4. Testing Investment Lag Hypothesis 
The second major theory of investment cycles refers to the length of the gestation period of investment goods. It plays, for example, a very important role in Josef Goldmann's theory of quasicycles although he combines it with the theory of planning cycles. Goldmann argues that the planners' attempt to speed up economic growth beyond the tolerable limits results in a situation in which the economy, and especially the building industry, cannot digest the increased amount of investment and, as a result of the "pipeline effect", the finishing of investment projects is delayed, plans cannot be fulfilled, the rate of growth of output begins to decline, investment is cut down, and the whole economy slips towards the bottom of the quasicycle. The essential part of Goldmann's theory is that this wavelike behavior of investment will reappear with a lengthy time lag, caused by the 810 years long gestation period of the large investment projects, in a similar wavelike behavior of finished construction.  
2 Goldmann argues that the number of large and essential projects which were started at the peak of the previous investment wave and got stuck in the pipeline will be finished 810 years after the previous wave. 'this sudden increase of completed productive capacities would release the tension in the economy and help to speed up the new upswing which started in the meantime, thus contributing to the reinforcement of the investment cycle. Goldmann offers some empirical evidence about the cyclical fluctuations in the completion of capital construction  which is, by the way, also clearly visible from Diagram 3, but his claim that the capacities completed in one peak stem from the previous cyclical peak remains to be proved.  
3 To test Goldmann's hypothesis about the length and, consequently, also about the role of the gestation period, we estimated Almon's polynomially distributed lag between investment and newly completed capital construction. This was done separately for total investment, investment in "buildings and construction", and investment in "machines and equipment". Table 3 shows three alternative estimates of lag equations  differing in the degree of the lag polynomial for each of these three categories. In all cases it was assumed that the lag is distributed over a period of 6 years and in the seventh year the lag function was forced through zero. We also estimated shorter and longer distributed lags and lags without the final zero restriction. However, except for some perverse and economically meaningless cases the results were quite similar to those reported in Table 3.  
1Table 3. Estimates of Investment Lags
 
 
 
We ought to stress that because of the lack of any other information we had to use annual rather than quarterly or monthly data, as would have been desirable for the estimation of such a short distributed lag. In this light the obtained results are not so bad. Except for a few negative but statistically insignificant lag coefficients, the lag functions look quite reasonable. The zerolag coefficient is always, the one year lag frequently, and the two year lag coefficient in three cases, significantly different from zero. In no case is the 3, 4 or 5 year lag coefficient significant. The best statistical results were obtained with the quadratic lag function. However, the quadratic function may be too restrictive on the shape of the lag distribution, so that we should be cautious in interpreting the results. What do the estimates show? First, they show that the mean lag on total investment was slightly less than one year and for investment in buildings and construction slightly more than one year. We must, however, judge very carefully, because the standard errors of mean lag estimates are relatively large No conclusion about the mean lag of investment in machines and equipment can be reached, because the standard error is extremely large and because negative lag coefficients can make the mean lag meaningless.Second, the estimates of individual lag coefficients  at least the first two are quite revealing. They say that: 4456 percent of total investment are spent on projects which are finished in the same year, and 1928 percent are spent on projects finished in the immediately following year. For investment in buildings and construction this percentage is slightly lower: 4050 percent of investment are finished in the same year and 2026 percent in the following year. For machines, however, 5675 percent of all yearly investment are put into projects completed in the same year.Third, the percentage of investment which needs more than three years to be completed is very small in all three categories.  
Although we estimated the investment lag, i.e. the lag between investment spending and the introduction of newly completed capital, and not the full gestation period, which Goldmann is using in his theory, our findings still considerably question the validity of his conclusions. If our estimates are right, then it follows that the completion of capital construction should fluctuate with a relatively short lag after the fluctuation in investment spending. The investment peak and the resulting completion peak are separated only by about one year and not by an 810 year distance. In other words, both peaks are part of the same cycle, and therefore the investment lag can hardly play the role of a propagation mechanism which carries over the cyclical movement from one cycle to the other. This does not mean that the investment lag is of no importance in the explanation of cycles. However, its role must be interpreted differently. We shall return to that later. 



