Transportation–Communication Capital and Economic Growth: A VECM Analysis for Turkey

Abstract

This paper analyses the short- and long-term relationships between the transportation–communication capital and the output for Turkey. The study applies a Cobb–Douglas production function under the assumption of constant returns to scale and employs co-integration analysis by estimating a vector error correction model (VECM). As a result of the VECM estimation, one co-integrating relationship is detected. The results based on the impulse response function analysis imply that per labour transportation–communication capital appears both to have been a crucial input in the Turkish productive process and to have had a positive crowding in effect on the per labour non-residential total capital formation. Moreover, the results support the argument that the transportation–communication capital has a lagged impact on economic growth. The long-term accumulated elasticity of output to transportation–communication capital has been found to be 0.59. The long-term accumulated marginal product was also calculated. It implies that a 1 Turkish Lira increase in per labour transportation–communication capital results in a long-term rise of 1.45 Turkish Liras in per labour output. All these findings suggest that transportation–communication capital may be a powerful tool for policy-makers to promote long-term per labour real output growth in Turkey.

Acknowledgements

The authors would like to thank Assoc. Prof. Dr Ozan Eruygur very much for his valuable comments and assistance. He inspired and guided us at every step of the paper.

Notes

According to authors' own calculations based on SPO (1970–2008), this figure is 20.2% in average for the period of 1968–2006.

Due to the non-separated official data of TurkStat, we had to process transportation sector with communication sector under the name of transportation–communication sector.

For a critical survey and the other classification of the studies on this subject, see, Romp and De Haan (2007).

The initial stock of capital has been obtained using the standard perpetual inventory procedure following the seminal paper of Harberger (1978). This approach uses the neoclassical growth model prediction of a constant capital-output ratio over time and hence is based on the assumption that the country was at its steady-state capital-output. The procedure is built on the accumulation equation which implies that . The left side of the expression is the growth rate of capital stock, g. Hence, one can write . Neoclassical growth theory proposes that investment and capital grow at the same rate in the steady state. Thus, the growth rate of capital can be approximated by the growth rate of investment (Kamps, 2004, p. 14). In our study, following the approach of Kamps (2004), the growth rate of capital is approximated by the average growth rate of investment over the period 1963–2006. While the steady-state assumption of Harberger procedure is strong, it is probably better than assuming that an initial capital stock of zero (Beck et al., 1999). Setting the value of initial capital to zero might generate significant measurement error in application with short time series.

Recall that , where t = 0,1,2, …, T. Taking the natural logarithm of both sides yields . Rearranging we get . For the growth rate between the initial and last year, it is . A useful approximation is that for any small number g, . Hence one can write , which is nothing but the expression used in our study for the annual average growth rate of total fixed investment. Note that a log growth rate is a continuous rate of growth with continuous compounding. However, a percentage change is specific to a particular time horizon. When time spans are short, a percentage change and a log difference often produces very similar answers. However, when the process evolves over many years or decades (as in our case) the accuracy loss from using percentage changes can be large.

We think that without access to all the detailed official national and sectoral accounts and without collecting depreciation rate specific data with necessary surveys, the depreciation rates cannot be estimated accurately. In fact, the national statistical institutes should provide the depreciation rate estimates.

This database involves sectoral and national figures about the consumption of fixed capital from the beginning of 1960s for 12 OECD countries. The data for consumption of fixed capital are given for the following countries and periods (OECD, 1998): Australia (1966–1995), Belgium (1970–1996), Canada (1961–1997), Denmark (1966–1992), Finland (1960–1996), France (1970–1997), Germany (1990–1997), Germany (West, 1960–1994), Italy (1980–1994), Norway (1978–1997), Sweden (1980–1994), UK (1984–1994) and US (1960–1997). Most of the studies use depreciation rates of US only, but in our study we used the arithmetic mean of all countries. Hence, the averages are taken for several countries over a very long period of time.

We also included dummy variables to account for the break points suggested by the Lee and Strazicich (2003) Minimum Lagrange multiplier unit root test. However, neither of these dummy variables was found significant.

Note that the equality of the coefficient of k ₁ to the coefficient of k ₂ does not imply that k ₁ and k ₂ are perfect substitutes (i.e. identical inputs). Consider the constant elasticity of substitution (CES) functional form as where ρ ≤ 1, ρ ≠ 0, r > 0, α and β are share parameters between 0 and 1, r is homogeneity coefficient, and is elasticity of substitution. As known, if ρ = 0 (σ = 1, unit elasticity of substitution), the CES function reduces to Cobb Douglass function since . Hence the equality of the coefficient of k ₁ to the coefficient of k ₂ implies equal share parameters for inputs (α = β). Consequently, according to the above reported estimation results, the share of transportation–communication capital (in production) alone seems to be around the share of all other capital.

Generalized impulses, as described by Pesaran and Shin (1998) constructs an orthogonal set of innovations that does not depend on the VAR ordering.

Excluding residential and transportation–communication capital.

In the variance decomposition analysis reported in Tables 5–7, the following Cholesky ordering is followed: k ₂ k ₁ y. The reason of this choice is that, in terms of impulse responses this ordering gives the closest figures to that of generalized impulses method which constructs an orthogonal set of innovations that does not depend on the VAR ordering.

In fact, this analysis can be much more interesting in this respect if our study would distinguish the private and public capital. However, for the sake of simplicity and the degrees of freedom concerns due to the limited observations, this type of analysis could not be detailed here. For a good example of the discussion of complementarity hypothesis in terms of private and public capital using impulse-response and variance decomposition analyses, see Ramirez (2004, pp. 169–172).

For similar findings about the total public capital instead of transportation–communication capital, using the same methodology with us, see Kamps (2005, pp. 549–551).

Excluding transportation–communication capital.