I ‘ve already commented on the Greek construction share of Gross Value Added. In this post I ‘d like to look into the subject from the building permits angle which are the ‘supply side’ without any price contributions.

Overall, since the stock of housing (close to 4.75mn) is much larger than annual permits (their peak was at 56,000 in 2005) and the housing depreciation rate is very low, housing should exhibit large inertia while the main drivers should be population and price dynamics.

To test the above hypothesis, I ‘ve calculated net construction as new house building permits minus demolitions for 2000 – 2012. Independent variables candidates were the previous year net construction, labor force change in the 30-44 and 45-64 age brackets, real compensation per person employed change, total population and population aged 15-64 movements and real house inflation (deflated with a constant – rather high – 3% rate of inflation).

As it turns out, only net construction, labor force changes for 30-44 and real house prices are statistically significant. Real compensation seems to suffer from endogeneity problems since it is highly positively correlated with housing prices (correlation coefficient of 0.79 with an R² of 0.62). Although certain variables are not I(o), the results point to a strongly cointegrated relationship. The regression outcome is as follows:

regressionWhat is clear from the above is that construction seems to be a strongly auto-regressive process with negative dynamics. Absent any other factors, this year’s construction will be 75% of last year. Coefficients for labor and price dynamics are quite low. An average annual growth of 20-30 thousand persons in the 30-44 labor force will result in an increase of 3-5 thousand while a (maximum in this time series) real price appreciation of 10% will only lead to 4000 more houses.

Since on average till 2007 the annual net construction was close to 40,000 houses, labor and price dynamics were only capable of maintaining that level since the AR(1) characteristics would lower permits by almost 10,000 in the first year. In other words, the system dynamics are strongly negative while the construction sector does not react much to outside forces but has structural dynamics of its own.

As a result and in contrast with other sectors, economic developments will mostly be reflected in price adjustments for the housing sector rather than a change in quantities. House prices might be an bubble, but the actual construction sector will not, which will also be reflected on its contribution to value added and GDP.

Since 2008, net construction has fallen from 29,000 houses to less than 8,000 which is mainly attributed to the large fall of more than 23% in house prices and to the negative internal dynamics of the sector. The open question is to what extent policies (tax treatment of property, credit availability etc) have contributed to the price fall and not final demand.

From a technical point of view, one can interpret the regression results as follows: If the intercept for net construction is allowed to change due to outside forces each year (attributed to the structural pressure of labor force movements and demand side push from house prices), the actual series appears to be stationary. Even at 99% confidence level, the net_construction(-1) coefficient appears to be lower than 1:

Greek construction regression confidence intervals

Actually, a Wald test for restricting the coefficient of net_construction(-1) to 1 yields a p-value of 0.006, pointing that the series is stationary (under the specification examined).