Comments on “The Link between Product Market Reforms and Macro-Economic Performance” by Rachel

Griffith and Rupert Harrison

Stephen Nickell

Bank of England Monetary Policy Committee

November 2005

European Commission Workshop on Structural Reforms and Macroeconomic

Performance, Brussels, 18 November, 2005

• This paper makes a significant contribution to the expanding literature on product market regulation and macroeconomic performance.

• The strategy is first to relate product reforms to a measure of the mark-up of price over marginal cost. Second, this is then related to various performance measures. In some cases, these two steps are merged.

• The reported results suggest that product market reforms are associated with increased employment and investment.

• However, the lower mark-ups generated by reforms seem to be associated with lower levels of productivity and reductions in R and D and productivity growth rates.

• These latter results are based essentially on time series correlations. Some of the cross-section correlations tend to tell the opposite story in the sense that countries with lower rents tend to have higher productivity. However, cross-section correlations are easily corrupted by omitted variable bias.

• On the other hand, some of the time series results must also be treated with caution. In particular, identifying shifts in trend productivity growth rates from 15 years of data is highly problematic.

Some of the empirical modelling omits important variables. Consider employment.

A basic model can be written as follows. Production : LKAfY , (1)

Profit max. : PWLKAf L /, (2) (W = wage, P = price, = mark up) Assuming constant returns, labour demand is 0,/ ' gPAWKgL (3) How do we eliminate W/P?

A long-run wage equation is wZPWgL ,// 1 (4) - - where wZ are wage pressure factors (eg. benefits, unions, labour taxes etc.) and is the population of working age or the labour force.

Using this to eliminate W/P yields wZKAhL ,/,/ (5) + + -

Key points are that employment in each country is driven by the population of working age and all the wZ variables are important. The employment

equations estimated in the paper omit all these variables. Turning to investment, the model above implies that the long-run capital stock

will be determined by 0,/ '

22 gYAPcgK (6) where c is the cost of capital. This sort of model suggests that the investment equation is a dynamic version of e

o YAPCI ln/ln/ln, 32 (7)

Again, the investment equations in the paper omit most of the relevant variables

Any productivity models should be based on the production function. In logs, we might have

dynamicscyclehoursskillmixaky ,,, (8) , trend TFP, might be specified as ta o 321

Taking differences yields dynamicscyclehoursskillmixalky ,,,

where ta 321

Points to note. (i) When analysing these models, the levels equation (8) and the growth equation

(9) should always be consistent. This is not the case in the paper. (ii) Data on skill mix is important here. The education composition of the

workforce is available. (iii) Annual hours per worker is crucial. The rate of decline of this variable since

1983 varies significantly. For example, it is 0.5% pa in France, Germany, Italy, 0.2% in Scandinavia, 0.1% pa in UK.

(iv) I would specify as a long moving average in the equation explaining trend TFP. Indeed, in the growth term, I would personally take as a constant for each country since I do not see how we can detect changes in trend TFP growth in 15 observations.

• Finally, when computing the rents, I would not use a constant cost of capital over a period when long-term real interest rates have declined significantly.