We have fitted explanatory functions and trends to several time series of Czechoslovak data, and plotted the ratios of actual to fitted values.  
Diagram 1 shows these ratios for production functions using national income or, alternatively, gross social product as a measure of output.  
Diagram 2 shows fluctuations of basic inputs (labor and capital) around their trends.  
Diagrams 3 and 4 demonstrate fluctuations in investmentrelated data and some other selected statistical series.  
The estimates of parameters used in calculating trends and fitted values are shown in Table 3 and in the Appendix. Already at the first glance one can see that many of the time series clearly exhibit fluctuations which seem to have a cyclical pattern. But such an intuitive approach may be misleading  the apparent cycles could just be a product of random disturbances.  
How can we distinguish endogenous cycles from purely random deviations? Clearly, at least by two criteria:
 
Looking at Diagrams 1, 2, 3, and 4, criterion (b) seems to be satisfied. One can see two major downswings (one in 195354 and the other in 196265) in many variables. Later, we shall present estimates of a lag structure between at least two variables, but a more rigorous test of lags will be done elsewhere. If the point (a) is correct, then the residuals in estimating the trend coefficients would be serially correlated. And, surely, almost all the regressions indicate a very high degree of autocorrelation. A purely random cause of the apparent cycles seems to be unlikely.  
This does not mean that random or exogenous shocks are not important. However, the primary interest in a theory of cycles generally and hence also in the theory of growth cycles is the endogenous mechanism that is responsible for the propagation of exogenous shocks through the system over time so that the initial disturbances reappear periodically either in damped or strengthened form (see Frisch). 



