I have written a few times about the Keynesian predictions that when WWII ended, the economy would be plunged into Depression and mass unemployment. In Fight of the Century , John Papola and I have Hayek say:
When that war spending ended your friends cried disaster
Yet the economy thrived and grew faster
Those friends were people like Paul Samuelson who had written  (HT: David Henderson):
When this war comes to an end, more than one out of every two workers will depend directly or indirectly upon military orders. We shall have some 10 million service men to throw on the labor market. We shall have to face a difficult reconversion period during which current goods cannot be produced and layoffs may be great. Nor will the technical necessity for reconversion necessarily generate much investment outlay in the critical period under discussion whatever its later potentialities. The final conclusion to be drawn from our experience at the end of the last war is inescapable–were the war to end suddenly within the next 6 months, were we again planning to wind up our war effort in the greatest haste, to demobilize our armed forces, to liquidate price controls, to shift from astronomical deficits to even the large deficits of the thirties–then there would be ushered in the greatest period of unemployment and industrial dislocation which any economy has ever faced.
(From Paul Samuelson, “Full Employment after the War,” in S.E. Harris, ed., Postwar Economic Problems, 1943.)
John Cochrane writes  (HT: Arnold Kling ) about Lawrence Klein’s 1946 article wondering why the Keynesians went so wrong. W.S. Woytinsky responded to Klein  in the April 1947 issue of the JPE. It’s a fascinating self-described polemic. A few highlights:
The first step in an objective appraisal of the results of the test of forecasts should be to locate the errors. Klein labels the unlucky forecasters as “government” or “Washington” economists. Actually, various projections were prepared during the war in Washington, New York, Chicago, Boston, and other cities; some were made by government econo- mists, others by private experts. Some referred to 1950, some to 1946-47, others to a year described as “194x.” The worst of the projections-with postwar unemployment of 20 million and more- were prepared by research units of labor unions and by journalists crusading for full employment through deficit spending. Thus, the distinction between bad and good predictions did not follow the line of Washington versus the rest of the nation or of government economists ver- sus private economists. Apart from the purely political projections, too poor for a post-mortem, a deflationary spiral and mass unemployment after the end of the war was, in fact, predicted by R. G. D. Allen (London School of Economics), Michal Kalecki (International Labor Office), Jacob Mosak, Robert Nathan, John Pierson, Beardsley Ruml, Arthur Smithies, Everett Hagen, economists as- sociated with the National Planning Association, and many others. Opposite forecasts were prepared by Richard M. Bissell, Edwin B. George, Sumner H. Slichter, Rufus S. Tucker, economists of the Brookings Institution, of Fortune, and of the Committee for Economic Development, the present writer, and others. All in all, about thirty forecasts have come to my attention in recent years, some of them very general, others more elaborate. While many forecasts pointed toward depression and mass unemploy- ment, about half of them suggested inflationary tendencies after the war. For brevity, the predictions of the second group will be handled as “correct” in the following analysis and those of the first group as “erroneous,” although there were different degrees of correctness and error in both groups.
The outstanding feature in the distribution of errors is that nearly all forecasts classified as “erroneous” were sup- plied by proponents of the Keynesian theory, while most of the predictions originated by other schools-in the gov- ernment or private agencies-proved to be correct, at least for the transition period. The probability that this distribu- tion of errors is unrelated to theoretical differences is extremely small indeed.
This is followed by the incredibly large margin of error in the predictions of post-war calamity and a lengthy discussion of the attempts of the day to specify and estimate the consumption function. Here is Woytinsky’s summary of such techniques, which reminds me of what is wrong with say, the CBO’s “estimates ” of job creation based on multipliers estimated from past relationships between government spending, output and employment:
The preceding reasoning does not pretend to prove that consumption expenditures and savings are not functions of disposable income. Nearly any time series may be expressed as a function of other time series; excellent fits may be obtained between such series as the number of visitors in our national parks and the number of automobile accidents in Australia or of divorces in Paris. Unfor- tunately, these functions are purely descriptive of observed data and permit no extrapolation. Consumption functions like those used in predictions of mass unemployment in 1945-46 belong to the same class. Klein may meet my criticism by a suggestion that new and better regression formulas be developed, relying on a longer series of observations and taking into account a number of inde- pendent variables. Such formulas certainly may be presented, and some of them will probably show excellent fit with empirical data. It is not clear, however, whether this will improve the technique of projection. Based on a mathematical elaboration of several independent series, the new formulas will defy any extrapolation unless it is assumed that the interrelation of observed features will remain the same in the future as in the period surveyed. In practice, after having extrapolated such a formula, one must use his judgment in deciding whether the result is good. If one likes the result, he accepts it; otherwise, he adjusts it to his own ideas. Regression or no regression, if the judgment of the analysis is sound, he has a good chance of hitting the nail on the head. If his judgment is poor, he will produce projections not much better than those which have covered with immortal glory the American-Keynesian school.
Woytinsky’s analysis of how the Keynesians came to make such poor predictions, is also very timely, and reminds me of what I have written  on ideology and macro:
The authors of the false predictions rank too high as economists to warrant the belief that this was an accidental error. It seems more plausible to believe that they used the pattern of reasoning because it led to conclusions which appealed to them as correct. The process may have developed subconsciously, but it appears fairly certain that many people were attracted to the consumption-function theory not by its mathematical justification but by its social and political implications.
Moreover, such has been the origin of all spectacular false prophecies. Apprehension of the approaching terrible events comes first. Then, some dark passage is dug out from an Old Book that can be interpreted (extrapolated) in such a way as to point toward the approaching danger. The text catches the imagination of those who believe in the forthcoming calamity and becomes the foundation of their faith. Although the technique used for fitting the text to the coming developments is utterly irrelevant, it should be recognized that in most cases it is strictly honest, solemn, and mysterious. If the apprehension is correct, the prophecy will be fulfilled; if it was fallacious, it will be forgotten. In the particular case discussed here, the apprehension was not correct and the prophecy cannot be saved by a reinterpretation of the fatal passage.
It is remarkable how obscure the post-WWII Keynesian predictions had become until recently. They were forgotten. We’re trying to recover those memories. Interested readers with access to Jstor may also enjoy this analysis  of the post-war predictions by Everett Hagen, a Keynesian who like Klein, tries to figure out what went wrong.