Real Regional Productivity Differentials for Use in Transport Appraisal
D Johnson, J Dargay, ITS, University of Leeds, UK; K Reilly, Leeds University Business School, UK
This project identifies regional productivity differentials which can be used for the calculation of GDP impacts not typically captured as welfare gains in the appraisal of transport projects.
The Department for Transport appraises projects in a sustainable development framework, setting out components of social welfare benefits and costs of a scheme, relative to not implementing it. Whilst most welfare gains from transport schemes are increases in GDP, some are not. As part of the ongoing development of this methodology there has been a strand of work devoted to the Wider Economic Benefits of a project which focus on GDP impacts not captured as welfare gains. Such effects include the changes in agglomeration following an investment, benefits arising from changes in competition, and economic benefits from increased employment and productivity. It is this latter effect which we are concerned with in this study, commissioned by the Department for Transport. Specifically this effect, the increase in labour productivity in a particular sector, or GP3, can be measured as the sum over all areas of the product of the change in employment in each sector/region, the level of productivity in that sector/region and the national average industry GDP in that sector. The aim of this project is to identify regional productivity differentials which can be used for this calculation. These differentials also tell an interesting story themselves about the distribution of economic activity in Britain. Since transport investments often have their greatest effect within a relatively small geographic area, the productivity measurements need to be highly disaggregate. In particular we focus on the Local Authority District (LAD) level for Great Britain (405 districts in England, Scotland and Wales).
Productivity studies are concerned with modelling output per unit of input, where output can be measured based on a single input such as labour, or in the context of multiple inputs using total-factor productivity or multifactor productivity techniques. Given the data available and time constraints we use labour productivity as our productivity measure. Under the assumption that the labour market is in equilibrium, there is equality between the hourly wage and marginal product of labour. Using this as the starting point we can borrow directly from labour economics a well developed framework for estimating wage equations.
Using data from the Annual Survey of Hours of Employees (ASHE) for 2004-2006, a reduced form wage equation is estimated using regression analysis of hourly wage on vectors of individual characteristics (including age, gender and occupation), employer characteristics (industry and firm size) and a vector of location dummies at the LAD level. Estimates of the coefficients on the regional dummies represent the additional productivity of an employee in particular area, relative to the excluded region. We estimate two models - with and without interactions to provide a test as to whether the regional effects are caused by differences in industrial composition amongst regions. Our approach allows us to estimate each region?s productivity relative to the national average.
Our parameter estimates are consistent with the literature on earnings, notably that. ceteris paribus, there exists significant concave relationships for both age and tenure, women are paid significantly less than men and that wages increase with firm size.
In addition, the coefficients of all occupational dummies are highly significant similar in both specifications. All regional dummies are also highly significant in both models. Whilst controlling for regional industry composition increases the range of differentials, we find that the standard errors of the regional coefficients are considerably smaller in the model without interaction terms. This reduces the number of regions where productivity is statistically different from the national average.
With both models, London areas LADs dominate the most productive areas, with the City of London, Tower Hamlets and Westminster being the first 3 LADs in terms of productivity, and the same 8 LADs appear in the top 10. Regarding the LADs with the lowest productivity, both specifications agree that Isles of Scilly and Penwith are the bottom two, and that the Isle of Wight, North Norfolk, Kerrier, Torridge, Arun and Hastings are among the bottom 10. There is a larger geographic spread of the least productive regions.
We also show how our results have a lower spread over regions than raw earnings differentials, ie by controlling for personal and firm level characteristics, the regional differentials in productivity are lower - use of the difference between regional wages and the national average as a measure of a regional labour productivity will provide a significant over- or underestimate of regional productivity differentials. This suggests caution should be used when employing income based estimates of labour productivity defined in terms of Gross Value Added in appraisal.
Association for European Transport