Box 1.1. 
Forecasting long-term growth in transition economies

Details

Convergence at risk

The “productivity catch-up” phase associated with opening up to the outside world and international integration has ended in most transition economies. Much work remains to be done to bring their institutions and market structures up to the level of mature market economies. However, the way in which growth relates to capital stocks, human capital and institutions in the transition region should no longer be very different from other market economies.

It is therefore possible to analyse the long-term growth potential of transition economies in a standard growth accounting framework using a large sample of advanced, emerging and transition countries.¹ Growth, physical capital, total factor productivity, the saving rate and foreign direct investment are determined inside the model, whereas geography, demographic variables, institutions and human capital are treated as exogenous.

The following assumptions are made:

• TFP growth depends on human capital, FDI, the distance from major economic centres and the quality of political institutions (measured by constraints on the executive), as well as initial levels of TFP.
• 
The saving rate depends on demographic variables, natural resources and financial openness.
• Growth in the physical capital stock (investment) depends on the saving rate, FDI, the quality of political institutions and the initial level of capital.
• Finally, FDI depends on trade and financial openness, law and order (as a proxy for economic institutions), the shares of services and manufacturing in GDP, and the initial level of GDP.

The fact that growth in physical capital and TFP are functions of their initial levels implies that the model allows for “factor-specific convergence” – that is to say, the possibility that capital and TFP growth may slow as their levels rise.² The results suggest that this is indeed the case.

This system of four equations is estimated by three-stage least squares using a world sample of 88 countries over the period 1982 to 2011. The panel consists of five six-year intervals with period averages for all contemporaneous variables and the values of the final year of the preceding period for all initial conditions. Not all data are available for all countries over the entire period – data for transition countries typically start around 1990 – resulting in an unbalanced sample of 361 observations.³

The results support the contention that political and economic institutions play a crucial role in determining the prospects for growth. Variables related to policies (trade openness and financial openness) or institutions (constraints on the executive, and law and order) are significant in all four equations (see Table 1.1.1). For example, countries with stronger constraints on the executive are found to have a higher rate of TFP growth and faster accumulation of physical capital, while more open trade policies are associated with greater FDI inflows.

In addition, the levels of human capital and FDI are found to be important determinants of productivity growth. The negative coefficient for economic remoteness suggests that being close to global centres of economic activity promotes productivity catch-up. This is in line with the experiences of CEB and SEE countries, whose proximity to western Europe is widely viewed as having helped them to catch up.

The model is used to predict long-term growth rates based on specific assumptions about developments in the exogenous variables. In order to evaluate what the continued stagnation of reforms would imply for the growth prospects of transition countries, the baseline forecasts assume that institutions and openness will remain at their current levels, while human capital continues to grow at its current rate. The remaining variables are held constant, with the exception of demographic characteristics, which evolve in accordance with United Nations projections.

In this scenario the model predicts that transition countries will not sustain their pre-crisis growth rates in the long term. Chart 1.1.1 shows that in virtually all countries the average growth rate of output per worker is projected to be lower over the next two forecasting periods⁴ (that is to say, from 2012 to 2023) than it was between 2000 and 2011.⁵ In absolute terms, growth in output per worker is projected to be modest in most countries between 2012 and 2023 – between 2 and 4 per cent – and to decline further, by about one to two percentage points, between 2024 and 2035. The initial slow-down occurs despite the fact that the preceding period includes the deep recessions of 2008-09. The drop in growth rates is primarily due to diminishing TFP growth. For most economies shown, the slow-down in output per worker will be compounded by a stagnation or decline in employment as populations age.20 The exceptions here are the SEMED countries and Turkey, where the growth rate of GDP will remain significantly above that of output per worker as a large number of young people join the workforce.

The main finding of this analysis – the fact that, under their current policies, most transition economies can expect a significant slow-down in long-term growth relative to the past – is robust to variations in how exactly “current policies” are defined. For example, modest improvements in political institutions (such as a 1-point improvement on a 10-point scale) will not change the main result, and neither will a slow continuation of financial opening. To make a difference, large improvements in political and economic institutions are needed, as described in the main text.

1. The analysis assumes a human capital-augmented Cobb-Douglas production function in which output is a function of TFP (denoted by A), physical capital (K), human capital (h) and labour (L):
   
   Y_t = A_t K_t^α (h_t L_t) ^1-α  and y_t = A_t k_t^α h_t^1-α
   
   
   where y is output per worker (^Y/_L) and k is capital per worker (^K/_L).
   
   ∆ln(y) = ∆ln⁡(A) + α∆ln⁡(k) + (1 – α)∆ln⁡(h)[back]
2. 
This approach draws on recent literature on long-term conditional growth projections; see Lee and Hong (2010) and Chen et al. (2012). Data sources include the Penn World Tables (for capital, TFP, human capital, labour shares and growth data), the World Bank (for natural resource rents and sector shares), UNCTAD (for gross FDI), the Chinn-Ito index database (for financial openness), ICRG historical data (for law and order) and the Polity database (for executive constraints). Openness to trade is structurally adjusted using the adjusted trade intensity approach employed by Pritchett (1996). For further details, see Lehne and Zettelmeyer (2013). [back]
3. Dropping the measure of law and order allows a larger sample (455 observations), with a longer time horizon (1976-2011) and more countries (99). Estimating the model on the basis of this sample does not change the results for the other variables in the system. Neither does dropping the observations for the transition economies prior to 2005, a period when (as argued in the text) they may have been undergoing a unique catch-up process that made them structurally different, in terms of the model coefficients, from other countries. Further robustness checks are conducted in Berglӧf, Lehne and Zettelmeyer (2013). [back]
4. Separate forecasts are generated for each six-year interval from 2012 to 2035. [back]
5. Hungary and Slovenia are two exceptions. They experienced particularly weak growth between 2000 and 2011, which the model expects will be partly corrected in the next period. [back]
6. Eighteen transition countries are expected to see their working age populations decline by 
the mid-2020s. [back]

• Prev
• Next