Acta Univ. Agric. Silvic. Mendelianae Brun. 2018, 66, 1383-1391

https://doi.org/10.11118/actaun201866061383
Published online 2018-12-19

Disparity in Performance of the Czech Construction Sector: Evidence from the Markov‑Switching Model

Václav Adamec

Department of Statistics and Operations Research, School of Business and Economics, Mendel University in Brno, Zemědělská 1, 613 00 Brno, Czech Republic

Activities in the construction sector are assumed to be influenced by inflow of mortgage funding in the private housing sector and public finances targeted at large infrastructure projects, apart from climate variables. In this study, we modeled seasonal time series representing monthly output in the Czech construction sector in CZK mil. during 2000:1 through 2016 : 12 (T = 204) adjusted for calendar variations and seasonal movements via TRAMO‑SEATS and then transformed to natural logarithms of gross returns. A Markov‑Switching model with two states, no intercept, average monthly temperature, average monthly precipitation and parameters of first‑order autoregression process was specified and estimated by the Expectation‑Maximization. In State 1 of regular performance, the log‑differenced returns were significantly and positively influenced by precipitation levels, but not by ambient outdoor temperature. In State 2 of non‑standard operation of the construction sector, the transformed series was unaffected by precipitation levels, but instead by ambient outdoor temperatures. First‑order autocorrelation dependency in both regimes was established. Changes in legal and macroeconomic environment pertinent to tax law amendments affecting VAT or corporate tax, country’s accession to EU or large construction project deadlines were shown to induce nonstandard regime in the construction sector (State 2). The model classified 91 % observations in the first state, while only 9 % data belonged to the State 2. Transition probability matrix indicates that change from model State 1 to State 2 is difficult to attain. At the same time, once State 2 was established, it tends to persist or change to State 1 with near equal probability. Ability of the Markov‑Switching model to identify both states is reasonably good.

References

16 live references