Browsing by Subject "Spectral analysis"

Now showing 1 - 2 of 2

Business cycles and institutions : empirical analysis
(2017) Kufenko, Vadim; Hagemann, Harald
The cumulative dissertation covers diverse aspects of empirical analysis of business cycles and institutions. There are three research questions in focus. To address the interplay between business cycles and institutions, the first research question is formulated: could the Malthusian cycles be present in a frontier economy with abundance of land and which institutions could be responsible for the Malthusian regime and the transition from it? In order to consider the far-reaching implications of economic cycles for the development of economic thought, the second question is stated: can economic fluctuations quantitatively influence research output? To address the methodology of business cycle analysis, the third question is brought up: how may spurious periodicities emerge and how could one test for them? The main findings in the cumulative dissertation can be summarized as follows: i) it is shown that institutional arrangements may form economic constraints or build-up on the existing ones, responsible for the regimes in which cyclical fluctuations take place; ii) the interaction between the economic cycles and fluctuations in bibliometric variables representing research output in Economics as a science is analysed, and empirical evidence suggests the downswings of cycles stimulate more publications on the topic of crises and business cycles; iii) spurious periodicities emerge close to filtering bounds for real and simulated data after detrending, and it is demonstrated that simultaneous significance testing of spectral density peaks against the noise spectrum across different types of signals may help to reveal spurious periodicities.
Spurious periodicities in cliometric series : simultaneous testing
(2016) Kufenko, Vadim
In this paper we revisit the methodological aspects of the issue of spurious cycles: using the well-established clinometric data, we apply an empirical strategy to identify spurious periodicities and cross-validate the results. The analysis of cyclical fluctuations involves numerous challenges, including data preparation and detrending. As a result, there is a risk of statistical artifacts to arise: it is known that summation operators and filtering yield a red noise alike spectral signature, amplifying lower frequencies and thus, longer periodicity, whereas detrending using differencing yields a blue noise alike spectral signature, amplifying higher frequencies and thus, shorter periodicity. In our paper we explicitly address this issue. In order to derive the stationary signals to be tested, we perform outlier adjustment, derive cycles from the series with the asymmetric band pass Christiano-Fitzgerald filter using the upper bands of the Kuznets and the Juglar cycles as cut-offs, and obtain detrended prefiltered signals by differencing the series in the absence of fractional integration. Afterwards, we simultaneously test whether the spectral densities of filtered and detrended prefiltered signals are significantly different from the spectral density of the related noise. The periodicities from the Kuznets range were not simultaneously significant, and thus are likely to be spurious; whereas ones of the Juglar and Kitchin ranges were simultaneously significant. The simultaneous significance test helps to identify spurious periodicities and the results, in general, accord with the durations of the business cycles found in other works.