Comparative Cyclical Behavior of Short-Term Interest Rates in Four Countries, 1878-1938

This PDF is a selection from an out-of-print volume from the National Bureau of Economic Research

Volume Title: International Financial Transactions and Business Cycles

Volume Author/Editor: Oskar Morgenstern

Volume Publisher: Princeton University Press

Volume ISBN: 0-870-14091-4

Volume URL: http://www.nber.org/books/morg59-1

Publication Date: 1959

Chapter Title: Comparative Cyclical Behavior of Short-term Interest Rates in Four Countries, 1878-1938

Chapter Author: Oskar Morgenstern

Chapter URL: http://www.nber.org/chapters/c9463

Chapter pages in book: (p. 74 - 123) CHAPTER III

COMPARATIVE CYCLICAL BEHAVIOROF SHORT-TERM INTEREST RATESIN FOUR COUNTRIES, '878-1938

Section 1. The Role of Short-TermRates Short-term interest rates play a prominent rolein many cycle theories. This corresponds to their busine great signiflca ifl the national money markets. There is furthera widespread belief among economic quantities they are that exceptionallyexposed to international influences and that theirmovements may frequenuy produce favorable or unfavorable conditions abroad.For that reason they are frequently included in studiesof foreign Their role was undoubtedly exchang. great during the classicalperiod of the gold standard, i.e., between1871 and 1914. It is tive that we study their cyclical therefore impera. behavior. In thischapter we show the behavior ofsome typical series, and prepare the groundfor further operationsto be carried out withthem in chapters. subsequent Short-term interestrates are highly sensitive toinstitutional changes. Some ratesare operationally important of merely historical at one time, but interest later althoughthey may still be It is thereforeimportant to find quoted. out when such breaksin their significance occur andto what theyare to be attributed. this cannot be Often determined ina satisfactory manner; if then we mustat least exercise this happens great reserve ininterpretation. The ratesare determined by a number of elementswhich need not be identical inall countries. Among them themost important are the supply ofmonetary funds on the money marketsand the willingness of businessto borrow at the former isinfluenced by the prevailing rates. The reserve position of thebanks, the movements of gold, andthe actions of exists. The latter the central bank,if one are the result of theinterplay of little forcesoperating through understood the the "expectations"of businessmen comparison ofpresent and expected and with the future profit corresponding conditions conditions Most of these of borrowing. determinantsare themselves pendentupon national predominantly de- conditions. The linkwith other was chieflyestablished by gold economies movements and shifts ofshort-term 74 THE ROLE OF SHORT-TERM RATES capital. Policy measures may intervene one way or the other, though when the strict gold standard was in force they were infrequent andpossibly--of minor importance. We do not know, at present, how to estimate the relative significance of the domestic and foreign influences. This is a general dilficulty discussed above in Chapter II which will keep on cropping up. But it seems likely that the domestic influences outweigh the others in the case of short-term interest rates; the day-to-day fluctuations of the domestic economy would affect the domestic interest rate structure more frequently than the occasional disturbance of foreign origin, because distance alone creates a certain isolation from minor shocks coming from other economies, when there are no really large disturbances in any one country. This would, for example, be true toward the end of a cyclical contraction and at the early stages of an expansion. If a strong central bank is supplying ample funds, then there is another domestic factor frequently overshadowing international influences. So when several countries are simultaneously in these stages, it is plausible that the movement and the level of the various short-term rates are essentially the product of domestic forces. It may be quite different, however, when international financial crises occur. The interaction of short-term interest rates of different countries therefore need not have been of a stable nature; it may have been weak in some years, strong in others. We have here an instance of the possibility of varying cyclical dependence, mentioned briefly in Chapter I, section 8. The interdependence of interest rates within each money market is presumably very great, because of the possibility of translating a long-term rate into a series of successive short-term rates, e.g., by repeated borrowing or lending.' Because of this systematic interrelation of interest rateswhich isaffected also by other factors than the one just mentioned, e.g., by variations of risk among short-term and long-term ratesit is permissible to speak of an interest ratestructure,even if its properties are far from fully known. To avoid misunderstandings about the scope of the following 'Cf., for example, F. A. Lutz, 'The Structure of Interest Rates," Quarterly journal of Economics, Vol. LV, 1940, pp. 36-63. For general interest the important work by F. B. Macauley, Some Theoretical Problems Suggested by the Movements of Interest Rates, Bond Yields and Stock Prices In the United States since 1856, National Bureau of Economic Research, 1938, may also be referred to. 75 SHORT-TERM INTEREST RATES

observations on interest rates it is emphasized again thatthis isno a study of the function of interest, its influence upon the allocation of durable goods, its control of investment on money andcapital markets, etc. These arc far-reaching problems. We aim mainlyat description of a few interest rates and theircomparative, inter. national behavior. If it were our aim merely to get a picture of themovement of interest rates, i.e., of their cyclical behavior, it would probablynot be very important which short-term, medium-term, andlong-term rates we chose. The members of a given classof ratesmove closely together. But other considerations ar enter into our selections.One of themto be discussed in Chapter IXrelates to the absolutelevels of rates in different ti(J countries. In measuring suchdifferences in level the comparability ofrates is highly important. th Our data cover two main periods,the first from February ab to August 1913 (426 months), 1878 em the second fromNovember 1925 to December 1938 (167 months).In some of the subsequent and measurements different tables periods, i.e., parts of theabove, are P1 occasionally used; the technicalreasons for such deviations th noted on each are one of the tables or in thetext. There is notably ma subdivisi0of the postwar data a rat for the period ofthe moreor less general applicationof the goldor gold exchange standard. to We now formulate the criteria for the selectionof the short-term interest rates. Whatinformation do to give? we expect our interestrates First, we are primarily I concerned with cyclicalbehavior. Hence oux interest ratesmust be representative trth of the category ofshort- term rates as faras movement is concerned. chi Second, we shouldlike our ratesto be quantitatively trib in the sense that significant, LIS t a large volume of (domestic)business be trans. acted in the obligationsto which they hasi pertain. This isan important consideration in theevent that international in influences change the rate and thus directlyaffect a large CS1) volume oftransactions. ratp Third, we shalllater wantto form differences interest rates of between short-term In one country and thoseof another. The of course, thatthese differences idea is, for even primary, influence may have an important,perhaps 9 an upon short-term capital movements be- whic tween nations. Theinterest rates should tional significance therefore have interna- in thesense that a sufficient the rates of difference between from two countrieswould cause (liff(t These desiderata2 a flow of funds. are difficult to fulfill a simultaneously especially It is clear thatthey to (2l automatkally 1nsure thecomparability of our series be clu 76 FREQUENCY DISTRIBUTIONS since they should be continuously fulfilled for hundreds of months, i.e., taking possible institutional changes into account. This latter condition is especially hard to meet after World War I, even during the few years of the gold standard. But fortunately the task is not hopeless, if a reasonable latitude of view is adopted.3 Then it turns out that the internationally significant rate is generally one of great domestic importance too and hence relevant to the cyclical behavior of all short-term rates. Thus we find that the contacts among the large money markets are multiple: The domestic factors influencing the rate of interest such as monetary policytransmit repercussions to other national markets by affecting the differentials of the rates, while at the same time the domestic rate is subject to influences from abroad, because this rate is in turn exposed to similar repercussions emanating from the other centers. The rates selected were for the United States the commercial paper rate and the New York call money rate; for GreatBritain the London open market discount rate; for France the Paris open market discount rate; and for Germany the Berlin private discount rate. A brief description of these five will be found inthe appendix I to this chapter.

Section 2. Frequency Distribution$ It is of interest to investigate to some extent the frequencydis- tributions generated by the interest rate series, even though our chief concern is with their cyclical properties. Frequency distributions are often neglected in cyclical studies, but theyhelp us to gain a deeper insight intocyclical behavior by showing some basic properties, almost of an institutional character. Weneed, in any case, to know the average standof the different rates, especially in Chapter IV, when the question of absolutelevels of rates enters into the picture. In this first instance we determine thefrequency distributions for both the seasonally uncorrected and corrected series(Tables 9 and 10). This comparison itself brings out someinteresting facts which otherwise might be overlooked. For thereader who is in-

from country to country, though they may in each caserelate to somewhat different institutional conditions. * For the United States we shall even consider two interest rates, in order to check on a belief that one of them, the NewYork call money rate, should be chosen. 77 SHORT-TERM INTEREST RATES TABLE 9 Frequency Distributions Short-Term Interest Rates Prewar Period, January l876-Jtily 1914 Seasonally Uncorrected Data (per cent) Glass N.Y. Great Brstarn France Berlin, 0-. 0.39 commercial N.Y. call open market open market pthat. 0.40- 0.79 money rate discount rate discount rate Class paper rate (lisCOunt tute 0.80- 1.19 0- 0.39 1.20- 1.59 0.40- 0.79 17 1.60-. 1.99 0.80- 1.19 21 34 7 2.00- 2.39 2.40- 2.79 1.20- 1.59 36 51 32 23 1.60- 1.99 45 35 79 2.80- 3.19 42 3,20- 3.59 2.00- 2.39 59 55 100 50 2.40- 2.79 4 45 86 135 3.60- 3.99 63 4.00- 4.39 2.80- 3.19 17 40 40 52 84 3.20- 3.59 28 31 54 33 4.40- 4.79 64 4.80-- 5.19 3.60- 3.99 69 22 57 17 32 4.00- 4.39 51 5.20- 5.59 30 25 4 29 5.60- 5.99 4.40- 4.79 57 27 11 1 45 4.80- 5.19 6.00- 6.39 62 27 11 3 18 6.40- 6.79 5.20-. 5.59 68 15 1 9 5.60- 5.99 6.80- 7.19 54 6 4 4 7.20- 7.59 6.00- 6.39 30 14 1 7.60- 7.99 6.40- 6.79 8 3 1 1 8.00- 8.39 6.80- 7.19 4 5 7.20- 7.59 1 8.40- 8.79 3 6 8.80- 9.19 7.60-. 7.99 2 7 9.20- 9.59 8.00- 8.39 3 3 9.60- 9.99 8.40- 8.79 1 3 8.80- 9.19 10.00-10.39 1 10.40-10.79 9.20- 9.59 2 9.60- 9.99 10.80-11.19 1 11.20-11.59 10.00-10.39 1 11.60-11.99 10.40-10.79 1 10.80-11.19 I 3 11.20-11.59 1 11.60-11.99 I fx' 12.00-12.39 2 Anthnietic in 14.00-14.39 1 Mt'clian 14.40-14.79 I Standard (1UVi 14.80-15.19 1 Coefficient of 15.60-15.99 1 variation 16.40-16.79 1 20.80-21.19 1 teres Total 463 463 463 463 463 2fx 2242.685 seas 1723.485 1241.085 1131.855 1464.685 fx2 11425.521575 9348.233575 3955.217575 2972.629575 5145.621575 B Arithmetic mean 4.84% 3.72% 2.68% 2.44% 3.16% peri Median 4.83% 3.04% 2.63% 2.44% 3.05% Standard deviation 1.10% 2.52% mea Coefficient of 1.17% 0.67% 1.05% -an variation 0.23 0.68 0.44 0.27 0.34 78 FREQUENCY DISTRIBUTIONS TABLE 9, concluded Postwar Period, Januaryl925-December 1938 N.Y. Great Brstajn France Berlin, Germany commerjnJ N.Y. call open market open market Class paper rate private money rate discount rate discount rate discount rate 0-.O.39 6 2 0.40- 0.79 87 10 57 o.8o-L19 21 56 20 9 1.20- 1.59 11 11 1 16 1.60- 1.99 6 20 2.00- 2.39 5 10 12 20 2.40-2.79 7 9 7 20 2.80-3.19 8 1 15 44 3.20-3.59 4 2 1 23 9 3.60-3.99 20 11 9 6 27a 4.00-4.39 23 18 31 9 3 4.40- 4.79 7 8 14 3 18 4.80-5.19 3 6 1 6 10 5.20-5.59 8 2 8 4 3 3.60-5.99 2 1 4 6 5 6.00-6.39 6 2 1 0 9 6.40- 6.79 2 2 14 6.80-7.19 4 2 8 7.20- 7.59 1 1 5 7.60- 7.99 1 5 8.00- 8.39 1 8 8.40- 8.79 2 1 8.80- 9.19 3 9.20- 9.59 1 9.60- 9.99 10.00-10.39 10.40-10.79 10.80-11.19 11.20-11.59 11.60-11.99 1 1 188 168 168 168 168 Ifs 447.160 464.380 409.560 518.360 804.760 If 1109.444200 2127.832200 1559.740200 1980.732200 4363.788200 Añthinetic mean 2.66% 2.76% 2.44% 3.09% 4.79% Median 2.62% 2.04% 2.13% 2.78% 4.42% Standard deviation 1.76% 2.24% 1.83% 1.51% 1.74% Coefficient of 0.66 0.81 0.75 0.49 0.36 variation

terested in it, this gives at the same time a check on the nature of the seasonals and on the consequences of their elimination. Beginning with the New York commercial paper rate (prewar period), we find only slight differences between the arithmetic means-4.84 per cent (uncorrected) and 4.87 per cent (corrected) -and the medians-4.83 per cent (uncorrected) and 4.86 per cent 79 SHORT-TERM INTEREST RATES

TABLE 10 Frequency Distributions Short-Term Interest Rates Prewar Period, January lS7GJuly 1914 Seasonally Corrected Data (percent)

N.Y. Great Britain France erlln, commercial N.Y. call open market open market Class Pnvatc paper rate money rate discount rate discount rate rt 0- 0.39 0.40- 0.79 13 0.80- 1.19 12 29 2 2 1.20-. 1.59 31 36 32 1.80-1.99 48 47 84 2.00-. 2.39 46 69 105 2.40- 2.79 4 35 59 125 2.80- 3.19 11 52 63 63 3.20- 3.59 104 25 37 53 28 3.60- 3.99 45 41 34 13 51 4.00-4.39 66 38 31 4.40- 4.79 9 69 29 19 2 4.80- 5.19 75 22 6 14 5.20- 5.59 81 19 15 2 g 5.60- 5.99 39 11 6.00- 6.39 1 7 24 8 1 8.40-. 6.79 13 6 6.80-. 7.19 6 7 7.20- 7.59 2 1 7.60-. 7.99 2 8.00- 8.39 5 8.40- 8.79 4 8.80- 9.19 3 9.20- 9.59 1 3 9.60-. 9.99 10.00-10.39 1 1 10.40-10,79 10.80-11,19 1 11.20-11,59 1 11.60-11.99 1 2 14.80-15.19 1 16.00-16,39 1 LI.O()-1U.JJ z 463 46.3 463 fx 463 46.3 225&085 1717.085 124.3.485 11489.497575 1132.285 1472.285 Arithmetic mean 8582.457575 3870,793575 4.87% 3.71% 2958,225575 5080.585575 Median 2.69% 2.45% 3.18% 4.86% 8.28% 2.65% Standard deviation 1.01% 2.42% 3.11% Coefficient of 2.19% 1.08% 0.21 0.59 0.84% 0.95% variation 0.40 0.26 0.29

80 FREQUENCY DiSTRIBUTIONS (corrected).4 It is not of course to be expected that seasonal adjustment will affect the general level of the series;it may however reduce the series' variability, and the standarddeviations indicate that this happens: they are 1.10 per cent(uncorrected) and 1.01 per cent (corrected). This reduction invariability is still more marked in the NewYork call money rate. The arithmetic means of 3.72 per cent (uncorrected)and 3.71 per cent (corrected) are substantially lowerthan those for the commercial paper rate. This is not surprising in view of the nature of the tworates (e.g., in the matter of collateral, duration ofloans, etc.), but even this American rate is considerably higher thanthe average of the highest European rate. This point will reappearlater (cf. Table 26). The coefficient of variation5 o/X is almost three times ashigh for the call money rate as for the commercial paperrate (uncorrected data). With the elimination of the seasonalsthis difference becomes smaller. But it is true of all the rates in Table 9that the elimination reduces the coefficient of variation,the standard deviation of which is 2.52 for the uncorrected seriesand 2.19 for the corrected series in Table 10. The New Yorkcall money rate and the Berlin private discount rate showed the greatestreductions. Before the war the French openmarket discount rate had the lowest mean and median, the NewYork commercial paper Tate the highest. In the interwar period,the British rate was the lowest, both the Paris and Berlin rates risingrelative to prewar, the latter now becoming thehighest. The two New York ratesaveraged lower than before the war. The sensitivity of the rates, as indicatedby the standard deviation and coefficient of variation, showedthe New York call money rate the most sensitive before World WarI, the French rate the least sensitive; this latter evidenceof greater stability of theFrench economy will reappearfrequently in this study. Even in the interwar period this relationship continued,6although all rates varied more than before 1914. The British rate wasthe second most sensitive in both periods, evidenceof its central position in theinternational

4These are obviously very line distinctions.Their value dependsas in the following instanceson the accuracy ofthe basic data. It is possible that it is not very good, but error estimates arenot available. This coefficient is usually writteniOO(a/X), i.e., expressed as a percentage, instead of a ratio. In order not to confusethis percentage with the unit of the interest rate, which is also apercentage, we use the ratio form. Except that the Berlin coefficientof variation became the lowest. 81 SHORT-TERM INTEREST RATES

money market, as a focal pointfor major Jinancial infiuencfrom abroad.

Section 3. Seasonal Variations Now we turn to a brief examination of the seasonal variations of our interest rates. Here we touch upon an important and tricky question. There are many methods by which to calculate and eliminate seasonals. Frequently this is done in an altogethertoo mechanical way. Criteria of what constitutes a satisfactorysmooth ing are for the most part merely qualitative, even visual,which are quite inadequate in view of the deep-lying problems in- volved.7 The merits of each method must be judged fromthe problem on hand and from the properties of the series.Sometimes it may be doubtful whether seasonals should he eliminatedat all, depending on the use to which the corrected anduncorrected series are to be put.8 If it is a question, e.g., of comparing therelation of turning points of various series, and if seasonalcorrections shift the original turning points to different monthsas they often do, then a very troublesome situationmay arise. It is hard to make general decisions, if not impossible. For that reason we shall sometimesuse corrected data, sometimes uncorrected, oreven both. This will be determined by the needs of the occasion. An importantaspect is also the speedofthe repercussions which may occur betweentime series. Finally there is the vast and largely neglected field of the interdependenceof seasonal variations of differentactivities. This is of importance in international financial affairssince, for example, the seasonal harvests of one country cause seasonal changes inthe exchange rate and interest rate ofanother county,e.g., Argentina, Great Britain. Similar interdependenceis found domestically. The seasonal variationsof our interest ratesare best characterized by their indexes Chart 4 and Table 11.Considering first theprewar period we note a common property: Therates are generally highest during November and December, lowest in Juneand July. There are exceptions to this over-allpicture which showat the same 'Cf. in particular thework by A. Wald, "Berechnung a Saisonschwankungen" in und Aussehaltung von Beltrage zurKonjunktur,orscliung Vol.x, Vienna, ac 1938, whIch is theprofoundest study of the 'Cf. 0. Morgenstcrn, subject yet. See footnote 34, Chap.!. an On the Accuracyof Economic Observations, Princeton University Press, 1950,for a discussion of th curacy o the data. the relation ofcomputation and ac-

82 SEASONAL VARIATIONS time a certain amount of instability, provided the indexes are accurate. These are of two kinds. First, there is the exceptional July value for France for 1882-1898, the exceptional June value for Germany for 1884-1895, and the exceptional April value for the United States call money rate for 1898-1911; they might be due to an imperfection of the method or of the data. Second, there is the greater parallelism between the British and French and between the German and American indexes. The latter are frequently below 100 in the first three months of the year while of the first pair there is only one small instance (March for France, 1882-1898). There may not be anything deeper behind these relationships; at least we have not been able to find any clear reasons for the groupings, and the two correspondences may be due to chance. However another phenomenon is definitely not random and of greater significance: it is the difference between prewar and postwar indexes for all countries. The change can be seen in thetotal deviations as well as in the distribution of the monthly deviations. We know what caused the shifts andfor Germany and the United Stateseven the entire disappearance of seasonal deviations after 1928 and 1929 respectively for the commercial paper rate (1931 for the call money rate): it is the greater control over the money market exercised by central banks and treasuries. This is achieved either by direct contraseasonal intervention, or by the deliberate creation of an abundance of short-term credit at certain periods of the year, or both. From then on these institutions could do as they pleased regarding seasonal variations,and since the latter are generally consideredundesirable the decision was obvious. It will be noted that the decrease in seasonals and their final disappearance in some instances coincides with the period of a weakened application of the principles of the gold standard (with the gold exchange standard in some cases) and their final abandon- ment from 1931 onward. But from this it does notfollow that a strict gold standard requires wide seasonal fluctuationsof interest rates. Agreements between central banks caneasily be reached to abolish them within the givenframework. There exists a connection between the seasonals of these series and those of the exchange rates which ultimatelyreflect seasonal activities in foreign trade. This is particularlythe case if we have an industrial country onthe one hand and an agricultural one on the other and if these two are trading directlywith each other (e.g., as mentioned above, GreatBritain-Argentina). The extent 83 CHART 4 Seasonal Indexes of Short-Term Interest Rates, Prewar and Postwar, 1871-1939 Great BrItain: FrancE Germany: London Open Market Paris Open Market Berltn Priote OL5COUflt Rote Dtscount Role Discount Rate 120 430

140 110 1873-81 120

130 100 110 1876-93

120 90 100. I 1816-98 110 eo 90

100 90

90 70

80 130

TO 120 1884-95 60 110

100

130 120

120 110 1895-1902 1899-1913 1899-1913 110 100

100 90

130- 70 120 -

110

100

130

120 lb 1925-31 o I, 100

tO 0 JF eo .1 FMAu.1.1* SQN JFMAMJJiliili ASONOtiii CHART 4, concluded

United States: United States: New York New York Comm.rctaI Poper Rate Cofl Money Rote lao 140

110 130 \7785 1871-85 10 120

90 Ito

8 100

120 90 t1 a

12 It- 1886-9?

12 7

1898-1911

III

11 91

a 120-

11 12 1912-18

11 0 0- IL:,Z23

0- 1922-33 - 01

It 1924-31 to- 0I

I iIIii I i II I i iJ I ii I I I I JEMAMJJASOND JFMAMJJAS0N0 Source: Table 11. SHORT-TERM iNTEREST RATES

TABLE 11 Seasonal Indexes of Short-Term Interest Rates

GREAT BRITAIN YRANCE 1878.- 1899- 1873- 1882-1899- 1876-. 1895 owms 1898 1913 188! 1898 1913 1883 1895 1902 1913 Jan. 111 110 110 112 117 107 TI(jF- Feb. 113 103 103 101 102 77 7 87 Mar. 101 101 100 93 100 79 85 93 193 Apr. 84 91 101 101 104 87 80 87 May 88 87 98 92 94 89 89 90 91 June 70 81 88 90 89 96 102 96 July 62 18 87 103 87 93 83 85 Aug. 86 86 87 89 79 88 83 868781 Sept. 100 95 87 88 89 120 111 104 193 Oct. 116 115 109 101 109 123 121 112 113 Nov. 134 127 115 112 116 125 127 124 Dec. 134 128 117 112 115 116 142 133 131 Total deviations 219 164 110 96 125 182 208 152 151 1914-1934- 1925-. 1932- 1925-1931' 1933 1939 1931 1938 Jan. 100 87 106 92 96 Feb. 101 85 100 110 91 Mar. 99 94 96 113 94 Apr. 96 97 92 114 F May 95 94 90 112 99 M June 95 118 89 121 A July 102 100 100 93 99 104 Aug. 101 97 101 90 103 Sept. 102 98 102 85 jti Oct. 103 104 104 106 84 106 Ai Nov. 104 96 113 89 103 Sc Dec. 104 131 112 92 104 0 N Total de- D viations 31 95 60 137 50 T Uncorrected data used for 1932-19.39.

86 SEASONAL VARIATiONS

TABLE 11 (concluded)

U.S. Commercial Paper Rate 1877 1886 1893-.. 1899 1904 1909 1922-.- 1933b MONThS 1885 1892 1898 1903 1908 1914 jVo3 ioo 89 100 111 94 101 Feb. 98 90 88 90 103 88 99 Mar. 100 96 102 96 107 89 101 Apr. 100 111 94 95 89 101 May 89 91 94 89 91 89 98 june 84 87 82 87 86 89 96 july 82 91 92 93 90 98 96 Aug. 103 103 114 98 94 111 97 Sept. 109 110 127 110 104 114 101 oct. 113 113 113 114 105 118 104 Nov. 111 111 95 112 106 115 104 Dec. 108 114 93 117 109 106 103

Total deviations 94 101 134 106 89 128 29

U.S. Call Money Bate 1871 1886' 1898_l 1912' 1919i 1924' 1885 1897 1911 1918 1923 1931 Jan. 121 96 113 92 96 95 Feb. 91 78 77 97 104 97 Mar. 120. 94 98 95 100 94 Apr. 115 94 104 90 97 94 May 87 88 78 87 96 99 June 71 71 73 95 103 102 July 64 74 78 95 101 100 Aug. 68 100 65 91 92 99 Sept. 90 123 92 107 94 105 Oct. 111 120 101 112 107 105 Nov. 121 127 141 120 107 98 Dec. 133 131 113 119 103 112

Total deviations 204 202 271 116 50 48 b Uncorrected data used for 1934-1943.'Starting April 1919. Starting June 1885. 'Starting May 1924. Starting July 1897. h Starting June 1931, the original data were used. Starting February 1912.

87 S SHORT-TERM INTEREST RATES

to which this is reflected depends on the quantitiesiflvolvcl on the degree to which their ecoflornies are interlockJ.o

Section 4. Domestic Cyclicalehavior After the elimination of their seasonal variations theSpecifr cycles are determined for our four respective interestrates'o A can he seen from Chart 5 and Tables 12 and 13, thereare niarke cyclical fluctuations often withconsiderable amplitiide. occur in all series, but the rhythm is differentfor the countries. Before specific cycles are examined, Variots thequestion must be raised whether thereare fluctuations other than cyclical. seasonal First, there is the question of long-runtrends. These well known, of great rarity are, as is in interest rates, butthere is at least one instance here which vi1l bemore fully dealt with in IV, i.e., the United States Chapter commercial paper rate. Thetrend is weak influenceon monthly data, and if the a determined from specific cycleswere a series where the trend hasbeen eliminated in the traditionalmanner,'1 it would Probably haveno influence upon the turning points of interestrate cycles, which interest at present. are our only Second, randomfluctuations areconsidered in the selecting specificcycle turning points. process of This is done ina qualitative way on the basis ofknowledge about the background of the institutional, historical series. It must beadmitted that especially in a studyof international sometimes, turbances can he relationships, randomdis- extremely importantand more revealing the ordinarycycles. than Third, there might conceivablybe minor cycles, swallowed up by the which are specific cycles.According to the of a specificcyclet2 certain definition minor fluctuationsarc not taken into account. As a rule theyare not very interest rates. Our noticeable. So it is withthese purpose is tocompare their cyclical behavior. flw seasonal of the six exchangevariatiotis of our interestrates arc nuy tinies rates between the as large as those accompanying discussion. four coulntrjq.sCf. Table 42 and the '° discussior, of the technique bywhich this is Mitchell, op. cit.,pp. 56-71. The concept achieved is in Burns and a considerableamount of qualitatjv itself is simple; thetechnique iiivokes value of all subjective quantitatj measurements based jiidgnient. This limits the "This is not theway in which Specific upon them. "Cf. Burns and cycles are obtained. MithhJ op. cli.,pp. 57-58. 88 Turning Point Dates of Specific Cycles of Short-Term Interest December 1938, Seasonally Corrected Data TABLE 12 Rates, January 1876July 1914; January 1925 Peak DISCOUNTOPEN RATE MARKET LONDON Trough Peak DISCOUNTOPEN RATE MARKET PARIS Trough Peak BERLINDISCOUNT PRIVATE RATE Trough Peak COMMEBCL4.LPAPKE RATENEW YORK Trough Peak NEW YORK CALLMONEY RATE Trough JulyAug.Dec.Aug.July 18831878 189018991883 Sept.Oct.Oct.Sept. 18781879 189418851892 Oct.Jan.Aug.Mar. 187918821888 18911893 JuneApr.Sept.Feb. 18781892188918881879 Feb.Apr.MayFeb. 1885188218871877 1890 Jan.Sept.Feb.May 1888 1884187918761886 Dec.JulyJuneMayOct. 18771887188318901880 Aug.MaySept.June 1878 188918811885 Sept.Feb.Aug.Jan. 18781876188118791882 Jan.MayOct.Sept. 1880188218781876 1885 Apr.Nov.MayJuly 190319101907 1913 Sept.Oct.Mar. 190119041908 1911 Aug.JulyJan.Aug. 190019031907 1913 Aug.Sept.May 190119041895 1909 Aug.Aug. JulyAug.Mar.1893 1896 190719001903 MayApr.JulyMayDec. 1892 1902189719051894 Apr.Oct.JulyAug.Mar. 189618931898 19001903 JuneMar.Jan.Apr.Oct. 1892 1899189418971901 Oct.JuneSept.Oct.Aug. 189918931887 189619021890 JulyJulySept.Oct.MayAug. 1,897 19041894189219001888 May 1925 July 1928 Aug. 1928 Nov. 1925 MayFeb.Aug. 191319291910 JulyApr.July Nov.1928 19111914 1908 Oct.JuneDec. 1926191019131907 Nov.JulyJan. 1905190919111927 Oct.Feb.Dec.May 1907 192619101912 Nov.Sept.Nov. 1911 19081927 Jan.Oct.Dec. 19341929 1931 'After May 1931 original data used. JuneOct.Aug. 19801938 1933 JuneJulyDec.jan. 1929 193319371935 Dec.Nov.JulyOct. 1936 193219271934 Aug. 1931 Aug. 1930 Aug. 1929 Sept. 1931 Dec.Mar. 1931'1929 May 1931 CHART 5 Great Britoin, London open market discount rote Timing of Specific Cycles of Short.Term InterestPrewar and Postwar, 1876-1914, 1925-1938 Rates. Germany,France, ParisBerlin open open market market discount discount rate rate United States,I New York commercial paper rote I I I I I I I I I I I I I United States, New York call money rote 1q76 77 '7$ '19 80 81 '82 '83 '84 8 '86 CHART 5, corPuded'8? '08 '89 z7'90 '81 '82 '93 94 '95 '80 '97 '9$ '88 1900 '01 '02 '03 Great Britain, London open market I I discount rate I I CHART 5, concludedI France, Ports open market discount rote Germany, Berlin open market discount rate United States, New York commercial I I I I I I I paper rate I I United States, New York call money ---- i rote '10 I '11 jI '12 I '13 I '14 '2 '26 '27 I '28 I '29 I '30 I '31 I '32 '53 '34 '35 '36 '37 __.I. I '38 1902 '03 '04 '05 '06 '07 I_I '08 '09 correspondence between betweenline,threeSource: no countries correspondence. the Table U.S. In 12. contraction;and eliFor threeFor the the three United European States, countries, the three types European countries. dash line, correspondencethe in heavy expansion; ,olid line representsof lines indicate correspondence or no and the ne solid correspondence S SHORT-TERM INTEREST RATES

TABLE 13 Number and Duration of Specific Cycles of Short-Term Interest Ba Four Countries, Seasonally Corrected Datj of

AYEiLi AVERAGE DURATION OF' PER NUMBER OF: IXJRATIONO, (months) SPEj-. cy Expan- Cont rac- Full Expan- Contrac-j1 sions lions cycles' slons lions cycles Prewar, February 1878-August 1913, 426months Great Britain 8 8 7 84.9 17.5 52.6 France 9 8 8 29.6 20.0 Cermany 47.5 597 11 10 10 20.5 17.9 U.S. commercial 38.6 53A paper rate 12 11 11 20.2 15.9 U.S. call money 38.6 56.0 rate 12 11 11 19.1 16.5 38.2 53 Postwar, January l925-December1938, 167 months Great Britain 3 4 3 11.3 30.7 39.7 France 4 3- 100 26.9 731 . io.o a3.2 Germany 2 2 41.6 3&4 2 2.3.0 51.5 74.5 U.S. commercial 30.9 ,i paper rate 3 3 2 12.0 33.3 550 26.5 73. 'Thiration. of only completecycles, expansions, and tive columns contractions are included in their b The re. percentages are calculated froma base equal to the average duration that of contractions. Thissum is equal to the average duration of espansionj of full cycles is thesame as the number of expansions of cycles only whenthe nuinie 'Only complete cycles and contractions. are counted; parts of cyclesat both ends of a series are measared from troughto trough. Similarly only are dmppej Cydr counted in theirrespective columns. complete expansions andcontracti

This means obviouslythose cycles which to the ordinary are related by duration business cycle. Thatis precisely what the cycles purportto be. Therefore specific there is no needto look for smaller cycles here. Withthis remark we do not rule themout altogether. On the contrarywe shall introduce the another notion of short cyclesiii connection where itwill prove helpful certain possible in elucidating causal relationships(cf. Chapter VI). We begin thediscussion of the tion of their specific cycles witha considera- number andduration, making about theprewar and post-World separate statements we have two American War I periods(Table 13). Since series we referin the followingto American 92 DOMESTIC CYCLICAL BEHAVIOR

TABLE 14 Phase Comparison of Specific Cycles of Short-Term Interest Ratesand Their Respective Reference Cycles Four Countries

SAME PHASE IN: SAME PHASE IN: Expan- Contrac- DiFFERENTExpan- Cant rae- DIFFEFENT sions tions Total PHASE sions tfons Total PHASE (months) (percentages) Prewar, February 1878-August 1913, 428 months 17.6 Great Britain 234 111 351 75 54.9 27.5 82.4 18.1 France 207 142 849 71 48.6 33.3 81.9 18.1 Germany 192 157 349 77 45.1 36,9 81.9 United States 155 106 261 165 36.4 24.9 61.3 88.7 U.S. call money 177 142 319 101 41.5 33.3 74.9 25.1 Postwar, January 1925-December 1932, 96 months 37.5 Great Britain 10 50 80 36 10.4 52.1 62.5 20.8 France 35 41 76 20 36.5 42.7 79.2 75.0 25.0 Cennany 33 39 72 24 34.4 40.6 39 51 90 8 40.6 53.1 93.8 8.3 United States 21.9 U.S. call money 28 47 75 21 29.2 49.0 78.1

call money rates in parentheses while theunbracketed figures refer to the commercial paper rate. Beginning with the prewar period, wefind notable differences for the four countries. There are 7, 8, 10, 11,(11) cycles respectively for Great Britain, France, Germany,United States. Their average duration necessarily varies, and in this case in inverseorder to the number,'3 from the maximum of 52.6months to 36.3 (36.2) months. There are more cycles than referencecycles (cf. Table 6), Great Britain and the United States eachhaving one more full specific cycle, France two, and Germany four morethan reference cycles. This does not mean necessarily thatthe correspondence between the specific cycles of the short-term rateswith their own reference cycles would be found highest in theUnited States, and lowest in Germany (cf. Table 14). The difference among countries inthe average length of full specific cycles is considerable. It istherefore noteworthy that the differences among the duration ofthe contractions is not nearly '3This need not be so, of course, sincethe periods covered by complete cydes in each series are not precisely the same. 93 SHORT-TERM INTEREST RATES so great. It varies only between 15.9 months (United States) 20.0 months (France), while the values for Great Britainand Germany are practically identical: 17.5 and 17.9 months. and consequence the dispersion of the average duration of a expansio is much greater. No relationship of this sort could heobserv for the reference cycles of the various countries. The average per cent duration of the two phases (adding up to 100), computed from the above figures, shows forGreat Britain almost twice as long upswings as contractions of the short.ter rates; this is a more marked version of the same relationship found in the British reference cycles. In all countries,however the expansions were longer than the contractions. In the absence of trends this has to be interpreted as indicating that gradual rises of rates were briefly but sharply interruptedby steep falls rates. The long rise and sharp fall (best of seen from Chart 5)seems to be a characteristic of freely moving short-termrates. The post-war situation is widely different. The periodcovered is much shorter, and the three full cyclesfor Great Britain France and the two for Germany and and the United Statesare too few from which to obtain significantaverages. But in spite of the necessary caution we can point out thatnow the average per duration of contractions is larger cent than that ofexpansions. The figures for Great Britain and the United States (commercialpaper rate only) are almost identical; bothshow the greatest of contraction proportion over expansion. It is clear whatthis general shift means: on the one hand there was a drop from the exorbitanthigh rates in a country like Cermanyas an after-effect of and inflation; a lost war on the other hand therewas the deliberate policy of reducing interest rates,coupled with a genuine funds in the United increase in liquid States, especially in thel930's, and demand for commercial a lack of paper. In some instancesthe policy of depressing interestrates was even understandings reached internationally coordinated by among central banks. Whilethe postwar reference cycles hadalso shown of the whole cycle significant changes in duration or parts of it, theywere not nearly as pro- nounced, and there themeaning is quite different. France occupies a somewhat singularposition in the general picture, Although theretoo the reversal of the average duration the relation between of the two phasestook place, it limited (41.6per cent for was narrowly expansion and 58.4per cent for contraction). Even beforethe war there difference between was comparatively little the shares ofthe two phases(59.7 and 40.3 94 DOMESTIC CYCLICAL BEHAVIOR per cent). The French rate, as we know from the frequency distributions (ef. Table 9), had not only the lowestaverages hut also the second smallest coefficient of variation of all of them (in the prewar period) and this would tend to be reflected in our present statistics. We do not want to attribute too much significanceto these figures, but it will be seen later again and again that French financial affairs in many ways showed greaterautonomy of movement than those of other countries. It is important now to study the behavior of the short-term interest rates relative to the reference cycles of the same country. We want to find out whether the conformity of movement, i.e., of turning points, is the same in all four countriesor whether any significant differences emerge. Table 14 contains the result of the comparison. The figuresex- press, either in months or percentages, how the reference cycle and the short-term rate were or were not simultaneously in the phase of expansion or contraction.14 For the prewar data it is found that the three European countries show the greatest total covariation between the respective national reference cycle and short-term rate. The values, 82.4 per cent, 81.9 per cent, and 81.9 per cent for Great Britain, France, and Germany are almost identical. For the United States we get less correspondence: 61.3 per cent (74.9 per cent). That the call money rate has greater covariation is noteworthy especially since in the postwar period just the opposite was the case. The great role played by the call money rate in the United States before 1914 is well known and this is partly reflected in better correspondence of movement. The post-World-War I data need be dealt with only briefly. This is partly justified by the much smaller sample,'5 although the results vary far more than could be attributed to this difference in samples. Very large changes took place, an indication of the enormous influence the war exercised. It seems safe to assume that it is not only a post hoc but a propter hoc. The American commercial paper rate, which before the war

In the following sections the correspondence of the movement of pairs of short-term rates will be investigated. It is clear that if two rates correspond well to each other and each shows a high degree of covariation with its respective reference cycle, then we have also high covariation between the first interest rate with the reference cycle of the second and vice versa, i.e., there is transitivity in these relationships. The significance test of section 2 in Chapter II applies the necessary changes to the present situation. 95 a SHORT-TERM INTERESTRATES showed the least correspondence,afterward was in the with American reference cyclesin not less than same Phase 93.8 perC all months. For the callmoney rate the improve,ne of spondence is minor. incorfl. Most interesting is the change for Britain after thewar. There we find the lowest instead of thehighest per cent, just slightly higher than the corresptdefl.625 lowest ewar figure per cent, United States commercialpaper r..tte). But the (61.3 interesting fact is that only.10.4 per cent (out ofa total of 62.5 of all the months for per cent) which the Britishshort-term rates reference cycles and British were in the same phase,were they both ing (before 1914: 54.9per cent out of a total of expand. This is in 82.4 percent). agreement with all other series;while before pansion dominated without 1914 ex- exception, contractionnow domjnatj The mosteven distribution of the two phases is for Germany. correspondence betveenthe To some extent leads and lags affectedthe degree of The New Yorkcommercial paper rate agrceme United States tends to lagbehind the reference cycle. TheNew York cai showed sometendency to lag money rate at the peaks before1914, vher after World WarI it tendedto lead at the peak. War I period the In thepre-Worid. London short-termrate tended to lead ence cycle at thetrough, while the Paris the refer- It would be rate lagged at thepeak. interesting to discoverwhether the betweenagreement in expansions complete reversal andcontractions was due central bank policyaiming at reducing to interest rates duringbusi- ness depressions (atleast doingso faster or earlier and keepingthem stable during than before) tend to loosen businessexpansions. This theconnection between would Specific cycles. reference cyclesand these As pointed out elsewhere,short-term interest a large role inmost business rates have played cycle theoriesand are still of importancethough decreasingly considered cal gold so. For the periodof the classi- standard theywere often held cance, notably in to be of primarysignifi- the earlierwritings of R. 0. clear thatduring the Ilawtrey. Now it is prewar period insimilar interest ratesbehaved differently Countries, similar to the at differenttimes and in national businesscycles.18 respect Increasing thenumber of cycles If we hadtaken the shown up amplitud into ac(1J,Int still more. The the diversitywould have amplitudes would coefficient5 of ShOW thesame. variatiol) make thislear. Cyclical 96 DOMESTiC CYCLICAL BEHAVIOR generally available for establishment of a theory to 36 (before 1914 but 46 up to 1938) does not make easier theoretical inter- pretations. On p. 52 it was pointed out that a strong probability of international transmission of the fluctuations as shown by reference cycles might be inferred from the maintenance of a high degree of parallelism when there was an alternation between short and long cycles. This worked only for the European countries. The same question arises now about specific cycles. Chart 5 and Table 12, showing the specific cycles of short-term interest rates, indicate the same conclusion, with certain limitations. To begin with, the good correspondence does not really set in before 1892. Then the first trough of a short specific cycle begins, the cycle lasting to the next trough in 1894-1895. In this case even the American cycle is in step. The next cycle is long, lasting to 1901- 1902; the three European countries are again in high correspondence. Then a short specific cycle isfollowed by one more long specific cycle. When the American cycle is in agreement with the others, it is quite clearly a phenomenon of mere coincidence, because in general it seems to follow its own,quasi-independent rhythm. The measures of concentration and dispersion of turning points below (pp. 110 if.) make these statements sufficiently accurate, without however showing fully the change in pattern here mentioned. The case is not nearly so clear as for the reference cycles; this is an interesting observation, in consonance with the fact that the international covariation of short-term rates is not so marked as that of the reference cycles.Apparently these rates are very delicate and therefore sensitive to many different influenceswhich come both from several other countriesand from purely local conditions. Thus they do not give much indication of anydirect influence of one national rate on another. There may howeverbe a very powerful interaction workingthrough a separate medium and involving other economic quantities. If this be the case it may even be unwise to look for repetitionof the pattern of the reference cycles, because dissimilarity rather than similarity, or atleast systematic displacement in time, might have to beexpected among these rates. This concludes the comparison of the cyclical behavior ofshort- term interest rates with the reference cyclesof their respective countries. A certain degree of covariation was tobe expected, 97 SilORTTERM INTEREST RATES because interest rates in general should be givensome weight obtaining the average which constitutes i the referencedates. But the covariation is high, thoughit differs among the Countries and was subject to sharp modification after thewar. The differences were probably much larger than could beattributed to in the institutional setup. Thereremains the possibilitydiflerences interest rates in the various nations that these were not independentof each other. There may havebeen influence ofone on the other vice versa. Such international and transmission is notobvious, cannot simply be read off from i.e., it our charts. On the otherhand know in a generalway about the interlocking we capital markets. of moneyand The reader will recallwhat was said inChapter 1, section the possibility that 8 about some parts of the nationaleconomy are closely related to thesame or other parts of other more to different spheres ofthe national economies than be explored economy. This possibilitywill now; it will crop up repeatedlyin later chapters.

Section 5. InteruotionolCovariation We now take up the direct phasecomparison of the our short-term interestrates, first for cycles of for the six pairs, groups of countriesand then in the samemanner in which reference were compared in Chapter cycles II. Our discussionshould be read the light of theresults obtained in T1r test of section 2 there and withthe significance in mind. Justas was then the covariation does not case, high or low in itself meananything for the of transmission.There could simply wider question action of be parallelismwithout inter- any kind. Clearlyeven if purely random produce a high factors could parallelism of the Four we would be bound reference cycles oftwo countries, to get someparallelism of the series, if thesemove at all in constituent cycles. That they correspondence withtheir reference 8. do somove was just sequent analysis demonstrated. Soour sub- neglects at firstthe problem 1. it will then have of transmission.But to be decidedwhether similarmovements or not. we have independentbut T Themeasurements are containedin Tables 15 1 less cyclicalparallelism in the to 19. There is in that of movement ofour interest rates than ]O. 1 reference cycles,i.e., Tables 11. F lion still 3, 4 and 5. Butthe covaria- remains high and,subject to the hardly beconsidered above qualification,can random. Somenew and rather a Ft' peculiar bN1 98 INTERNATIONAL COVARIATION features emerge which,while difficult to interpret, are unmistakable. They possibly point toward elements of astructural relationship which only a very extensive statistical studycould fully reveal. This feature lies, as we shall see, inthe great uniformity of the percentage of the numberof months in which some of the countries Our aresimultaneously in the same phase, notably in contraction. measurements so far are insufficient toelucidate fully such a possible structural relationship, since theyhave not taken care of amplitudes nor sufficiently of the difficultproblems of the absolute levels of interest rates in differentcountries.1T The total phase correspondence-asshown in Table 15, which is based on Table 12-isdefinitely high. The four countries taken together are in the same phase in48.6 per cent of all prewar months. If the call money rate isused, the figure drops to 42.7 per cent. Omittingthe United States altogether, the percentage 'TAs stated before, we shall return tothese matters, especially in Chapter IX.

TABLE 15 Phase Comparison ofSpecific Cycles of Short-Term InterestRates

PREWAR PERIODS POSTWAR PERIOD" NumberPer cent Number Per cent of monthsof total of monthsof total

Three European countries 173 40.8 10 6.8 1. All three expand 20.0 50 33.1 2. All three contract 85 258 60.6 60 39.7 3. All three in same phase(1 +2) 91 603 4. In different phase 168 100.0 151 100.0 Total 426 Four countries expand, United 5. Three European countries 31.5 10 6.6 States expands 134 expand, United 8. Three European countries 9.1 0 0.0 States contracts 39 United 1. Three European countries contract, 2.9 8 5.3 States expands 12 8. Three European countries contract,United 73 17.1 42 27.8 States contracts 9. Three European countries in samephase, 51 12.0 8 5.3 United States in opposite phase (6 +7) 48.6 52 34.4 10. Four countries in same phase (5 +8) 201 219 51.4 99 65.6 11.FourcountriesindifferefltphaS(4+6+l) 426 100.0 151 100.0 Total February 1878.-August 1913, 428months. bNovember 1925-June 1938, 151 months. 99 SHORT-TERM INTEREST RATES S

TABLE 15, continued Phase Comparison of Specific Cycles of Prewar Short-Terni Interest i for Groups of Countries, February 1878-August 1913. 48montl1,

SAME PHASE IN: s.snIIIASE IN: Expan. Contrac- DIFFERENT Expan-Cmtr Di sions tions Total PhASE sions th, Total (months) (percensagej Great Britain- France 232 107 339 87 54.5 25.1 79.6 Great Britain- Germany 194 109 303 123 45.5 25.8 71.1 France-Germany 183 117 300 126 43.0 27.5 70.4 U.S.-Creat Britain205 103 308 118 48.1 24.2 72.3 U.S.-France 186 103 289 137 43.7 24.2 67.8 U.S.--Gennany 158 115 273 153 37.1 27.0 84.1 United States call money rate

Four countries 120 62 182 244 28.2 14.6 42.7 U.S.-Creat Britain 187 97 284 142 43.9 22.8 66.7 U.S.-Francc 164 95 239 167 38.5 22.3 80.s U.S.-Ccrrnany 153 123 276 150 35.9 28.9 64.8 U.S. call money- U.S. commercial paper 178 131 309 117 41.8 30.8 72.5

rises to 60.6. Of this, 20.0 per cent is accounted forby contraction. This agreement was somewhat lower thanthat for the reference cycles (cf. Table 3).18 In the postwar period radically differentresults obtain. Taking first the period November 1925-December 1931, all fourcountries are in the same phase 27.4 per cent of themonths and the three European countries show thesame correspondence, 27.4 per cent. This percentage is thesum of equal Components, of expansion and contraction 18.7 per cent, in both instances. Use of the callmoney rate weakens thecorrespondence still further to 24.7per cent. If the postwar period to June 1938 is used, thecorrespondence rises for all four to 34.4 per cent and for thethree European countries to 39.7 per cent. The percentages compared by pairs ofcountries in the prewar period vary from the maximum of 79.6per cent for Great Britain- France to the minimum of 641 per cent forthe United States- ' There wasno observable pattern in the Interest rates, which also leads and lags among the four accounts for the relatively highagreement. 100 INTERNATiONAL COVARIATION Gennany.19 In view of the generallyknown interaction of the British and French money markets thishigh percentage might be expected. We would hardlyassume a similarly close connection between the American and Germanmoney markets. In the first case there may have been much interaction, with sufficientlyrapid transmission of effects between the twocenters to cause parallel movements. In the second casewe have no particularly good intuitive reasons to assume a similarly high degreeof interaction. There would therefore simply be theexpectation of a parallelism based on the belief that parallelmovements occur in similar countries. Such a view would seem natural because ofthe already noted similarity of the reference cycles of thesetwo countries. For the longer postwar period,up to June 1938 (151 months), the size of the sample is twice that of the shorterperiod through December 1931, though even the larger sample isnot much more than one-third of the prewar. The changes fromprewar to postwar are so considerable that they can be attributed to influences other than the size of the sample; theyare furthermore consonant with the preceding discussion. The total covariation was considerably lesstaking the smaller sample firstthan in the prewar period,except for the United StatesFrance, where it is 87.7 per cent (Table 16),or 19.9 points higher than prewar and thus not only the highestpostwar covaria- lion (even if the longer period up to 1938 is also considered) but the highest in all periods. We have pointed outa similar phenomenon for the reference cycles of this pair of countries (cf. Table 4) and have noted that there was little known about post-World-War I economic relations, which would account fora highly parallel movement between them. This holds for the general business cycles. But short-term interest rates are presumably highly susceptible to international influences, so that the repetition of the correspondence here is noteworthy. The correspondence between Great Britain and France on the other hand, which with 79.6per cent was the highest before 1914, fell after the war to the second lowest and insignificant position of 50.7 per cent, only somewhat better than the United StatesGerman covariation of 45.2 per cent, which is below the 50 per cent probability level. It is not difficult to attribute to this change a real meaning in the light of thereasons for the high prewar correspondence and its profound disturbance "Confrast this, however, with the quite divergent ranking of the correlation coefficients of Table 21. 101 SHORT-TERM iNTEREST RATES

TABLE 16 Phase Comparison of Specific Cycles of Short-TermInterest Rates, for Groups of Countries, November l925-Deeeinber 1931, 73month SAME PHASE IN: SAMEPHASE IN: Expan- Contrac- DIFFERENTE77Taojr sions tions Total PHASE sions tions Total (months) (percentages) Four countries 10 10 20 53 13.7 13.7 27.4 Three European 72.e Cr countries 10 10 20 53 13.7 13.7 27.4 Great Britain- 72.6 Gr France 11 26 37 36 15.1 35.6 Great Britain- 50.7 49.3 Fr Germany 20 18 38 35 27.4 24.7 52.1 France-Germany 20 18 38 47.9 U.L 35 27.4 24.7 52.1 U.S.-Creat Britain 16 26 42 47.9 U. 31 21.9 35.6 57.5 U.S.-France 27 37 64 42.5 U. 9 37.0 50.7 87.7 U.S.-Cermany 12.3 20 13 33 40 27.4 17.8 45.2 54.6 United States call money rate Four countries 8 10 18 55 11.0 13.7 IJ.S.-Creat Britain 24.7 75.3 15 31 46 27 20.5 42.5 U.S.-France 63.0 37.0 20 36 56 17 27.4 U.S.-Cermany 49.3 78.7 21 20 41 32 28.8 27.4 U.S. call money_ 58.2 43.6 U.S. commercial paper rate 21 32 53 20 28.8 43.8 72.6 after the war. The cooperation of these hvomoney markets apparently sufferedmore than might have been drop in covariation inferred from the in general, as shown bythe reference cycles. For the longerpostwar period therewas an improvement the shorter in the over covariatjon in fourcases, a deterioration for United States-Great Britain and practicallyno change for France- Germany. Great Britain-Germany is in thelead with 73.5 cent. There was per a very close financialcontact between London and Berlin, namelythe huge British the Reich and then short-term investmentsin the great effortsto extricate them again mediately before and im- after Germanexchange control duced. Clearly high was intro- covariation under suchconditions has a different significance thanin other cases. Up months beyond to December 1931 (six German introductionof exchange control!) correspondence was only the 52.1 per centor virtually nil and before the war 71.1per cent.

102 INTERNATIONALCOVARIATION

TABLE 17 Phase Comparison of Specific Cyclesof Short-Term Interest for Groups of Countries, Rates, Postwar November l925-June 1938,151 months SAME PHASE IN: SAME PHASE IN: Expan.. Cotstrac- Dwsur xpan- Cont rac- DIFFERENT sions lions Total pa*sx (months) slons lions Total PHASE Great Britain.- (percentages) France 15 66 81 70 9.9 Great Britain.- 437 53.6 46.4 Germany 20 91 111 41) 13.2 France-Germany 60.3 73.5 26.5 20 59 79 72 13.2 U.S.-Creat Britain 39.1 52.3 47.7 16 86 102 49 10.6 57.0 U.S.-France 32 67.5 82.5 69 101 50 21.2 45.7 66.9 33.1 U.S.-Ceimany 20 78 98 53 13.2 U.S.-Great Britain- 51.7 64.9 35.1 May 1920-June 1938 (217 mos.) 23 108 131 86 10.6 49.8 60.4 39.6

Extending the postwar period of observationto 151 months thus leads in general tomore parallelism, though not as muchas for the prewar period. However, theimprovement over the results for the 73 months must not be attributedto the size of the sample, because of the obvious break in theinstitutional background against which our statistics must beseen. We do not know whether there were similar but hidden events during theprewar periods, which we consider homogeneous. At presentwe shall continue to assume homogeneity, which we infer principally fromthe fact that the principles of the gold standardwere applied and were violated only to minor degree (chiefly by the central banks throughapplication of the "small means" of gold policy,see Chapter VIII). In one respect the extension of thepostwar period is genuinely revealing. That is the change in distributionof the total percentages between expansion and contraction. Contractionis without exception overwhelmingly predominant,even more so than was the expansion in the prewar period. It is questionablethat, this is due merely to the fact that the depressionsaveraged longer after World War II. There remains a remark on the substitution ofthe American commercial paper rate for the New York callmoney rate. We note first the covariation between these series themselves.It is not veiy high (for these interest rates) though practicallyidentical in prewar 103 ,

SHORT-TERM INTEREST RATES and postwar: 72.5 and 72.6 per cent. The great interdependence of interest rates in the same money market (especially when both are short-term) would have led one to expect much higher percentages. In the light of this the values for the rates of markets of different nations assume greater importance.2° If the callmoney rate is included in the four-country comparison, values decrease. The same was true before the war for the covariation of France and Great Britain; also in the postwar period. The relationship (and its change) between the two phases is maintained. We add the measures for the call money rate because of a need for it later (cf. Chapter V) when, in the study of the interaction ofmoney markets, we come across the argument that this rate rather than the commercial paper rate provided the link with foreignmar- kets.21 From seasonally uncorrected data we have computed thecorre- iation coefficientsfor the short-term interest rates of our six pairs of countries (Table 18)22 Uncorrected datamay be expected to give higher correlation coefficients than corrected data (ifvery successfully corrected), if the seasonal movements affect thetwo time series simikrly This need not be the case,e.g., when the seasonal movements of one seriesare the exact opposite of those of the other. Table 11, which shows the seasonal indexes forour series (which we know must be cautiously used) should be consulted in on with Table l& The long period covered and the fairly frequent changes of the indexes probably reduce the significance of the seasonal movement. Another point of much wider interest deservesmention here. Seasonally corrected series sometimes show peaks andtroughs in places differing by one to several months fromthose of non- corrected series. This may occur for two seriesinversely if their Besldes the rates of diff&ant countries are n* Identical, In regard to theuou t,pe of traneactiori- to which they relate, but only"Imilar; this would Part Cl the different results for the callmoney rates arises froni its lead Over the cianmerdal paper rate,psrticeluzlyatthep Th time covered IsighiIy Mf.t from that used earlier. But thisIs immaterial. The number Of menths, 463, isa little larger than the other, which is 426. These are, Indeed, vy large zausbers for ounelaticicoethdenti and such differences disappear fully. At any ratewe ha'e deterinbied the cuifldeaice limits In all these cues. For methods used, d. ILCramer, Mdheesagici,J Meth- ods of Seetlet 1943,_pw -4I1 bw enpItlo. of coe6denoe limits forerdatim eo#&4'i' and ft A. Fisher, StetI1Method, for ResearCh Worker,, 10th Ed., 1943, Chapter vi TabJaiv for 1-distethution and Table v-B for tnun&mmetloanb.tw,,uif and,. iNTERNATIONAL COVARIATION

TABLE 18 Correlation between Short-Term Interest Bates, Seasonally Uncorrected Data

CONFIDENCE LD.f ITS' RATE Lower Upper Prewar Period, January 1876-July 1914,463 months London open market rate-Berlin private discount rate +0.73 0.61 0.78 London open market rate-Paris market discount rate +0.67 0.60 0.73 Berlin private discount rate-Paris market discount rate +0.62 0.54 0.69 N.Y. commercial paper rate--London open market rate +0.45 0.35 0.54 N.Y. commercial paper rate-Berlin private discount rate+0.40 0.29 0.49 N.Y. commercial paper rate-Paris market discount rate +0.36 0.25 0.46 Postwar, January 1925-December 1938,168 months London open market rate-Berlin private discount rate +0.84 0.77 0.89 London open market rate-Paris market discount rate +0.34 0.15 0,50 Berlin private discount rate-Paris market discount rate +0.13 -0.07 +0.82 N.Y. conunercial paper rate-London open market rate +0.93 0.90 0.95 N.Y. commercial paper rate-Berlin private market rate+0.77 0.68 0.84 I N.Y. commercial paper rate-Paris market discount rate+0.34 0.15 0.50 N.Y. call money rate-London open market rate +0.84 0.77 0.89 For confidence coefficient of 99 per cent. peaks or troughs are very near but do not coincide. In that case there would be a considerable shift and less phase correspondence than if the uncorrected data had been taken. If a comparison is made between the covariation of phases on the basis of seasonally corrected data, and correlation coefficients are computed from uncorrected data, the correlation coefficient may give a different re- suit because of the absence of such shifts (aside from other influences operating toward variation of results). The seasonal cor- rection of our short-term rates has, to be sure, frequently led to temporal shifts in peaks and troughs, but mostly in those minor fluctuations which occur between the turning points of specific cycles. Thus there would be no influence on covariations of cyclical phases. The issue becomes of importance only when highly sensitive actions and reactions are studied. The correlation coefficients provide an ordering of the various pairs of money markets. In view of the length of the series each one must be considered significant, and a significance test for their mutual comparison would not upset the ordering of Table 18. On 105 SHORT-TERM iNTEREST RATES top of the scale for the prewar periodwe find Great Brjtaj_ Germany with +0.73, followed by Great Britain_Francewith 4-0.67. This sequence is the inverse ofthat found byour phase covariations (Table 15). This is repeated forthe next two Here we have Germany-France, countries + 0.62, followed by UnitedStates_ (commercial paper rate only)-QreatBritain, +0.45, which wide gap. The total phase is a correspondence shows much lessvaria. tion. The smallest degree of relationshipis indicated for States-France, +0.38. United For much of the time after WorldWar I thereare no seasonal movements in our series, so the considerations aboveconcerning their common influence donot apply. The order of the now change, placing New York-London coelficiefib (for thecommercial paper rate) definitely first with0.93. The main result isthat the Londo. Paris relation has fallenvery low, the other coefficients lack of correspondence. showing Table 19 givesa sign correlation of the nique used in the following same series. The tech- correlations amountsessentially toa

TABLE 19 Sign Correlation:Short-Term Interest Rates of FourCountries Correlat Prewar and Postwar

Period Palsofcu,j N Z lag toC Dec. l87l-JuIy 1914Great Eritain.Fmnce Jan. l876-July 1914 511 283.5 0.02< 0.03 0.110 Jan. l878-JuIy Great Britain_Germany 462 271.0 0.00 3 1914 France_Germany < 0.0001 0.173 1 Dec. 18(39.-July 1914 462 257.5 0.02<0.03 United States_GreatBritain535 296.0 0.02 0.115 2 Dec. 1871-July 1914United States_France < 0.03 0.107 4 Jan. 1876-July 1914 511 274.5 0.12<0.13 0.074 United States_Germany 462 251.5 0.08 6 <0.09 0.089 5 Jan. l925-Doe. 1938 Great Britain_France 167 Dec. 1924-Dec. 1988 92.5 0.21 <0.22 0.108 5 Great Britain_Germany 1(38 Jan. 1925-Dec. 1938 93.5 0.18 <0.19 0.113 Germany_France 167 4 Dec. l924-Dec. 1938 85.0 0.83 <0.84 0.018 United States_GreatBritain 168 6 Jan. 1925-Dec. 1938 96.5 0.08 <0.09 0.149 3 UnIted States_France 1(37 Dec. 1924-Dec. 1938 102.0 0.01 <0.001 0.222 United States_Germany 168 107.0 0.01 2 N <0.001 0.274 1 number of pairs ofdifferences observed Z = number oftimes change in one-half). each series showed same sign (a zero changeis counted as p probability thatan observed deviation to chance. from theexpected value as great C or greater is due coefficient ofcovariation

106 INTERNATIONAL COVAR1ATION modified form of the familiar chi-square test. it was developed for the purpose of this investigation by A. Wald and has a number of advantages over the ordinary chi-square tcst, chief among which is that of easy computation. The method is described in the footnote.

Letnbe the number of time points at which two economic variables x and y areobservedLI.e., we have the observationsx1,...,x. and y, ., Consider the two sequences of differences:

X*Xt, X1XL, ..., ' F y, - yi,s - ... - - From (1) we derive the sequence (2) where z = 1 if sigmim (y + =signum (riri). If this does not hold, thenzL= 0. (We exclude the case that any of the differences * - and x.+ 1 Xmight be zero.)

Let

For largenthe statistic z is normally distributed if the observations on xand y form two independent random series.

SinceEZL= we have EZ

If- 2, thenZ,and z are independent. Hence the variance e' of Z is given by r' (-ni)a,+2(n-2)uz1z, , 1 Cs1

1 10 1 CZZI= Ez,z, - - 4 = n1 2(n_Z)_lInl3 Hence 36 - 36 ni EZ_ 2 lln-13 - 38 C=2(_2- uni3 Cv =9(ni)'' 107 SHORT-TERM iNTEREST RATES It can easily be shown that our coefficient of covariat C closely related to the familiarmeasure T=Nvx2 where N is the number ofobservations2l of differences between successive original data, i.e., exchange rates anddifferentials. Ccan vaiy from 1 to 1. T isa frequently used convenientmeasure of contingency (1'2 is knownas the mean square contingency) developed by Tschebycheffand Yule. It should be noted that, since both the sign correlationand the phase comparison measure only the direction of themovement of a series, they are more closely related thaneither is to the coefficient of correlation, which compares corresponding absolute figures oftwo series, regardless of the directionof the movement. As can be seen froma comparison of Tables 18 and 19, the ment is about the agree- same for the two correlationmeasures, the only significant shift being in theNew York relationships andto a lesser extent the Paris rate for theprewar years. We shall have furtheroccasion to discuss some correlation efficients and co- compare them with othermeasures of certain time series.some like those employedhere. The latter vital to our theme topic is more since it more directlyinvolves probable causal relationships betweeninterest rates. Here the tween the series is interdependence be- assumed to bea possibility without any hypothesis as yet Concerning how it operates. Thequestion must be left open, whether thedetermination of cyclical stresses only turning covariation, which points and duration ofcycles and phases, necessarily leads tomore divergent results than does relation. If such ordinary cor- a disparity does arise,as is the case here,a deci- Sion must be madeas to which of these we are more interested measures to prefer. Clearly in the cyclicalaspects and must therefore be inclined to viewthese measures as more adequate. Thisrefers to specific cycles, i.e.,fluctuations more nearly amplitude of the general of the length and business cycle.Correlation analysison the other hand takesinto consideration which by the also those 1)ossiblvminor cvcks definition of thespecific cycle series may have are discarded. Two strong opposite trendsbut closely harmoniotis correspondence This would phase reduce the correlationcocfflcints hut The n- 1 in footnote 23 equals N the original series. in the fonnula for T,since n refers to

108 INTERNATiONALCOVARIATION

TABLE 20 Sununaryof Reference Cycle and Short-Term Interest RateCorrelations London- London- Berlin-New York'- New York'- New Berlin York' Paris Paris London Berlin Paris Reference cycle phase compai- isoa___prewarb 90.2 88.2 89.7 postwarb 64.9 62.3 61.1 60.5 68.8 61.1 65.6 Short-Term interest rate: phase 55.4 70.1 comparison_prewarb 71.1 79.8 70.4 72.3 84.1 67.8 postwarb 73.5 53.6 52.3 67.5 64.9 66.9 Sign correlation-prewar 0.173 0.110 0.115 0.108 postwar 0.089 0.074 0.113 0.108 0.002 0.149 Correlation coefficient- 0.274 0.222 prewar +0.73 +0.07+0.82 +0.45 +0.40 postwar +0.84 +0.38 +0.34+0.13 +0.93 +0.77 +0.34 Three European countries Reference cycle phase comparison-prewar' Four Countries 88.1 535 postwar' 45.2 Short-Term interest rate: phase comparison_prewar' 35.7 60.6 48.6 postwar' 39.7 34.4 'Commercial paper rate. Per cent in agreement.

yield good covariatfonpercentages. Or, the specific cyclesmay be similar, but one series alonemay have a trend. The matter of trends is ofsome importance here, especially for those pairs where the Americanrate enters. Table 18 shows that the three prewar correlationcoefficients involving the United States are substantially lower than those for the threeEuropean pairs. This may be attributedto the fact that there is a downward trend in the Americanseries and nothing like it in the European data. This then may explain whywe have the lowest correlation coefficient for the pair United States-France(less than one-half of the highest coefficient) buta phase correspondence (for both phases) of 67.8per cent, which compares with the highest percentage of 79.6 and is higher than the lowest of64.1 per cent.25 For the sake of conveniencesome principal measurements made thus far, includingsome of Chapter II, are summarized in Tables 20 and 21. They requireno further comment since they give a corn- The French rate showed the least regularityin its downward movement, experiencing many upward movements while the UnitedStates rate was drop- ping, thus making fora low coefficient of conformity. 109 SHORT-TERM INTEREST RATES

TABLE 21 Summary of Rank Ordering, of Reference Cycle,and of Short-Term Interest Rate Corrclatjons

ThPJOHATION (_.) REL&T!ON, 1879..1914 1879-1914 1925-1938 l925-ij Ref. 8R0RTTERM Ref. SHORT-TEIthI INTEIT RATE ccle E.Esr RATE c11C18 INTERL%T MTh p phase p ose p pile_se corn-corn- corn- Corn- COIH- corn. pat. pan-signcorr, pan- pan-signcorn. pan- pan.. son son corn.coeff. son son corn.coeff. son SC)?) co,vcci4 London- Berlin 1 3 1 1 5 1 4 2 + London-. Paris 8 1 3 2 2 5 5 4.5 Berlin-. Paris 2 4 2 3 4 6 6 6 New York- London 4 2 4 4 3 2 3 1 + New York- - ++ Berlin 5 6 5 5 6 4 1 3 -+ ++ New York- Pans 6 5 6 6 1 3 245 + Not significant (0.99confidence regions oerlap).

prehensive picture of thepairwise behavior of the four cial centers with which great finan- this work isconcerned. It will gradually become clear how dicult it is to establisha ranking of the degree of interaction betweenmoney markets. The final measurement is the determinationof the dispersion andconcentration of specific cycle 23). The method turning points (Tables 22 and used was explainedin Chapter II, where it was applied pages 65ff., to reference cycles;our present measurements should becompared with those of ing the period July Table 7. The calculationscover- i876-September 1913are for the four countries, twice using thetwo American rates, and for the threeEuropean countries. The periodis substantially the same as that of the other SI 110 INTERNATIONALCOVAR1ATION computations. Such differencesas there are arise from turning points are not the fact that necessarily foundat the very first months for whichwe have data.20 or very last In interpreting Tables22 and 23, the discussion in Chapter reader is rernjnd of the II, pages 53 if.It is importantto guard against excessiveinterpretation, especially of deriving international regarding the possibility turning points ofshort-term interestrates. On the other hand thetables reveal a considerableamount of in- formation, especially ifviewed together cycles. We can also with those for thereference reconsider somequestions that were raised Chapter II, merelyon the basis of general in expectations, about the greater or smallerconcentration and dispersion troughs. It will furthermore of peaks and be desirableto compare the incidence of the average peaksand troughs with shall consider only the that observed before.21We prewar conditions, therebeing too few cycles after World War1. The comparison for all fourcountries is made twice, because the substitution of the of American commercialpaper rate for the call money rate in thesecond case. In the first casewe have an unbroken and troughs from sequence of average peaks October 1882 to June1913. The first peakin October 1882 shows thehighest dispersion ence cycles of Table 7) with (even including refer- 8.50 months. Thefollowing average trough of February_March1886 has a the largest of that mean deviation of 5 months, group and the second largestmean deviation of both peaks andtroughs (for the The smallest commercial paper set of data). mean deviation for allextrema is for August 1893 with 0.38 months. This is a date andeven a year that does not as a reference cycle date in occur Table 7, but thereis no doubt that in 1893 therewas a very sharp international shows up well here. financial crisis which The same observationmust be made for other dates and this leads us back to the argument,tentatively expounded in pages 96ff., of the possibility of someautonomy of both financial and international peaks and troughs whichare not revealed by the reference cycledates. The year 1893 is certainly oftenmentioned by historians ofbusiness cycles andnot only in relation to the solitary Americanpeak of January of thatyear.

Note in particular the slight difference In regardto Table 7. We hope the reader will concur withus that it Is not material. This will supplementour study of the covariation of the specific cycles with their interest rate respective reference cycles, carriedout above 111 SHORT-TERM INTEREST RATES

TABLE 22 Short-Term Interest Rates. Concentration and Dispersion of Specific Cycle Turning Points, July l878Septernbcr 1913, 422 months

COINCIDING I'EAKS' Average Mean tnea,j Averagedeviationdeviation Percentage peakb (months)(months_ciding Four countriesinclud- Oct. 1882 8.50 ing United States com- Sept. 1890 4.50 mercial paper rate Aug. 1893 0.38 Feb. 1900 1.25 2.79' 68 July 1903 1.00 Sept.-Oct. 1907 2.00 June 1913 1.87 Four countriesinclud- Aug. 1890 3.50 ing United States call July-Aug. 1893 0.75 money rate Dec. 1899-Jan. 1900 1.50 2.36 Apr. 1903 3.63 59 Sept. 1907 1.50 Apr. 1913 3.25 COINCIDING TROUCnS' Average Mean mean Averagedcviatlo,i deviation Percentage trough'a (months) (months) coinciding Four countrjesjnclud. Feb.-Mar. 1886 5.00 ing United States corn- June-July 1892 1.25 mercial paper rate Jan. 1895 3.63 Oct. 1901 4.3.8 3.50 62 Dec. 1904 2.75 Feb. 1909 4.00 Four countries.includ June 1892 1.38 ing United States call Jan. 1895 3.38 money rate Sept. 1901 5.88 3.20 51 Oct. 1904 2.88 Dec. 1908 2.50 Average mean deviation ofcoinciding peaks as per cent of corresponding average for troughs Four countries__including UnitedStates commercialpaper rate Four countries__including United 80% States call money rate 73% 'Coinciding turning points: turning points. no opposite turning point withinthe range of b The arithmetical mean of thecoinciding turning points average falls exactly between two is computed. if the months, bothare reported. TIthe average falls more closely Into one month, onlythis month is reported. 'Exdudlng 1886, average mean deviation= 3.20. 112 INTERNATIONAL COVARIATION

TABLE 23 European Short-Term InterestRates, Concentration and Dispersion of Specific CycleTurning Points, July I878-September 1913, 422months

COINCIDING PEAXS Average Mean mean Ave raredeviation deviation Percentage peak (months) (months) coinciding Three European countries July 1882 7.78 Aug. 1890 4.67 Aug. 1893 0 Jan.1900 1.11 2.67 72 July 1903 1.11 Sept. 1907 1.56 June 1913 2.44 COINCIDING TROUCRSS Average Mean mean Averagedeviation deviation Percentage troughb(months) (months) coinciding Three European countries Aug. 1879 1.78 Apr. 1886 4.22 July 1892 1.56 Feb. 1895 4.00 3.21e 78 Jan.1902 4.67 Nov. 1904 3.11 Dcc. 1908 3.11 Average mean deviation of coinciding peaksas per cent of corresponding average for troughs Three European countries 83% tCoinciding turning points: no opposite turmng pointwithin the range of turning points. The aritlunelica] mean of the coinciding turningpoints is computed. If the average falls exactly between two months, bothare reported; if it falls more closely into one month, only that monthis reported. Excluding 1879, the averagemean deviation = 3.44.

The mean deviations of otheraverage peaks are likewise small, July 1903 (1 month) and February1900 (1.25 months). These values are notably less than those for reference cycles,and from 1895 on eachone is smaller than the deviation of the corresponding trough. Thuswe have for the entire series I) = 2.79 months and D1= 3.50 months. This bears out an expectation raised on page 68, when discussing reference cycles, whichwas then not 113 SHORT-TERM iNTEREST RATES

confirmed, i.e., that there would be greater dispersion aroundthe troughs than around the peaks. The opposite was true of reference cycles, but the proposition holds for short-term interest rates. This is of considerable interest because it is plausible toassume that interest rates of different countries react more sharplyon each other when there is financial tension, i.e., when ratesare high, than in times of low activity. The lower turning pointsmay therefore depend much more on domestic business conditions(and domestic economic policies, if any) than theupper turning points. If we exclude the somewhat exceptional values for1882 and 1886, we obtaini= 183 and 15= 3.20, a more marked ratio than before. There is at least one drawback to this besides theobvious one of smallness of numbers: thereare the low values of p/P = 68per cent and t/T = 62 per cent. The latter is identicalwith the respective figures for reference cycles, the formerreference cycle ratio was 71 per cent. No correctionor addition of any sort is pos. sible of course, since this is merelya question of the wealth ard poverty of data.2 Yet we can makea test, by looking at the second set of figures, computed for the callmoney rate. This gives one average peak andtrough less. There isan unmistakable tendency for theaverage peaks to occur earlier than when the commericalpaper rate was used. The averagemean deviation for peaks is now only 2.36 months, for troughs 3.20 months,which is identical with the previousfigure after elimination of1886; otherwise it is 10 per cent smaller. From this and the otherdata of Table 22 (notably the lowvalue p/P= 59 per cent) it cannot be decisively argued that thesubstitution of the callmoney rate im- proves the picture very much. Asmentioned elsewherewe would not have made this substitutionhad it not been for the that the call contention money rate wasbefore l914moreimportant for the regulation of flows of short-term funds to and fromthe United States. If this were true we would expect to finda higher cyclical correspondence for it-.--providedthat short-termrates between countries where the proper rates have been chosen doshow high cyclical correspondence.The one neednot entail the other. In the prewar period the peaks and troughs of thefour short-term interest rate cycles always followed those of thefour reference cycles. This lag, however, was less marked for the threeEuropean countries, particularly in thetroughs. DThe absolute number of turning points is largerin the case of interest rates. 114 INTERNATiONAL COVARMTION We have, however, onemore argument in favor of the assump- tion of a genuine interaction betweenour interest rates, made on page 52. There it was stated that, if there isinvariance of a high degree of covariation betweentwo or more time series against an alternation of short and long cyclesin any or several of the many combinatorial complications, then thereis every reason to exclude the possibility that this covariationwas not due to interaction. In- spection of Charts 4 and 5 will show conditionshere. We do not establish a particular measurement of the degreeof this occurrence. We have it in the measure of coinciding peaks andtroughs. If it is high, and if at the same time short and longcycles alternate as the dates of turning points of the specific cycles(Table 12) disclose, then the conditions for makingour claim are given. Such alternation of short and long specificcycles exists: those from 1892 to 1897 were shorter than those following,up to 1908, when at least one more short cycle occurred.As usual the length of the American cycles is much steadier, i.e., theyare generally shorter than the European. So the alternation of shortand long specific cycles is again chiefly a European affair.Hence the transmission hypothesis looks more acceptable for these three thanfor all countries or for any pair which includes the Americanrates of interest. As to the satisfactory expression of the phenomenonjust mentioned by our measures,a qualification is needed: if we have high concentration of turning points with low mean deviations, there might still be exceptions which could fall preciselyinthose regions, where a change over from short to long cycles (Orvice versa) occurred. The ratios p/P and t/Tassume therefore added significance. The higher they are the better isour argument sup- ported.29 Tables 22 and 23 have shown fairly good values for these ratios and better ones still for the European countries. However we still need further evidence for interaction. The alternation of short and long cycles forour interest rates should be compared with that of the reference cycles, if itoccurs in both instances, it need not be in thesame order. Not even the exhaustion of all peaks and troughs by theaverage peaks and troughs is a guarantee for that. So we can makesure only by actually comparing the years in which the various turning points fall. Much has been said about this implicitly alreadyon p. 96 when comparing our interest rates with their respective reference cycles. There was complete exhaustion of all available peaks and troughs for the three European reference cycles. 115 ShORT-TERM INTEREST RATES We conclude this chapter by making five separate remarks which either take up points already raised or point to subsequent investigations which will subject our data to different approaches and place them in close relationship with data of other activities in the international financial sphere. First. At the beginning of this chapter we raised the important question, for the first tinie,3° whether the absolute level of short- term interest rates, its changes, or its cyclical behavior were more strongly subjected to domestic or to foreigii influences. It was then stated, on general credibility grounds, that the domestic factors might be the more important. In this most economists would probably concur, although perusal of the literature does not disclose anything decisive. There is one important exception inthe literature_though it does not cover the entirerange of this prob- lemthat is, the study of the interdependence of short-terminterest rates with exchange rates. This subject willOCCL1 us more fully than in any previous discussion (cf. ChaptersV and VIII). Second. There is the larger question of institutionalhomogeneity throughout the period covered, that is, thequestion whether the behavior of the data shows significantvariation revealing such changes in the institutional backgroundas were not already taken into account in the selection of theparticular interest rates, and discounted in our continuoususe of them. Such a break actually occurs and conforms with that observed for thereference cycles. It happens after WorldWar I. The break exists, however,whether the observations are restricted to the short period of thecontinued application of the principles of thegold standard orare extended beyond that period. Thecausation is naturally complicated, the war as such having little to do withit directly. In fact itwas more the emergence ofstate intervention with interestrates and all those international factors in the fieldof capitalmovements and foreign trade which produced less synchronizedcyclical behavior. This craving for greater autonomy on the part ofmany states has become a permanent feature of the years since World War I and villhave to be referred torepeatedly. For that reason our study must place chiefemphasis upon the period up to 1914. Theimplications of this the validity of the institutional change for various theories of thebusn55 cycle and of Cf., however,page 24 if. where the probleni foreign influenceswas put in general of separatiiig hnnttic and terms and wherecertain ticep-lying methodological difficultieswere pointed out. 116 INTERNATIONAL COVARIATION international trade are clear. Further reference to this is made in the fifth remark. Third. There is the question of the absolute levels of the interest rates. Tables 9 and 10 disclosed considerable differences in their absolute levels and in their variability. The problem is whether these differences__chiefly of the absolute levelshave any close bear- ing upon their covariation. Again this question cannot yet be an- swered. There is however an interesting phenomenon to note: the American call money rate is more nearly at a level corresponding with that of the other (foreign) rates. Yet its agreement with the three European rates over the cycle is decidedly less than that shown by the commercial paper rate, as far as covariation is concerned. But this complex of questions awaits study of the ampli- tudinal behavior and in every way transcends, at the level at which it has to be carried out, the scope of this investigation. Some aspects of the question whether the principles of interdependence of interest rates, both short-term and long-term, require that they vary only within certain rigid limits are dealt with extensively below, especially in Chapters V and IX. Fourth. Closely connected with these points is the set of problems arising from the study of the differences between short-term interest fates. Realizing that the rates will not in general be identical with each other, it might be possible to assign to their differences certain maximum values. If there is to be interaction between the rates, it would have to take place through shifts within these maxi- mal ranges. All this would further be tied up with induced and au- tonomous movements of other factors, such as capital transfers, exchange rates, etc. These latter in their turn depend in the range and frequency of their fluctuations upon limits imposed by the rules of the gold standard to the extent to which they are or are not adhered to. All this will form the substance of a large part of the subsequent investigations, which will throw some light upon the problems of covariation and transmission. Fifth. On page 21 if. we have touched upon the question whether the cyclical behavior of economic activities in various countries is sufficiently similar to warrant the establishment of a uniform theory of economic fluctuations applicable to any or all countries. Whether the answer is yes or no in general will not depend entirely on one field alone. As far as interest theory is concerned, we should like to question whether such a theory exists in the shape compatible with the facts described above. There is a large field of verbal, even 117 SHORT-TERM iNTEREST RATES quasi-metaphysical, discussion above the "nature" of interest,etc., and there is a very much smaller numberof investigations the behavior of the principalrates over long periods showing ing Countries. in the lead.

It may be helpful at thispoint to summarize the specific findings of the chapter. more impoItant As a general rule, the international peaks andtroughs indicatcJ by the referencecycles are present in the short-term interestrate cycles, although foreach of the fourcountries there tended more cycles in the interest to be rate series. Unlike thereference cycles however, the short-terminterest rate peaks tend to show lessdis- persion than the troughs.The four countries do number of cycles, the not have thesame United States havingthe greatest Before World War number. I the British rate showedthe best phase respondence with its reference cor- cycle whereas afterthe war it the poorest, theUnited States was commercial paper ratenow showing the bestcorrespondence The British phase rate also showedthe best correspondence with therates in the other three before the countries war although not in theinterwar period. War I the UnitedStates rate showed Before World the leastagreement with the two continentalcenters, much less than centers with each other. did the threeEuropean Expansions were longerthan before 1914whereas they contractiom were shorter in theintenvar period. Using frequencydistributio5 to the rates, the gauge the relativestability of variation for theUnited States call much greater thanthat of the money rate was before World commercial paper rate,particularly War 1. TheBritish rate European rates, was the most unstableof the and thisinstability, i.e., the interwar variability, increasedin period, evidenceof the economy was laboring. tension under whichthat The stabilityof the French 1914 was reflectedin its short-term economy before less than that interest rate showinga variance of GreatBritain and variations decreased Germany. Ingeneral seasonal after WorldWar I. Before 1914 the United Statesrate was the highest whereas in theinterwar period of the four it was thelowest. Appendix. Description of theMonthib, Data Un ited States. Here we choosethe Sources: Seconda,, comnercjal paper rate.3' This Financji4 Review source, Macaulay,op. cit., Table 10; and Conime,jand Fhmnckl priniary sources, that these twosources are practically Chronicle Macaulay identical. states 118 DESCRIPTION OF THE MONTHLY DATA was one of the most important Americanshort-term rates, at least up to 1914, at which a large volumeof domestic and foreign business was transacted. The term applies to promissory notes"on which merchants and manufacturers borrow money for use in the ordinary course of their business. No stock or bond collateral is deposited to secure the loan?'32 After 1914 the notes were admitted to rediscount at the Federal Reserve banks under certaincondi- tions. Up to 1923 the rate applies to endorsed commercial paper ("prime double-name paper"); in the late 1920's the endorsement dropped out of use. The maturity of the paper was, up to December 1923, sixty to ninety days and thereafter the rate relates to paperhaving a maturity of four to five months. The tendency was toward paperwith longer maturities.84 The commercial paper rate is what its name implies. It hasde- creased in importance in the course of time, but more so in the post-World-War I period rather than before 1914, toward which our main interest is directed. Itwould be most important for us to know what the share of international transactions outof the total was. Unfortunately no data atall exist about this. We return to this matter below (cf. Chapter V).M Besides the commercial paper rate, we shall also consider to some extent a second American short-term rate,theNew York call mone, rate.The reasons lie In the disagreement as to whether this rate or the commercial paper rate was of greaterinternational significance for the shifting of funds, chiefly between NewYork and London,

NMacaulay, op. cit., p. A-344. NThis point may seem of small significance in thecharacterization of our series, but it will be seen below, Chapter V. that it assumesgreat importance. In order to have the characterization in one place, we mentionthis fact here, though it is not immediately significant. "Statistical procedure: the monthly rate is in generalthe average of the weekly rates. If a Saturday occurred on the first orsecond of the month, the week ending on that Saturday was considered asentirely in the preceding month. If a Saturday occurred on the fourth, fifth, orsixth, the week was considered as entirely In the later month. If It occurred onthe third, the week's average was considered a quotationfor half a week in the earlier month and half a week in the later month. In times of major disturbances the averagemonthly rate is an average of daily rates rather than of weekly averages. Cf.Macaulay, op. cit., p. A-350. As for the literature, we mention besidesMacaulay and the standard descriptions of the New York money marketsuch as M. C. Meyers, B. H. Beck- hart, J. C. Smith, and W. A. Brown Jun.TheNew York Money Market, 4 vols., New York, 1931-1932, the work by Albert 0.Creef, "The Commercial Paper House In the United States," Harvard EconomicStudies, Vol. i.x, Cambridge, 1938, a thorough study with an almostexhaustive bibliography. 119 SHORT-TERM INTEREST RATES and even in theprewar period. The call money rate36 is muchmore closely related to thestock market, whichsets it apart front rates chosen for the other the countries where thereis ICSS doubt their international, as commercial significance. '[herate covers to stock or bond brokers, "loans or to investment bankers whoplace in the hand of the lendersstocks or bonds,or both, as collateral. relates to a mixed The rate collateral which includesindustrial government securities."3' as svell as Those loans are made for indefinite periods,38to be repaid call." The rate isan official renewal rate made by at the moneycom- mission of the New YorkStock Exchange Clearing ordinary times Corporation. In some 95 per cent ormore of the call loans, and renewed,carry the renewal rate.3° The new the sameas for the commercial statistical procedureis paper rate (ef. footnote34). The callmoney rate was always a very importantrate on the New Yorkmoney market. The much on the New York Stock greater significance ofevents Exchange ascompared with any countiy accounts forthis. Also it European cannot be doubted thatmany movements of funds fromEurope across the by these ocean were oriented events. But this isnot the sameas being attracted pelled by thechanges and relative or re- the late l920's positions of the callrate. In the influx ofEuropean funds vated by expected to New Yorkwas moti- capital gainson the stock exchange by a differentialof European rather than there rates and the callrate. Before 1914 were occasionally stilldifferent institutions, accounfs.40 notably the joint Great Britain. Herewe chose the London rate.4' This rate is open market discot charged for thediscount of bank of exchangeissued or accepted bills, i.e.,a bill bank bill5 by a bank. Thediscount rates for are less than therates charged for R fine trade bills.The Sources: secondary source, MacaWay,op. cit., Table 10 primary sources,1878-1889 Commercj, and Appendix E.; Financial Rccjew andComml andand FinancEdClironkle. 1890-1919 eral Reserve Bulletj. Financial Chronicle;1923-1938, Fed- ' Macaulay,op. cit., p. A-337. TMCf. footnote 33. Maeaulay, op. cit.,p. A-339 These were accounts owned jointly by houses belongingto one and the the Londonand New York banking accounts were very important. same financialinterests. Moveii No material is of these of transactjo interest charges, available concerningthe volume to the great signillcanc etc., except that of theoperations. No participating firms attested could safely neglectthis institution statistics of capitalmovements Sources: primarysource, Econoj,j 120 DESCRIPTION OF THE MONTHLY DATA discounting is done by special discount houses, most of very old e standing. The olficial rediscount rate of the Bank of England, 0 briefly called "bank rate," applies to the same paper. A very large S volume of business was transacted at this open market rate and the e share of international transactions was always very large and con- e tinued to increase in the period before 1914 in conformity with London's dominant position in international finance. In the postwar period the rate was still of great importance, although Lon- don's position was no longer of this dominant nature. The maturity of the paper varied somewhat. The Economist indicates usually that its data relate to three months hank bills; but in other instances the heading reads: two to three months bank bills. This uncertainty about the actual or most frequent maturity does not deprive the data of their value; but it introduces, as was noted before (cf. footnote 33, a factor which will assume some importance later.42 France. Here we chose the Paris open market discount rate ("Escompte hors Banque")This rate is charged for the discount of bank bills, i.e., a bill of exchange accepted by the bank. it is very similar to the British open niarket discount rate. But there were no special discount houses in Paris; this was and is a peculiarity of the London market.44 Direct accommodation of business by banks also played a large role.

'The literature about the London money market is enormous. We indicate here some important sources, where not only descriptions of this money rate may be found but also the general setup of the market is analyzed. The order in which the books are listed indicates their more direct relation to the descrip- tive side, while from the analytica! point of view it might be quite different. George Clare, A Money Market Primer, London, 1891, 2nd ed., 1923. Ellis Thomas Powell, The Evolution of the Money Market (1385-1915), London. 1915. F. Lavington, The English Capital Market, London, 1921. E. C. Peake, An Academic Study of Some Money Market and Other Statistics, London, 1923, 2nd ed., 1926. W. T. C. King. Illstoi-y of the London Discount Market, London, 1936. Richard Sidney Sayers, Bank of England OperatIons, 1890-1914, London, 1988. a Sources: 1876-1888, Economist; 1889-1908, Statistics for Great Britain, Germany and France, U.S. National Monetaly Commission, Vol. 21, 1910, and Economist; Annual Reports, 1925-1938, Federal Reserve Board. "The literature about the Paris money market is large but not nearly as in- formative as that for other centers. That is a general experience relating to French monetary and other economic statistics. They are either completely lacking when they are available in other countries, or often of very doubtful quality without any possibility of checking and improving them. On the other hand there sometimes exist statistics in fields where, in other countries, it is 121 SHORT-TERM INTEREST RATES Gernany.Here we chose the BerlinprivatedLscovnt rate ("Fri. vatdlskontsatz").4e This rate is charged for loans, i.e., discount,on prime bills) that Is, on those of "recognized solidbanks and bankers. They must be payable ata bank place, i.e., In Berlin or some other town with a branch of the Eeichsbank. Theymust call for at least 5,000 Marlcs.'T This characterization hasremained essentially true since 1910 but as In other countries the role ofthe bills of exchange has decreased in Germany.This was associated with the rise of the large enterprises in industryand banking and the more direct aceommodation of business, especially big business, bythe banks. The private discountrate is comparable to the French and British rates to a greater exter:in that order.tlianto the commercial paper rate at New York. But this holds respectivelytrue for the three European rates. Thevolume of transactions carriedout in Cermany with bills of exchangewas considerable and a fairly high percentage of the bills either referredto international transactions arising from international tradeor became Instruments of capital movements. A frequently used practice of Germanbanks was to rediscount these bills in Paris whenthe rate there was lower and to repatriate them just before they felldue. This was the so-called 'pensioning" of bills. The duration of the loansvaried from fifty-six days to three months. This variation is similarto that found above for the other

impossible to obtain any reliable fosn3atlou at alL Seen against th.general background, this makes themvery Fortunatdy Interest rates (and later exchange rates)are generally lecurate but lee C3iepter V, are SOn)e important works: section 7. Rem Alfred Neymarck Di tOrgsnisstlamSi. Mardsó Fbwnclera téfranger, ParIs,1884. en Franca at a k Aupetft and co-authors, L Grand. Marchd, F'enci.rs,Franc. (P.1* at Province), Londres, Berlin, Ne.pYork, Pu 1912. Pierre Castes, Lea Gianda Marckd,Financier., Paris, 1922. H. Fischer, Dir Stat du frsorinirthp, Kradi.* Statistical procedure: for 1984. 1876-1868 and 1909-1914 theaverage monthly rates were unaputed by theNs4eel Bureau cI Ecomomic monthly average Is the Reseerththe a of as mo werAdy_quotmiow as them Fridays of the mimtk Fcc1 1906 the em the ?4atlonsl Bureau: for eri were utuisd by J 188G4aluuy 1905s wauber of Thur.. days In the mouth deiembxsd eamohor of March 1906-December19* qaoladora - for the wa1Iy averages was the wanber of £otriatlayi. For1 1968 were snmad the Nadonsle Seüass. TMSourcei: 18164910, Dli Rek-' 1875-19 0, Stogl.tlwljea Jahibsi4s Ill, d 191.1; 1911-1083, D.ut.th. R.k* 19844138,WWac.b4t s.d U.S. National Mcaatary . cli., VoL *1, p 901. DSCftlpfloN OF TEEMONTHLY DATA We have not been able to ascertain what theaverage was, certainly not theaverage weighted for the volume of transactio,u.*4$

'Satiwicl procedure; The monthlyaverage rite is computed on the basis d average daily quotations him the various sources. Th, latter Is theaverage of quotations fat iong-term and sat-termpaper, Le., within the above limits. If mote than one quotation occurred1 either for short-term or long-termpaper, then the higheot quotation wasuae The bteretvre about Carmenmonetary organization sInce 1879 Is ample and goumly accurste. We thaiane only a few works that rolate directly E.VtTh.r die HUt. dcr varach4.dm.i,Zfnaerten and IhT. wechsebeage iL5ism(. Jeer, 19(. Pburger Die Eurwkkiung des Zdoafws.sin Deutichiand von 1870- 19tS, FrnikMt, 1901 H. Albert, Die p.chlthdIcls. Eistwlckltmg des 1896 bIt l9$, I4prlg, 1910 ZMafmssses hi Deutschland von Paul WallIth, ledge zur Cnscliichte desZinsfusses von 1800 bls zur Cegenwart," J*rbuch flit Nattoedökonomk, Vol. 42,1911, pp. M9-312. 1. Poble1 Di. neonre Ewtscickivng des Zffiifwsei tied desElnfluss des Welt- krleges aeiciness Stand, Fra'ildurt, 1915. M. Palyl and P. QuIt*ner, Hmndw&terbuch des Bankwescis,BerlIn, 1933.