What are the Drivers of TFP Growth? An Empirical Assessment

Abstract

This chapter builds upon the related research that grapples with determinants of TFP, as the driving force of potential growth. In particular, we empirically estimate, in a homogenous and systematic manner, cross-country contributions of cyclical and structural determinants of aggregate TFP growth. Under a growth accounting framework, we compute TFP growth estimates for 41 economies over the 1992–2014 period. After selecting its main drivers by means of a Bayesian Model Averaging (BMA) approach, we exploit panel estimates to conclude that a substantial share of the growth underperformance in recent years was related to cyclical factors, mainly the output gap, but also: (i) over-indebtedness for advanced economies; and (ii) the decline in commodity prices for commodity exporters. In addition, the growth of IT capital and the convergence towards the technological frontier appear to be significant structural drivers of TFP productivity growth in emerging market economies.

Over time, productivity growth is the key determinant of improvement in living standards

J. Yellen, Chair of the Board of Governors of the US Federal Reserve System, June 6th, 2016.

We would like to thank Ángel Estrada, Sonsoles Gallego, Ignacio Hernando, Enrique Moral-Benito, Pedro del Río and seminar participants at the Banco de España for their helpful comments. Many thanks as well to Abdul Erumban and Klass DeVries for providing ICT and non-ICT capital stock series from the Conference Board. Finally, we are especially indebted to Marina Sánchez del Villar for her excellent research assistance. All the remaining errors are our own responsibility.

Appendices

Appendix 1. Commodities Export Price Index (CEPI)

In order to build the Commodities Export Price Index (CEPI), we have considered commodities defined by the UNCTAD classification. Each country-specific weight is calculated on an annual basis relative to the value of total exports. To maintain weights constantly, the final weight for each category in the index is the average of each product’s weight for all years in the country.⁷

We adjust the index for the US deflator (base 2005). Thus, our CEPI measure is calculated as follows:

$$ CEPI_{k,t} = \prod\nolimits_{j} {\frac{{w_{j} *p_{j,t} }}{{Def_{US,t} }}} $$

(8)

where $w_{j,k}^{E}$ denotes the weight of each sub-index j and $p_{j,t} \;i$ s the value of the sub-index at each time t.

Appendix 2. TFP Convergence

As a measure of distance to the productivity frontier, we calculate:

$$ Dist_{k,t} = A_{k,t - 1} - A_{frontier,t - 1} $$

(9)

where A denotes TFP in levels using the following approximation:

$$ A_{t} = \frac{{Y_{t} }}{{L_{t}^{\alpha } *K_{t}^{1 - \alpha } }} $$

(10)

In order to compute the TFP frontier, we have considered the average of the TFP level for the three countries with the highest values at the reference year 2005, CHE, USA and GBR.

While the majority of the countries in our sample have reduced the distance at which their productivity levels stand in the last 10 years, some cases stand out (See Fig. 3a). In particular, Denmark not only outperforms the frontier countries during this period, but its productivity level surpasses the frontier level. Other advanced countries that outperform the frontier growth are Ireland, Korea and Finland. In emerging economies, almost all the countries converge, with the sole exception of Saudi Arabia. Those economies converging faster are mostly from Eastern Europe, with comparatively higher levels of human capital.

Fig. 3

A (left) Distance to the TFP frontier between 2005 and 2014. B (right) Global and conditional convergence

To illustrate this point, it is worth mentioning that our baseline convergence path is linear (in logs) while once we add the human capital factor, $H_{t}$, the speed of adjustment to the frontier rapidly increases. Additionally, bear in mind that the presence of significant country-specific individual effects (not included in our model) would yield to a non-convergence path as shown in Fig. 3b.