RELATIONSHIP BETWEEN TRANSPORT INFRASTRUCTURE INVESTMENT AND PRODUCTIVITY: EVIDENCE FROM OECD COUNTRIES
ABSTRACT
KEY WORDS: infrastructure investment, economic growth, productivity, total factor productivity (TFP), multifactor productivity, panel regression, cointegration, causality.
1. Introduction
Infrastructure is regarded as the veins of a certain economy, connecting uses and sources together. It requires large amounts of investments, financed by public and/ or private resources. Enhancement of infrastructure network reduces production costs, increasing efficiency, improves market accessibility. It creates employment and impact the remaining economy by means of multipliers. Infrastructure is crucial to the location and scale of economic activity, and patterns of agglomeration and specialization (Baldwin et al. 2003; Krugman, 1991).
The relationship between infrastructure, production and growth attracted attention especially starting from 1980s. Aschauer (1989 a, b, c) were the pioneering studies, which found a strong link between core infrastructure and the productivity. These studies triggered complementary studies as well as critics. The critics were mainly related to 1) how to interpret the correlation (Fernald (1999)), 2) the direction of causality; 3) methodology issues, such as production function and aggregate data used, simultaneous equations bias –i.e. two variables on both sides of the regression equation influence each other at the same time-; 4) how to model the complex network characteristics of infrastructure investment (Hulten, C. R., & Schwab, R. M. (1991). Public capital is endogenous; as such productivity causes public investment rather than the other way round, when it is assumed that infrastructure investments are performed solely by state. Izquierdo (2005) supported the findings of Aschauer and affirmed that in the long-run, production capacity of an economy is influenced through increased capital stock and improved productivity.
Evans and Karras (1994) and Holtz-Eakin (1994) brought one more dimension to the discussion and affirmed that strong correlation illustrated by Aschauer reflect how increased income causes a higher government activity, rather than the effect of the latter to private productivity. Their findings do not contradict with the finding that infrastructure is productive, either. The link between public infrastructure and economic performance is affirmed to be weak. Contrary to these, Pritchett (1996) outlines that especially in the developing countries public investments may not create productive capital.
One of the reasons which make it difficult to establish causality relationship relatively difficult is that infrastructure investment tends to be pro-cyclical. This might be due to having more room for investments during economic expansions (Holmgren, Merkel, 2017).
Most of the literature focuses on making a direct link between infrastructure investment and economic growth. Straub (2008) affirm that using economic output as the regressant tend to find more positive effects of infrastructure than those using productivity or growth rates and such may imply that transitory effects are observed more frequently. Cross-country analysis is relatively scarce in the literature, which mainly focuses on regional and sectoral aspects.This article attemps to fill this gap by directly observing infrastructure and productivity in a cross-country setting.
We employ a Pedroni co-integration and Granger causality methods as well as dynamic ordinary least-squares (DOLS) in order to establish whether a long-term relationship exists and the direction of the relationship.
The following sections are presented in the paper: In the section two, theoretical references are provided. In the section three, research methods are presented. In the section four, the empirical results are illustrated. In the section five, we conclude.
2. Theoretical References
Among the earlier studies on the linkage between infrastructure and productivity, Hurd (1975) analyzed the behavior of prices of food grains in India in relation to railway expansion during 1861-1921. The result of the analysis is that prices converged to each other in different districts of India along with the extension of the railway network. Consequently, railway expansion in the country promoted the expansion of the food grains market.
In the subsequent decade, Aschauer (1989a, 1989b) observed that decline in the infrastructure investment was followed some time afterwards by sluggish aggregate productivity in the US and some developed countries in the 1970s. Aschauer (1989a) show that core infrastructure explains the productivity and that such capital stock is dramatically more significant to derive productivity compared to flow of such spending. In the cross-national analysis by Aschauer (1989b), higher public capital raises the national investment rate above the level chosen by rational agents causing a crowding out effect on private investment. On the other hand, an increase in the public capital stock also raises the return to private capital, which crowds in private capital accumulation.
Shortly after, an economic growth cross-country model by Barro (1990) followed by Cronin, Parker, Colleran, & Gold (1991), Lynde & Richmond (1992) and Easterly & Rebelo (1993) supported the findings by Aschauer. The model by Canning and Pedroni (2008) based on Barro (1990) showed in general a short and long term bi-directional causality; while the tendency of infrastructure to stimulate long-run economic growth varies significantly across countries. The authors highlight that infrastructure capital is an input into aggregate production. However, they draw attention to the opportunity costs involved, namely reduced investment in other types of capital, implying an existence of an optimal level of infrastructure which maximizes the growth rate. Different than earlier works, they used physical measure of infrastructure, such as kilometers of paved roads, rather than making an estimate on the infrastructure stock from investment flows and presented the results in terms of type of infrastructure. They reasoned this on the fact that monetary investment may be a very poor guide to the amount of infrastructure capital produced (Pritchett, 1996). Straub (2008) note that the effect of infrastructure may depend on the level of economic growth.
Morrison and Schwartz (1996) used a cost based methodology in which public capital expenditures yield cost savings which supercede the investment costs. This is explained by the substitutability between public capital and private capital. The authors undeline that over the longer run incentivising investment in private capital improves economic performance more effectively than public capital expenditures alone. In a similar methodology, Demetriades and Mamuneas (2000) considered infrastructure as a cost-reducing technology, rather than an input in the production of final goods, promoting specialization. The outcome of the model was that infrastructure can stimulate long-run growth, though non-monotonic. Infrastructure accumulation is very productive when the rate of infrastructure accumulation is very low and counter-productive when high rates of accumulation.
Fernald (1999) supplies evidence on the impact of road investment on productivity. This is drawn from the assumption that if roads are leading to more productivity, then the vehicle-intensive industries benefit more from road-building. The latter was confirmed by the analysis. However, opposite effect is observed in non-vehicle-intensive industries. Sustainable high productivity could not be concluded from the data. A plausible interpretation was that the massive road-building of the 1950’s and 1960’s in the USA provided a one-time boost to the productivity, rather than a continuous one. Conversely, Duranton and Turner (2012) concluded that a 10% increase in a city´s stock of (interstate) highways causes about a 1.5% increase in its employment over 20 years.
In more recent studies, Baum-Snow, Brandt, Henderson, Turner, and Zhang (2015) found strong evidence that enhanced access to domestic markets stimulated regional GDP and income growth. Additionally, improved access to international markets promoted population growth.
Holmgren, Merkel, 2017 affirms that infrastructure investment is often seen as a solution to depopulation of rural areas. The effects of infrastructure appear to vary depending on the type of infrastructure with investments in air transport having the highest impact for agricultural and service sector. Road investments appear to have the highest impact on production in construction and manufacturing. Investments in port infrastructure have the highest impact on the agricultural sector.
Zhang, Ji (2018) analyses whether infrastructure has transitory effects by lifting the level of aggregate output or a longer-term impact by stimulating the growth rate of output on the data on China. Infrastructure stock (except railways) had a stronger impact compared to other capital stock in improving output. Other important conclusion is the non-linearity in infrastructure's productivity depending on whether there is relative shortage or overprovision of infrastructure. However, their model gives little evidence of infrastructure's positive impact on long-term growth rates and as such the effect of infrastructure appears to be principally transitory.
3. Variables
We considered the annual time series data between 1995-2016 in relation to a total of 48 countries from OECD online database, 36 of which are OECD members (OECD (2018), Multifactor productivity (indicator) and OECD (2018), Infrastructure investment (indicator)). The multifactor productivity residual measure is usually interpreted as the rate of change of the technical efficiency parameter in the production function framework ((Hulten, C. R., & Schwab, R. M. (1991).
The dependent variable is multifactor productivity (MFP) per country the measured as an index where 2010 is selected as base year (2010=100), which is also called total factor productivity (TFP), and independent variable is the percentage of transport infrastructure investments in GDP. We use the natural logarithm of TFP series.