by machinery and equipment investment, value added, and output per worker.

Table 7 Relative prices of machinery and equipment and sectoral outcomes: five-year averages.

As a robustness check, Table 8 presents all the regressions presented in Table 7 but using annual data. As expected, the estimated coefficients are smaller in magnitude when annual data are used instead of 5-year averages. However, all the results have the correct signs and are statistically significant, except for sectoral output per worker. As an additional robustness, we ran all the regressions from Table 5 excluding the commodity sectors (agriculture, forestry and fishing; and mining and quarrying), and the estimated coefficients were all the same sign, statistical significance, and generally slightly larger magnitudes.

Table 8 Relative prices of machinery and equipment and sectoral outcomes: annual.

Across both country-level and sector-level regressions, the evidence that the relative price of capital goods matters for investment decisions is strong. It is challenging to obtain an unbiased estimate of the elasticity of real investment with regard to prices, given the endogenous nature of relative price changes and problems with measurement. With those difficulties in mind, Fig. 9—as a purely illustrative exercise—uses the estimated elasticity from the cross-country (instrumented variable) analysis and the post-1997 change in the relative price of capital goods in each country to decompose the change in real investment rate. The decomposition is based on Table 5, column 4, where the estimated coefficient on the relative price is the smallest, to be on the conservative side. These changes comprise the parts attributable to (1) the decline in real investment prices; (2) other factors, such as change in relevant policies, global trends in investment, convergence, and growth expectations; and (3) the residual. The figure confirms that the dramatic decline in the relative prices of tradable capital goods can explain a sizable share of the increase in investment in tradable capital goods in advanced and emerging market and developing economies.

Fig. 9

Source: Authors’ calculations

Contributions of relative prices to increases in real investment in machinery and equipment, 1997–2017. Note: The figure presents the contribution to the observed increase in real machinery and transport equipment investment-to-GDP ratios between 1997–2001 and 2017 from the relative price of machinery and transport equipment, other controls, and residual. Black square indicates the total change in real machinery equipment investment-to-GDP ratios. AEs advanced economies; EMDEs emerging market and developing economies.

5 Conclusion

Our analysis provides new evidence on the drivers of the relative price of machinery and equipment and its macroeconomic implications. Leveraging the dramatic changes in capital goods prices that have taken place over the past few decades across countries and sectors, and the latest available comparable cross-country data on prices from the ICP and detailed trade flow statistics, we provide evidence that trade costs and relative productivity both play an important role in shaping the relative prices of machinery and equipment across countries and over time. Across countries, those with higher trade costs and lower productivity in the tradable goods sectors tend to pay a higher price for a comparable basket of machinery and equipment in both absolute terms and relative to the price of consumption. Analysis of sector-level producer price data shows that, over time, reductions in distortionary trade policies and improvements in productivity both contributed to the decline in the relative prices of capital goods.

We also show that the decline in the relative price of capital goods has played a crucial role in increasing real investment rates over the past three decades. While exact quantification is challenging, empirical evidence suggests that a nontrivial share of the rise in real investment rates in machinery and equipment can be attributed to the dramatic fall in the relative price of these goods.

Taken together, our analyses suggest that the slowing pace of trade liberalization since the mid-2000s, and especially the possibility of its reversal in some advanced economies, could interfere with the tailwind to machinery and equipment investment generated by the falling price of capital goods. This finding provides an additional, often overlooked, argument in support of policies aimed at reducing trade costs and reinvigorating international trade.

Many emerging market and developing economies still maintain tariff and other trade barriers that significantly raise the relative price of investment paid by domestic investors.²⁵ Effective import tariffs on capital goods in 2011 were about 4 percent in emerging market and 8 percent in low-income developing countries, compared with close to zero in advanced economies. Fully implementing commitments under the World Trade Organization’s Trade Facilitation Agreement could mean a reduction in trade costs equivalent to a 15-percentage point tariff cut in less-developed economies (WTO 2015).

In advanced economies, avoiding protectionist measures and resolving disagreements without raising trade costs will be crucial to prevent further weakening of the lackluster investment growth since the global crisis of a decade ago.²⁶

Appendices

Appendix 1: Sources and Country Groupings

1.1 Data Sources

The primary data sources for this paper are the IMF World Economic Outlook database, the Penn World Table (PWT) 9.1 database, including supplemental datasets on national accounts and capital detail, the World Input-Output Database (WIOD) Release 2013 and 2016, including both Socio Economic Accounts and World Input-Output tables, and the EU and World KLEMS databases.

1.2 Data Definitions

Several sources of data on prices are used in the paper. The relative price of investment is defined relative to the price of consumption.

The cross-country stylized facts on relative prices and the associated analysis rely on the International Comparison Program (ICP) 2011, which provides the price level of machinery and equipment and the price level of consumption measured for a comparable basket of goods across countries in 2011.

The stylized facts presented in Figs. 1 and 3 and country-level panel regressions use data from the PWT 9.1 capital detail dataset, which provides data on deflators of various types of investment, and capital stocks. The corresponding consumption deflator comes from the PWT 9.1 National Accounts dataset.

The sector-level panel regressions, which examine the relationship between investment in machinery and equipment and its relative price, use data from the EU and World KLEMS databases. The relative price of investment is likewise defined as the ratio of deflators, in this case the machinery and equipment deflator and the country-wide consumption deflator.

The sector-level panel regressions, which examine the drivers of sectoral producer prices, rely on the sectoral gross output deflator from the WIOD Socio Economic Accounts database (Table 9).

Table 9 Data sources.

1.3 Country Groupings

The definition of advanced economies, emerging market economies, and low-income countries follows the October 2018 IMF World Economic Outlook’s definition.

Tradable capital goods sectors, which, for the purpose of this paper, include machinery and equipment and transport equipment, are identified in the following manner across data sources. In the WIOD database, sectors 400, 410, and 521 are considered capital goods producing sectors. In the Eora MRIO database, sectors 9 and 10 are considered capital goods producing sectors (Table 10).

Table 10 Sample of economies included in the analytical exercises

Appendix 2: Comparison of Absolute and Relative Prices of Machinery and Equipment in 1996 and 2011 ICP Vintages

Using the most recent, 2011, vintage of International Comparison Program (ICP) data on prices across countries, we find that the price of producer durables (machinery and equipment) is negatively related to GDP per capita—opposite of Hsieh and Klenow (2007) findings with the 1996 ICP vintage.

As shown in Appendix Fig. 10, panels 1 and 2, the relationship between the absolute price level of machinery and equipment and log GDP per capita is slightly positive and not statistically significant in 1996, but is slightly negative and statistically significant in 2011. Notably, Hsieh and Klenow (2007) report a likewise negative and statistically significant relationship using earlier 1985 ICP vintage (which remains negative and significant using black market exchange rates) and a negative, but not significant, relationship using the 1980 ICP vintage (which becomes positive and significant using black market exchange rates).

Fig. 10

Sources: International comparison program (ICP) 1996 and 2011; and authors’ calculations

Price levels across countries. Note: The absolute price of machinery and equipment is the price level of machinery and equipment relative to its USA level, derived by the ICP using a similar basket of products across countries. The relative price is the price of machinery and equipment relative to the price of consumption. GDP per capita is expressed relative to its USA level. All variables are in natural logarithms. AEs advanced economies; EMs emerging market economies; LICs low-income countries; PPP purchasing power parity.

Currently, given better coverage of low-income countries and the methodological improvements described in Feenstra et al. (2015), the 2011 vintage is considered to be of the highest quality available for cross-country comparison.

In contrast to changing relationship for the absolute price level of machinery and equipment, the relationship between log GPD per capita and the price level of consumption is consistently positive and strongly statistically significant across ICP vintages (panels 3 and 4). In combination, the price of machinery and equipment relative to the price of consumption (panels 5 and 6) is negative and statistically significant across both vintages.

Appendix 3: Drivers of Relative Investment Prices: Across Countries

This Appendix section provides technical details on the analysis, which compares the level of capital goods prices across countries. The analysis relies on the ICP 2011 data, which provides the price level of comparable baskets of capital goods for 168 countries. The ICP reports absolute prices as a ratio to the corresponding US prices. When analyzing relative capital goods prices, the absolute price of machinery and equipment is divided by the absolute consumption price.

To establish if there is correlation between absolute prices and various measures of trade cost, we estimate the following equation using ordinary least squares, with standard errors adjusted for heteroskedasticity.

$${ \ln }\left( {P_{I} } \right)_{\text{i}} = \alpha + \beta \cdot \ln \left( {{\text{Trade}}\,{\text{Cost}}} \right)_{\text{i}} + \epsilon_{\text{i}}$$

where $P_{I}$ is the absolute price of machinery and equipment in country i in 2011. A separate regression is estimated for each measure of trade costs.

We consider the following measures of trade costs: (1) distance to exporters of capital goods, calculated as the weighted average of a country’s distance to all other countries, where the weights are equal to the partner countries’ exports of capital goods as a share of global capital goods exports; (2) the UNCTAD liner shipping connectivity index, which captures how well countries are connected to global shipping networks based on five components of the maritime transport sector: number of ships, their container-carrying capacity, maximum vessel size, number of services, and number of companies that deploy container ships in a country’s port; (3) the Fraser Institute’s Freedom to Trade Internationally, which is based on four different types of trade restrictions: tariffs, quotas, hidden administrative restrains, and controls on exchange rate and the movement on capital; (4) the average applied tariffs on capital goods imports, from Feenstra and Romalis (2014); (5) the cost to import and time to import indicators, which measure the cost (excluding tariffs) and time associated with three sets of procedures—documentary compliance, border compliance, and domestic transport—within the overall process of importing a shipment of goods from the World Bank, Doing Business Indicators.

Appendix Table 11 provides the estimated coefficients as well as the percent change in absolute prices associated with a one standard deviation change in the alternative measures of trade costs.

Table 11 Absolute price of capital goods.

When examining the determinants of relative prices in the cross section of countries, we estimate the following equation using ordinary least squares, with standard errors adjusted for heteroskedasticity.

$${ \ln }\left( {\frac{{P_{I} }}{{P_{C} }}} \right)_{\text{i}} = {\alpha } + {\beta }\cdot{ \ln }\left( {\frac{{a_{T} }}{{a_{NT} }}} \right)_{\text{i}} + {\gamma }\cdot{ \ln }\left( {\text{TradeCost}} \right)_{\text{i}} + \epsilon_{\text{i}}$$

The trade costs considered (one at a time) are the same as discussed above. Labor productivity is measured as the ratio of the value added of the tradable goods producing sectors divided by the total employment in those sectors, and the value added of all non-tradable sectors in the economy divided by their employment. This measure is constructed using 2011 data from the Eora MRIO database and adjusted using 2011 ICP prices to make productivity levels comparable across countries.

Appendix Table 12 provides the estimated coefficients. The regression-based decomposition is based on Shorrocks (1982). The contribution of each variable is calculated as the covariance between the (1) product of the estimated coefficient and the value of the independent variable and (2) the dependent variable, divided by the variance of the dependent variable (Tables 11 and 12).

Table 12 Relative price of capital goods.

Appendix 4: Additional Stylized Facts and Robustness Tests

Appendix Fig. 11 shows that over-time change in tariff varies significantly across sectors. While tariff barely changed between 1995 and 2011 in a large fraction of country–sector pairs, its decline was greater than 5 percent in other cases.

Fig. 11

Density distribution of changes in import tariff from 1995 to 2011

Appendix Table 13 reports the estimates of a variant of Eq. (3) that excludes relative labor productivity, and the instrumental variable for import penetration is still import tariff.

Table 13 Relative producer prices and trade integration.

Appendix Table 14 reports regressions in which we regress relative producer price on input tariff, which, for sector i, is defined as the weighted average of tariffs on imported goods used by sector i as inputs. The weight for imported goods used by sector i and produced by sector j in other countries is the share of these goods in the value of imported goods used by sector i as inputs (Tables 13 and 14).

Table 14 Effects of input tariff on producer prices.