A sectoral approach to Okun’s Law

ABSTRACT

The study adds to the existing literature on the relationship between output and unemployment by using a sectoral specification of Okun’s Law to capture the differential sensitivity of the unemployment rate to output developments in the services and manufacturing sectors. Using quarterly data for the period between 2000 and 2012, we show that Malta’s unemployment rate has been more sensitive to output developments in the services sector than to those in the manufacturing sector. We use different equation specifications and the youth unemployment rate to show that the impact resulting from developments in the manufacturing sector occurs mainly via lay-offs while developments in the services sector affect the unemployment rate via the hiring of new entrants into the labour market.

KEYWORDS:

• Okun’s Law
• manufacturing
• services
• unemployment

JEL CLASSIFICATION:

• E24
• E32

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ We measure the unemployment rate as the share of unemployed in the country’s labour force while output is measured by GDP. Both labour market and GDP data were obtained from the National Statistics Office (NSO), with the former based on the administrative records of the Employment and Training Corporation (ETC).

² Okun (1962) originally estimated two versions of the relationship: a difference version (which relates aggregate output growth to changes in the unemployment rate) and a gap version (which relates the gap between the actual and natural rate of unemployment to the gap between actual and potential output). Experiments along these lines have led us to believe that using the gap version as an alternative to the difference version should not alter the conclusions of this article.

³ We also considered an equation specification with lagged GDP growth based on the assumption that firms may react to changes in output growth with some lag. However, the estimates yielded similar results.

⁴ There were other observations which exceeded the ±2 threshold for studentized residuals. However, we focus on the two observations with the highest values.

⁵ Whereas Anderton et al. (2014) disaggregate GDP by its expenditure components, we use their method to disaggregate GDP by sector.

⁶ The β_iλ_i estimates are obtained by multiplying the weighted β-coefficients with the respective sector’s weights in GDP (λ_i) which are presented in column (iv).

⁷ The β_iλ_i estimates show the change in the unemployment rate that is associated with a 1% increase in output of sector i.

⁸ Along the lines explained in Anderton, et al (2014), the sum of the ‘component elasticities’ approximately adds up to −0.11, in line with the Okun coefficient estimates in the aggregate version of the Okun equation.

⁹ The stronger unemployment intensity of the services sector suggests that we should observe a stronger Okun coefficient as the share of the services sector in total output increases. Although this is beyond the scope of this article, we note that rolling regression estimates presented in Micallef (2013) indicate a strengthening of the Okun coefficient which coincided with a gradual and significant rise in the share of services in total output. Specifically, as the share of services in output increased from 73% in 2001 to 81% in 2012, the estimates suggest that the Okun coefficient increased from −0.08 during the first 20 quarters of the period under observation to −0.15 during the last 20 quarters. However, we note that the strengthening of the Okun coefficient may reflect changes in other factors (such as productivity) rather than just the increasing share of services, and while the cross-sector productivity differences would be captured by our specification, the within-sector productivity changes go unaccounted for.

¹⁰ The youth unemployment rate is defined as the number of unemployed persons that are 16–24 years old. Data was obtained from NSO and is based on the administrative records of the ETC.