Trade liberalisation and growth: a threshold exploration

Abstract

Openness and trade liberalisation variables are consistently estimated to have significant positive coefficients in panel growth regressions. Many arguments have been advanced as to why and how more open or liberalised economies might grow faster, but the specific channels this process uses have begun to be investigated only recently. We continue these efforts by including a variable identifying the date of trade liberalisation in a system of equations that captures the determinants of growth in per capita income. Four ‘channels’ are considered: capital formation, the share of government, the economy's openness to trade and its price distortions. We include the liberalisation variable in the equation explaining each channel, and allow for thresholds on its coefficient depending on the ‘years since liberalisation’. These estimated coefficients can also differ by region. In this way, we can identify the channels through which trade liberalisation affects growth and uncover the timing of the adjustments involved.

Keywords:

• trade liberalisation
• economic growth
• thresholds

Acknowledgements

We thank the editor, referees and participants in the First Annual NTU/GDC Workshop on Economic and Policy Developments in East Asia for comments.

Notes

1. For the period 1980–1989 for example, 79% of all loans had conditions in the trade policy area, in excess of those which attached to any other policy (Greenaway, 1998).

2. See Singh (2010) for a survey.

3. However, their liberalisation dates are generally consistent with those of Wacziarg and Welch (2008).

4. Indirect evidence is provided by the recent literature on exporting and productivity (for a review see Greenaway and Kneller 2007).

5. The trade share regression includes the growth of per capita income, land area, population and the three trade policy indicators – the import duty share, NTB coverage and the Sachs-Warner indicator of liberalisation status.

6. The estimated indicators are: including all three variables TP1 = −34.73 (Import Duty share) − 0.22 (NTBs) + 11.26 SW; and dropping the SW indicator of liberalisation status TP2 = −60.91(Import Duty share) − 0.24 (NTBs).

7. This has its origins in the Sachs and Warner (1995) openness indicator which is a dummy variable, with a country being classified as closed if it displayed at least one of five criteria, namely; (i) average tariff rates of 40% or more, (ii) non-tariff barriers covering 40% or more of trade, (iii) a Black Market exchange rate (BMP) that has depreciated by 20% or more relative to the official exchange rate, on average, (iv) a state monopoly on major exports, (v) a socialist economic system. The liberalisation date is then defined as the year in which none of these criteria are met. The openness measure was heavily criticised by Rodríguez and Rodrik (2000), who argued that the information on the BMP and the state monopoly on major exports played the major role in its classification of countries as open or closed. They went on to argue that a high BMP is likely to reflect factors other than trade policy, including macroeconomic mismanagement, weak enforcement of the rule of law and high levels of corruption, while the information on the state monopoly of exports works like an Africa dummy. In updating this indicator, Wacziarg and Welch (2008) note that the liberalisation date is less subject to criticism, and are careful to cross-check their dates against case studies of reforms in developing countries.

8. There is some econometric analysis of the dynamics of trade liberalisation and income levels or growth. Greenaway, Leybourne, and Sapsford (1997) use a smooth transition model to test for a transition in the level and trend of real GDP per capita for 13 countries and relate these to liberalisation. While all displayed a transition in level or trend, the majority were negative, and the positive transformations generally could not be related to liberalisation episodes. Greenaway, Morgan, and Wright (1998, 2002) (GMW) use a dynamic panel model to examine both the short- and long-run impact of liberalisation on growth in a large sample of countries. Results using three measures of liberalisation suggest a J-curve effect, growth at first falls but then increases after liberalisation.

9. SUR allows for correlation in error terms across equations and should be more efficient than OLS.

10. Our choices of variables and hence channels is constrained in two ways. First, there are the usual data constraints. The unavailability of data on many variables of interest for developing countries is well known. Using thresholds based on years-since-liberalisation also benefits from the inclusion of early liberalisation episodes. Second, it is best to view our equations, like Wacziarg's, as reduced form equations where many endogenous influences have been eliminated. We are not attempting to estimate a structural model of an economy. While the inclusion of country and time fixed effects in all equations captures idiosyncratic country and time (e.g. international business cycle) influences, there remain many variables which might be included. We have concentrated on a parsimonious set of ‘exogenous' variables, similar to those chosen by Wacziarg. Some variables (e.g. political variables) which were not found to be significant in any equation have been dropped from the analysis, and are not reported.

11. To ensure that the threshold is related to the effects of trade liberalisation and not to other shocks and that the estimated threshold does not depend on the experience of a few early liberalisers, we only consider thresholds at 15 years or less since liberalisation.

12. These include geographical variables and other fixed characteristics (such as colonial status, ethnic diversity and resource endowments).

13. These coefficients are largely unaffected by the inclusion of thresholds, except that FDev becomes insignificant once the regional effects are included.

14. The exceptions are that CAD and RER are now statistically significant in the trade channel, and FDev is statistically significant in the investment channel.

15. This may reflect the small number of countries in this group and the lack of variation in their black market premia.

16. Indeed the BMP channel becomes significant in the high regime in the aggregate results.