Studies In Business Cycles - NBER

viiiPREFACEmethods and summarizing what they seemed to show about the cyclicalbehavior of the activities we had studied These efforts were usefulmainly in a negative way. They demonstrated first that we needed toenlarge our sample of time series in various directions; second that muchmore intimate knowledge of economic activities than I possessed wasnecessary to understand their reactions to business cycles; third that ourfindings could not be adequately presented in a single volume as I hadthe(suchothenaively expected. The upshot was that we enlarged our staff. The detailed interpretation of our findings was undertaken by a group of collaborators who were or became specialists in such fields as agriculture,construction work, transportation, merchandising, inventories, prices,labor problems, foreign trade, international finance, and banking. I stuckto the task of trying to see how the results irs all the fields fitted together,depending on others to supply knowledge I lacked. I\'Ieanwhile, Dr.Burns familiarized himself with the uses of our measures by preparinga preliminary analysis of our findings about construction work, andthen devoted himself to a final critique and revision of our statisticalmethods.The present volume is mainly his work. Though the basic featuresof my original design have been retained, Dr. Burns has made our tech-nique of measuring cyclical behavior a much better kit of tools thanit was when put into his hands. All of the tests of our measures wereplanned and executed by him. So also were the chapters on the time unitand our treatment of secular and random components. With minor exceptions, the drafts I contributed have been so much improved by himthat they have become virtually his products. I have taken advantage ofmy seniority to insist against his wish that the relative shares we havehad in preparing this book be represented by putting Dr. Burns' namefirst on the title page.The dozen chapters form three broad groups. Chapters 1.5 describeour methods, first in general terms, then in full detail. Chapters 6-8 elaborate upon three themes treated briefly in the third chapter: our insistence,upon using monthly data whenever they can be had, and our peculiarways of dealing with secular and random movements. Chapters 9-12examine critically the justification for using averages to express the typical characteristics of cyclical behavior.We have sought to make the book useful to several groups of readers.While the discussion is focussed throughout upon the National Bureau'stechnique, certain alternative methods of time-series analysis come infor examination. Also, the abundant illustrations of our results possesssubstantive as well as methodological interest. Students who desire earnestly to understand business cycles must feel a professional interest inthe design and efficiency of the tools used in observing cyclical behavior,On the one hand, these observations show what should be explained; onDsienare ria.tesasformmauieupparts0anaya.sistan.1

ixPREFACEyclicalusefulled tomuchthe other hand, they afford means for testing explanatory hypotheses. Forsuch specialists, there is no short Cut. To facilitate use of the book byothers, we offer a tabular guide.wasA Reader's Guidelat ourI hadbe de-Laymen and economists with a general interest in busi.ness cyclesStudents of business cyclesTime-series analysisMeasurement of economic magnitudesTesting hypotheses regarding time seriesandlisticalscribecuhar942typi-reau'sne inossesse earest inavior.d;tosee what economic statisticians' may contribute to'theory'Who have wider interests but are only incidentallyconcerned with statistical techniquesStatisticians whose primary interest istetherDr.paringturesr techthanwere[e unit[or cxhimage ofhavename1.2have little or no concern with the empiricalfoundations of the subject, but would likepricesstuckk,Chaptersrecommended1.4', 9b, 12'1-4', 9.122, 3, 5-84.2,9.l2d'Sec. IV of Ch. S may be omitted.'Perhaps add Sec. VIL-VIlI of Ch. 10, aiid Sec. VII of Ch. 11.'Omit Sec. III.dOmit Sec.Dr. Burns and I have received generous and varied help from manyfriends. Acknowledgments of specific suggestions on technical pointsare made at the appropriate points in the text, and need not be recapitulated here. But this is the place to mention obligations of a more general sort.Chester I. Barnard, W. Leonard Crum, and Oswald W. Knauthformed the Committee of National Bureau Directors who examined ourmanuscript. All three raised questions of which we have taken accountas best we could. Mr. Barnard's critique incited us to add a section toChapter I on 'The Symbols Used in Observing Business Cycles'.Our present and former colleagues in the National Bureau have readparts or all of the successive versions through which the manuscript haspassed, and aided in various ways to better it. We are especially indebtedto Milton Friedman, Simon Kuznets, Frederick C. Mills, Geoffrey H.Moore, Julius Shiskin, and W. Allen Wallis. We have benefited also fromcriticisms or suggestions by Moses Abramowitz, James 'SAT. Angell, G. 1-leb-erton Evans, Gottfried Haberler, Edward E. Lewis, Oskar Morgenstern,George Stigler, Albert Wohlstetter, and Leo Wolman. Harold Hotellingof Columbia University kindly advised us on some points of statisticalanalysis. Martha Anderson put her editorial skill at our disposal, andHanna Stern prepared the Index.We owe an especially heavy debt of gratitude to our statistical assistants. Karl Laubenstein was mainly responsible for compiling data

IPREFACEand verifying sources. H. Irving Forman, Sophie Sakowitz and DenisVolkenau, aided at times by several computers, carried through the extensive calculations. Sophie Sakowitz helped also in the preparation ofthe printers' copy, and assumed the main burden of reading the proofs.H. Irving Forman drew all the charts in the volume, and assisted withthe proofs. Sally Edwards typed our bulky and difficult manuscripts witha diviner's art. Without the care, patience, and skill of all five, our taskwould have been far harder and less pleasant.WESLEY C. ,V'TI'VIIIIxPLANIIIIIIIVAppern

IDenisthe ex'ion of1roofs.withwithr taskContentsPageViiPREFACECHAPTER 1WORKING PLANSIIIIIIIVVVIVIIVIII3—22The Point of DepartureQuestions Raised by the Definition35'Inductive Verification' of Business-cycle Hypotheses.The Data Needed for Observing Cyclical Behavior. .Requirements that Technique Must Meet.The Symbols Used in Observing Business CyclesRange Covered by the ObservationsThe Program as a Whole810.11.141721CHAPTER 2PRELIMINARY SKETCH OF THE STATISTICAL ANALYSISIIIIIIIVVVIVIIVIIIIX23—36Basic Features of the AnalysisReference Dates, Reference Cycles and Specific Cycles.Timing, Duration and Amplitude of Specific Cycles.Measures of Secular MovementsCyclical Patterns.Measures of Conformity to Business Cycles.Averages and Average DeviationsCharts of Cyclical PatternsComparison with Customary Techniques . . .232426282931333436CHAPTER 8PLAN OF TREATING SECULAR, SEASONAL AND RANDOM MOVEMENTS 37—55IIIIIIThe 'Cycle of Experience' as the Unit of Analysis.Limitations of the TechniqueNeed to Economize EffortIVTreatment of Seasonal VariationsAppendix: Notes on the Elimination of Seasonal Variationsxi.37404143.51

CONTENTSxiiCHAPTER 4PageDATING SPECIFIC AND BUSINESS CYCLESIIIIIIIVVVIDating Specific CyclesDiffusion of Specific CyclesDifferent Methods of Deriving a Reference ScaleA Tentative Schedule of Reference Dates.Difficulties in Setting Reference Dates Illustrated.Dependability of the Reference Dates .56—114-.-.56667176--.-81--.--94--.-CHAPTER 5THE BASIC MEASURES OF CYCLICAL BEHAVIOR.IIIIIIIVVVIVIIVIIIIxxXI115—202Positive and Inverted Specific CyclesTiming of Specific CyclesDuration of Specific CyclesAmplitude of Specific CyclesMeasures of Secular MovementsSpecific-cycle PatternsReference-cycle PatternsRelation between Reference- and Specific-cycle Patterns .Conformity to Business Cycles: Behavior during Fixed PeriodsConformity to Business Cycles: Timing Differences Recognized.Analysis of Quarterly and Annual Data 185197-ROLEICHAPTER 6EFFECTS OF THE TIME UNIT ON CYCLICAL MEASURES'HII . 203—269The Problem of This Chapter'Why the Time Unit MattersThe Direction of Movements in Time Series.InI]SECULASThe Number of Specific CyclesIIDuration of Specific CyclesTiming of Specific CyclesIIIIAmplitude of Specific Cycles in Annual DataAmplitude of Specific Cycles in Quarterly Data.The Secular Component of Specific CyclesSpecific-cycle PatternsDifferent Forms of Annual DataReference-cycle MeasuresConclusionsIVVVIIVIIVIIICYCLICCHAPTER 7EFFECTS OF TREND ADJUSTMENTS ON CYCLICAL MEASURESIIIMaterials Used in the TestsThe Number of Specific CyclesI270—309-.-.271-,-273IIIIIIVI

ITiming and Duration of Specific CyclesAmplitude of Specific CyclesReference-cycle MeasuresVariability of Cyclical IIIIxxXI203204210215220223229241245246252261of the TestsThe Number of Specific CyclesTiming and Duration of Specific CyclesAmplitude of Specific CyclesThe Secular Component of Specific CyclesSpecific-cycle PatternsReference-cycle PatternsMeasures of Conformity to Business CyclesRange-Variability of Cyclical ?s'IeasuresUncertainties irs Identifying Specific CyclesConclusions163CHAPTER 9ROLE OF AVERAGES IN THE ANALYSISII3—269302ConclusionsEFFECTS OF SMOOTHING ON CYCLICAL MEASURES I115299CHAPTER 8II15—202280294The Time Unit and Trend Adjustments.8194276III370—383Variability of Cyclical BehaviorFunction of Averages and Average DeviationsProblems Raised by AveragesCHAPTER 10SECULAR AND DISCONTINUOUS CHANGES IN CYCLICAL BEHAVIOR. 384—417IIIIIIIVVVIVIIvi"Duration and Amplitude of Specific CyclesReference-cycle PatternsOther Cyclical MeasuresDuration and Amplitude of Business Cycles-. . 384393398401Business Cycles and Economic StagesBusiness Cycles before and after 1914403Conclusions from TestsPreparation for Later WorkCYCLICAL CHANGES IN CYCLICAL BEHAVIORI273.406412413--. . 418—465- . .418427 . 431 .440.CHAPTER 11268271-IIIIIJVLong Cycles Marked Off by Long Waves in BuildingLong Cycles as Deviations from TrendsLong Cycles Marked Off by Long Waves in Prices.Long Cycles as Triplets of Business Cycles.

CONTENTSxivVVIVIIPageLong Cycles Marked Off by BoomsLong Cycles Marked Off by Severe DepressionsConclusions and Plans for Later Work448455464CHAPTER 12STABLE AND IRREGULAR FEATURES OF CYCLICAL BEHAVIORIIIIIIIVVIndividual Features of Successive CyclesStable Features of Successive CyclesInfluence of Extreme Items on AveragesCausal Interpretation of AveragesTest of Consilience among the XDivision of Reference Cycles into StagesSome Supporting DataSources of 7181920AnS

Page448455464List of Tables466474491506TableISeries Classified According to the Process Represented, Country, and26Periods Covered by American and Foreign SeriesNumber of Business Cycles Covered by American and Foreign SeriesCoke Production, United States, 1914—1933Sample of Table SI: Timing and Duration of Specific Cycles, Coke Production, United States, 1914—1932Sample of Table S2: Amplitude of Specific Cycles, Coke Production,7Sample of Table S3: Secular Movements, Coke Production, United8Sample of Table S4: Specific-cycle Patterns, Coke Production, United9Sample of Table S5: Rate of Change from Stage to Stage of Specific85405514510II121314151617181920the Time UnitPage2020212526United States, 1914—193227States, 1914—193228States, 1914—193229Cycles, Coke Production, United States, 1914—1932Sample of Table RI: Reference-cycle Patterns, Coke Production, United30States, 1914—193330Sample of Table R2: Rate of.Change from Stage to Stage of ReferenceCycles, Coke Production, United States, 1914—1933Sample of Table R3: Conformity to Business Cycles, Coke Production,31United States, 1914—193332Bituminous Coal Production, United States, 1905—1939Chronology of Specific Cycles in Employment, Ten Manufacturing Industries, United States, 1919—1938Dates of Cyclical Peaks and Troughs, Successive Versions of Three Indexes of 'Industrial Production', United States, 1919—193859Reference Dates and Durations of Business Cycles in Four Countries .Number of Monthly or Quarterly American Series Available at Decennial Dates since 1860Short-term Fluctuations around the Cyclical Turns of 1937 and 1938,23 American SeriesSequence of Cyclical Turns in the 1937 Recession and the 1938 Revival,40 American SeriesProduction, Employment and Prices in Three Countries, 1914—1918, ,xv68757882868891

LIST OF TABLESxviTable2122232425PageDirections of Movement in Successive Reference Phases, 1854—1933: 46American SeriesSummary of Movements in All Reference Expansions and Contractions,46 American Series, 1854—1933Summary of Movements in Successive Reference Phases, 1854—1933:Based on 46 American SeriesSummary of Movements in Successive Reference Phases, 1854—1933:Based ott Three Subsamples Drawn from 46 American Series .Summary of Movements in Successive Reference Phases, 1854—1933:Based on DifferentApplied to 46 American Series.-10852Cot119Du;14257Av tion, United States, 1907—193814558Coal Production, United States, 1908—1919Sample of Table S4: Specific-cycle Patterns on Inverted Plan, Slab ZincStocks at Refineries, United States, 1921—193814635The Computation of Specific-cycle Patterns lilustrated: Bituminous---United States, 1907—1938--1476015061Sample of Table S5: Rate of Change from Stage to Stage of Specific-Cycles, Bituminous Coal Production, United States, 1907—1938 .Adjustment of Average Specific-cycle Pattern to Show Relative Varia-tion of Rates of Change from Stage to Stage of Expansion and Con-traction, Bituminous Coal Production, United States, 1908—1938Sample of Table Ri: Reference-cycle Patterns, Bituminous Coal Production, United States, 1905—1938The Computation of Reference-cycle Patterns Illustrated: BituminousCoal Production, United States, 1908—1919Sample of Table R2: Rate of Change from Stage to Stage of ReferenceCycles, Bituminous Coal Production, United States, 1905—1938Samples of Table R3: Conformity to Business Cycles, Three American--44Dat.43Six56Sample of Table S4: Specific-cycle Patterns, Bituminous Coal Produc-42and1363441ForPaiChJ i.Three American Series, Unadjusted and Trend-adjustedSample of Table S3: Secular Movements, Bituminous Coal Production,40Daican.-333953American Series124Sample of Table Si: Timing and Duration of Specific Cycles, Bituminous Coal Production, United States, 1907—1938129Sample of Table S2: Amplitude of Specific Cycles, Bituminous Coal Production, United States, 1907—1938133Average Amplitude of Specific Cycles on Positive and Inverted Plans,.ReMeliiitheHoPh;Go;Samples of a Section of Table Si: Timing of Specific Cycles, Three38485129.SatIgn106United States, 1879—1933-4750283i-102Sat105tigators, United States, 1854—1938Turning Dates of Specific Cycles in Six American Series36101.Business Cycles Recognized by the National Bureau and Other Inves-32Un454927 Sar103Relation between the Amplitude and Diffusion of Business Cycles,3198.2630Table45---62159177Conformity Indexes for P Instances of Positive Conformity in N Cycles 184Possible Divisions of Reference Cycles when Three to Five Stages AreAssigned to vtAm63161163FreDatJotnuaFreAm65AvAn:66Coed-67cific

LIST OF TABLESTable45Page984647xviiSample of Table RI: Reference-cycle Patterns, Railroad Bond Yields,PageUnited States, 1857—1933190Sample of Table R3: Conformity to Business Cycles, Timing DifferencesIgnored, Railroad Bond Yields, United States, 1857—1933192Sample of Table R4: Conformity to Business Cycles, Timing DifferencesRecognized, Railroad Bond Yields, United States, 1857—1933.193-484950rvlethods Used in Analyzing Quarterly and Annual Series198Illustrations of the Dependence of Specific Cycles in Annual Data onthe Months of Cyclical Turn206How i'vlonths of Cyclical Turn Determine Whether Brief CyclicalPhases Remain or Disappear in Calendar.year SummationsComparison of the Directions of Movement of Monthly and 015061¶6263-1671771846465,6667.Data, Six American SeriesComparison of the Directions of Movement of Two Series in AnnualForm with the Directions of the Same Series in Monthly Form: EveryPair of Six American SeriesCharacteristics of Cyclical Phases Skipped by Annual Data, Six American SeriesJoint Distribution of Durations and Amplitudes of All Cyclical Phasesand Those Skipped by Annual DataNumber of Specific Cycles in Monthly, Quarterly and Annual Data,Six American SeriesDuration of Specific Cycles Measured by Different Methods, Pig IronProduction, United States, 1879—1933Average Duration of Specific Cycles in Monthly, Quarterly and AnnualData, Six American SeriesFrequency Distribution of Leads or Lags of Specific Cycles in MonthlyData, Six American SeriesJoint Distribution of Corresponding Leads or Lags of Monthly and Annual DataFrequency of Leads or Lags and Average Timing of Specific Cycles, SixAmerican Series: Monthly, Quarterly and AnnualAverage Timing of Specific Cycles Computed in Different Ways, SixAmerican Series: Monthly, Quarterly and AnnualAverage Timing of Specific Cycles during Brief Periods, Six AmericanSeries, Monthly and AnnualStandings at Peaks and Troughs, Cycle Bases, and Amplitudes of Corresponding Specific Cycles in Monthly, Quarterly and Annual Data, PigIron Production, United StatesFrequency Distribution of the Differences between Absolute Amplitudes, Cycle Bases, and Relative Amplitudes of Corresponding SpecificCycles in Monthly and Annual DataAverage Amplitude of Corresponding Specific Cycles in Monthly andAnnual Data, Six American SeriesCoefficients of Rank Correlation between Amplitudes of CorrespondingSpecific Cycles in Monthly and Other Data, Six American Series Average Amplitude of Corresponding, Noncorresponding and All Specific Cycles in Monthly and Annual Data, Six American Series. 7.238

LIST OF TABLESxviiiTablePage68Average Amplitude of Specific Cycles during Brief Periods, Six Amencan Series, Monthly and Annual2396970Variability of Amplitudes of Specific Cycles in Monthly and AnnualData, Six American Series240Average Per Month Amplitude of Corresponding, Noncorrespondingand All Specific Cycles in Monthly and Annual Data, Six AmericanSeries71Standings at Cyclical Turns and Amplitudes of Monthly Data Corn-7778United States, 1883—1933Characteristics of Cyclical Phases Skipped by Calendar- and Fiscal-year737475767980Monthly and Twelve Forms of Annual Data, Pig Iron Production,242244878889American Series25298.C274AD99AD100CD101AA102A256Absolute Amplitude of Specific Cycles in Unadjusted and Trend.-095ASize and Frequency of Leads or Lags of Specific-cycle Turns in Trendadjusted Data at Corresponding Turns of Unadjusted Data, Five Amer.277ican Series with Upward Tr

Studies in Business Cycles 1 Business Cycles: The Problem and Its Setting By Wesley C. Mitchell 2 Measuring Business Cycles By Arthur F. Burns and Wesley C. Mitchell 3 A merican Transportation in Prosperity and Depression By Thor Hultgren 4 Inventories and Business Cycles, with Special Re