The Federal Reserve Amendments of 1917: The Beginning of a Seasonal Note Issue Policy

Article excerpt

SEVERAL STUDIES HAVE FOUND that the seasonal behavior of interest rates diminished after 1914, the year the Federal Reserve began operations (for example, Shiller 1980, Miron 1986, and Clark 1986). Mankiw, Miron, and Weil (1987), Miron (1988), and Barsky, Mankiw, Miron, and Weil (1988) suggest that the Federal Reserve was responsible for the change in interest rates, while Shiller (1980) and Clark (1986) suggest that the Federal Reserve had little or no effect. Because seasonal movements are believed to reflect real economic activity, the underlying issue in these studies is whether anticipated monetary policy can affect the real sector of the economy. By analyzing what the Federal Reserve actually did during the first few years of operation, I will show that monetary policy between 1914 and 1917 was not seasonal and, in effect, was fairly restrictive. Discount operations were limited relative to the resources of the Federal Reserve, and, when notes were issued, the district banks tended to deposit gold with agents such that new notes were substantially offset by gold deposits. It is argued that gold inflows caused by World War I, not the establishment of the Federal Reserve System, are responsible for smoothing interest rates in the United States between 1914 and 1917. [1] After the passage of the 1917 amendments to the Federal Reserve Act, however, money growth became seasonal and it is possible that Federal Reserve policy influenced the seasonal movement of interest rates.

The 1917 amendments, signed into law on June 21, 1917, by President Woodrow Wilson, greatly increased the note issue power of the Federal Reserve. Under the Federal Reserve Act of 1913, Federal Serve agents, who represented the Federal Reserve Board, were required to retain eligible commercial paper as collateral equal to 100 percent of the value of notes issued and Federal Reserve district banks were required to maintain an additional 40 percent gold reserve against notes in circulation. In effect, a one dollar Federal Reserve note was backed by a minimum of forty cents of gold and one dollar to eligible commercial paper. In practice, the twelve district banks repurchased commercial paper from their Federal Reserve agents using gold, leaving the agents with gold instead of commercial paper as backing for notes issued. The gold deposited with agents was a constraint on the issuance of new notes because it could only be released to the district banks if a corresponding amount of notes were withdrawn from circulation. The 1917 amendments relaxed the new note issue constraint by allowing district banks to count gold held by agents as part of their 40 percent gold reserve requirement. This simple accounting change had an immediate effect on the gold backing of notes and the amount of notes issued. The gold backing of notes fell from 100.4 percent on June 1, 1917, to 78.2 percent on June 22, 1917. Notes in circulation rose from $464.8 million on June 1, 1917, to $1,246.5 million on December 28, 1917, while they had increased by only $244.9 million for the seven-month period preceding the passage of the 1917 amendments.

The 1917 amendments were passed to help finance U.S. participation in World War I. When the United States entered the war, on April 6, 1917, it was estimated that the first year's expense would be $4 billion, with $2 billion to be raised from bond sales. If the general population failed to support the war effort, the financial burden would fall on the Federal Reserve. The 1917 amendments, along with amendments passed in 1916, permitted the district banks to issue notes that were backed, although indirectly, by government bonds. I suggest that without the 1917 amendments the resources of the twelve Federal Reserve district banks could not have supported a $2 billion bond issue.

The paper first examines the operating behavior of the Federal Reserve between 1914 and 1917. During this period, there were many institutional constraints that had to be relaxed for the Federal Reserve to operate effectively. The analysis then turns to the amendments of 1917, how they increased the note issue power, and how the United States financed its participation in World War I. Following an examination of trade and gold flows during the war, the paper offers some empirical evidence that the 1917 amendments affected the seasonal behavior of notes issued and interest rates.

1. PROBLEMS WITH THE EARLY FEDERAL RESERVE SYSTEM

The Federal Reserve Act of 1913 created the Federal Reserve System, which began operations under what may have been the greatest financial crisis ever to face the United States and Europe. In the days immediately preceding the outbreak of World War I on July 31, 1914, disruption of remittances from European clients of English accepting houses to cover maturing bills led to a halt in discounting of foreign bills by London houses. That action interrupted the supply of sterling exchange to New York banks, which then used gold to pay foreign remittances, drawing down their reserves. The volume of foreign remittances soared as European investors sold stock to increase liquidity and avoid losses due to falling asset prices. The resulting financial imblance closed stock exchanges in Europe and the United States for several months. In response, the Secretary of the Treasury, William McAdoo, acting under the Aldrich-Vreeland Act of 1908 and the Senate Emergency Currency Bill of August 4, 1914, authorized the issue of $500 million in emergency currency. National banks found the emergency currency appealing because it met their demand for liquidity, fostered in part by public hoarding, and was a cheap source offunds. Initially, the cost of securing the emergency currency was only 3 percent, whereas the New York clearing house charged 6 percent for clearinghouse certificates. In only a few weeks the Treasury had "sold" over $382 million in emergency currency, which represented about 10.2 percent of the existing money stock (Annual Report, Comptroller of the Currency, 1915, p. 90).

The Federal Reserve was organized for the express purpose of providing currency relief in an emergency. When the operations began on November 16, 1914, however, the Federal Reserve offered only token new liquidity to the economy. Instead of providing inexpensive currency, the various district banks charged between 5.5 percent and 6 percent to discount commercial paper, while market rates for such paper were .5 percent to 1 percent less. (2) As a result, the discount operations of the Federal Reserve were limited to only $10.6 million in 1914, an amount that represented only about 4 percent of the resources of the Federal Reserve (Annual Report, Comptroller of the Currency, 1915, p. 137). The Federal Reserve increased its discounting activities to about 10 percent of assets during 1915, but the largest increases did not come until the second half of 1917 when the Federal Reserve assisted with war-related financing.

The reasons why the Federal Reserve choose not to offer much liquidity during the early stages of World War I are not entirely clear. Generally speaking, though, the Federal reserve did not "hit the ground running" because it lacked adequate operating rules. (3) As the Federal Reserve did evolve and more rules were established it became clear that several clauses of the 1913 Act created problems. The biggest bottlenecks of the Act were the clauses restricting the issue of Federal Reserve notes. (4) Sections 13 and 16 of the 1913 Act state quite clearly that the intent of Congress was to issue notes only as advances against maturing commercial transactions. District banks would present eligible paper to their agent in exchange for notes. Federal Reserve agents were required to hold the eligible paper as 100 percent collateral or notes issued in an account that was separate from the account of the district bank with whom they were affiliated. The notes were considered a liability of the district banks and they were required to set aside a 40 percent gold reserve against notes in actual circulation. The district banks were allowed to reduce their liability for notes by depositing gold or other "lawful money" with their Federal Reserve agents, but these deposits did not reduce the 40 percent gold reserve requirement if the notes were still in circulation.

With these restrictions, the 1913 Act had adopted the Real Bills doctrine. The Federal Reserve, however, found a way to increase notes issued even if the current volume of commercial transactions did not warrant it. Influenced by Governor Benjamin Strong of the Federal Reserve Bank of New York, the district banks developed a scheme, known as "reversing the pump," to increase the quantity of notes issued. For example, at the end of November 1914, before the scheme was adopted, notes issued totaled $2.7 million, yet discounts amounted to $7.4 million, indicating that the Federal Reserve had used less than half of the eligible paper it had acquired to back note issue. By June 1915, however, notes issued totaled $82.9 million, more than double the amount in discounts and bills purchased (Annual Report, Comptroller of the Currency, 1915, I, p. 137). By the end of 1916, notes issued exceeded discounts and bills by over $140 million (Federal Reserve Bulletin, 1917, pp. 145-47).

The "reversing the pump" scheme is described by Willis (1923, p. 859), who was Secretary of the Federal Reserve Board from 1914 to 1918:

The operation of the plan was as follows: A customer bank having discounted some commercial paper with a reserve bank to the amount of, say, $10,000, the reserve bank was then in position to issue $10,000 in notes in exchange for this paper collateral. The collateral would be turned over to the agent, who would thereupon issue the notes after going through the necessary formalities for the purpose of getting them from the board. As things stood, the reserve bank had parted with $10,000 of collateral which was in the hands of the agent, while the board had "issued" $10,000 to the agent who had transferred it to the bank. The bank set up a reserve of 40 per cent against the notes. Then, in the event that a new customer came to the reserve bank and asked for Federal Reserve notes, there would be no means by which the reserve bank was able to get the notes from the agent, since presumably its collateral was entirely in the hands of the agent. It, however, now resorted to the plan of redeeming its collateral by placing gold to an equal amount with the agent. The effect of this, then, was to release the reserve previously set up against the notes. The agent now had $10,000 in gold, while the bank had taken back its collateral. The collateral would in this way be presented over and over again, being returned to the bank after each operation just as if it were a new offer of commercial paper. The result was that the bank was able to "pump" out as many notes as it wished, so long as it had a single dollar of commercial paper of eligible variety; provided, however, that the agent went through the more or less elaborate routine that was necessary in getting the notes.

Large gold inflows, generated by war-related trade, found their way into the hands of the district banks and subsequently the Federal Reserve agents under this scheme. It is not clear, however, why the Federal Reserve sought to accumulate a large stock of gold. Governor Benjamin Strong suggested that the gold might be needed in order to satisfy international debt obligations in an emergency. (5) Willis (1936, p. 253) suggests that the scheme was adopted to meet the demand for "clean paper"; that is, member banks wanted to convert old and worn currency into new notes. Both explanations may have merit, but a simpler explanation may be that the entry of the United states into the war was considered inevitable, even though President Wilson and others in the government professed neutrality. The Federal Reserve then was attempting to husband the country's gold to help finance the war in the same manner that the central banks in belligerent countries had done. (6) This scheme created a problem because, under the 1913 Act, gold could not be released by the agent without withdrawing notes from circulation; that is, there was no leverage gained by holding gold. A swap of gold for notes would not assist with the financing of the war effort, and would remove gold from the Federal Reserve. The 1917 amendments relaxed this requirement, and thereby provided a vehicle for financing the war.

2. THE AMENDMENTS OF 1917

The Federal Reserve had a two-tiered accounting system. Each district bank kept accounts with its members and with the Federal Reserve agent assigned to that bank by the Board. The 1913 Act gave the agent control, through the Board, over the notes issued by a district bank. Each note, as mentioned, was backed 100 percent by "eligible commercial paper and 40 percent by gold. The agents were assigned the job of evaluating the collateral, holding it on their own account, and maintaining a record of the 40 percent gold reserve. (7) Once a district bank had received notes from its agent, the collateral deposited was no longer part of its balance sheet, which meant that the district bank removed commercial paper from its balance sheet in exchange for notes.

The 1917 amendments changed Section 16 of the 1913 Act in two places. The first change allowed gold and gold certificates to act as collateral for notes issued, whereas previously only eligible bills and bankers' acceptances were allowed. The second change restated the rules on reserves: "Every Federal Reserve bank shall maintain reserves . . . Provided, however, that when the Federal Reserve agent holds gold or gold certificates as collateral for Federal Reserve notes issued to the bank such gold or gold certificates shall be counted as part of the gold reserve which such bank is required to maintain against its Federal Reserve notes in actual circulation." Thus, the 40 percent gold reserve that was originally required combined with only 60 percent commercial paper became the effective limitation on the issue of Federal Reserve notes.

These simple changes had a singificant effect on the note issue capability of the Federal Reserve. For example, on May 25, 1917, about a month before the amendments became law, the Federal Reserve held $334.3 million in gold coin and certificates and $2.9 million in the gold redemeption fund, while the agents held $456.6 million in gold as "collateral" against notes issued (Federal Reserve Bulletin, June 1, 1917, pp. 491-94). The amount of notes in circulation on this date was $454.4 million. The 40 percent gold reserve requirement meant that the district banks had to retain an additional $181.7 million in gold against notes in circulation. This reserve restriction and the agents' gold holdings effectively impounded $638.3 million. If the changes to Section 16 mentioned above had taken effect on May 25th, the note issue power of the district banks would have changed from $843 million [($334.3 + $2.9) times 2.5] to $1,984 million [($334.3 + $2.9 + $456.6) times 2.5], an increase of $1,141 million. (8) Similarly, the note issue power of the Federal Reserve in July 1917 was estimated to have increased by $1,875 million because of these amendments. (9)

The district banks moved immediately to expand the issue of notes. Table 1 shows the notes in circulation and gold backing from May to December 1917. Over $80 million of commercial paper was added to and about $70 million in gold was removed from the agents' accounts during the week that the amendments became law. This change was almost entirely due to the Federal Reserve Bank of New York,

[TABULAR DATA OMITTED]

although other district banks followed its lead as they came to recognize the new note issue power of the June 1917 amendments. By the end of the year, notes in circulation had increased by 153 percent and the gold cover had fallen to 62.7 percent.

The entry of the United States into the war was an important factor encouraging the adoption of the 1917 amendments. On April 24th, President Wilson signed into law a bill authorizing the issuance of $5 billion in long term bonds and $2 billion in one-year certificates of indebtedness (Commercial and Financial Chroncile, Bank & Quotation Section, May 5, 1917, p. 14). The first issue of war bonds, known as the First Liberty bonds, was for $2 billion and paid 3.5 percent interest free of taxes, maturing on June 15, 1947, with a call option on or after June 15, 1932 (Commercial and Financial Chronicle, Bank & Quotation Section, June 2, 1917, p. 14). At the time, this was the largest issue of debt ever undertaken by the United States. Because of the amount of money involved, President Wilson and the Treasury were concerned that the first bond issue succeed and that it not be too disruptive to financial markets. Loan committees were established by each district bank, over three hundred bond salesman were recruited, and numerous, well-known bankers were asked to help make the issue successful. Secretary McAdoo assured participating banks that until needed the funds collected would be deposited at banks according to the distribution of subscriptions. Even with these measures in place, there was concern about the willingness to the general public to support the war effort. As late as June 8, 1917, seven days before the books were scheduled to close on the issue, Treasury Secretary McAdoo announced that the subscription fund was $700 million short of the $2 billion goal (Commercial and Financial Chronicle, Bank & Quotation Section, July 7, 1917, p. 13). If the public was unwilling to buy these bonds, then the burden would fall more heavily on the Federal Reserve. As already discussed, the note "capacity" of the Federal Reserve at the end of May 1917 was only $843 million. Without a change in the 1913 Act and a willing population of investors the bond issue could easily fail.

When the books finally closed on the First Liberty bond issue, it was over-subscribed by $1 billion. In retrospect, it is not difficult to understand why the First Liberty bonds were oversubscribed. Most of the subscriptions for Liberty bonds had come through National banks. The Federal Reserve amendments of 1916 permitted the district banks to "advance" funds for fifteen days to member banks on their own promissory notes if these notes were backed by "eligible" paper or government securities (Annual Report, Federal Reserve Board, 1916, pp. 135-36). The 1917 Act made it possible to use these loans as "eligible" collateral against the issue of Federal Reserve notes. (10) The discount rate at the New York district bank was 3 percent at the time. The First Liberty bonds offered a .5 percent premium to the current discount rate. As it was possible to rollover the fifteen-day promissory notes with the district banks, the National banks received income of .5 percent for subscribing. (11) If the member banks subscribed to the entire Liberty bond issue by this method, for example, they stood to gain $410,958 tax-free every fifteen days. Before the 1917 Act, the note issue capacity of the Federal Reserve would support only a very limited amount of rollovers by member banks, but after its passage it was possible for the entire loan to be financed by rollovers.

Table 2 shows just how important the fifteen-day loans were to the success of the Liberty bond issues. (12) The amount of fifteen-day discounts and total discounts by the Federal Reserve System for each month in 1917 changed greatly after the Treasury began its loan operations and after the passage of the 1917 amendments. (13) The bulk of the installment payments on the First Liberty bond issue was due in June--August, when the fifteen-day discounts took their first big jump. Although the Second Liberty bond issue, also equal to $2 billion, required payments mostly in November and December 1917, many subscribers paid in full when the issue closed on October 27, 1917, an indication that they had borrowed those amounts from member banks. However, the banks relied less on rediscounting their customer

TABLE 2
FEDERAL RESERVE FIFTEEN-DAY DISCOUNTS AND TOTAL DISCOUNTS, 1917
                 15-Day         Total       15-Day/
  Month        Maturity       Discounts      Total
January      $10,804,495     $18,326,286     58.9%
February      14,646,032      22,408,604     65.4
March         18,694,658      26,706,266     70.0
April         34,967,987      50,055,801     69.9
May           69,958,620      91,413,473    76.5
June         691,247,199     750,270,739     92.1
July         403,310,779     460,733,353     87.5
August       164,538,077     220,838,942     74.5
September     488,347,299    548,164,104     89.1
October    2,601,438,919   2,681,165,854     97.0
November   3,014,997,799   3,206,486,771     94.0
December     616,232,779     892,237,774     69.1
  Source: Fedeal Reserve Bulletins, selected issues, 1917-18.

loans at district banks than on their own fifteen-day promissory notes secured by government war obligations, as the second jump in the latter in October and November shows. After the United States entered the war, Treasury certificates were issued almost monthly in amounts ranging from $200 million to $700 million, and help account for some of the increases shown on Table 2.

By releasing gold held by the agents for new note issue, the 1917 amendments played an important role in the financing of war loans. The resource cost of discounting to the Federal Reserve, however, was the same both before and after the amendments; the U.S. Treasury bore the cost of the "low" discount rate policy. (14) The amendments, however, made it possible for the limited resources of the Federal Reserve to be leveraged in order to support these large loan issues.

3. INTRA-WAR TRADE AND GOLD FLOWS

It is not surprising that trade with Germany, our second largest trading partner, fell sharply duringthe war, but these losses were more than offset by increased trade with the United Kingdom and other countries. Real exports, as measured using Romer's (1988) implicit price deflator, increased from $2,466 million in 1913 to an intrawar peak of $4,666 million in 1916 as measured in 1913 dollars (Historical Statistics of the United States (1975), series U317-U334, p. 903). Real imports also increased, but not as much. The favorable balance of trade during the war led to large gold inflows. In 1915, 1916, and 1917, the United States received on net $420 million, $530 million, and $180 million in gold, respectively, whereas for the twelve months before the war, the United States had lost $165.2 million in gold (Banking and Monetary Statistics, p. 536). Much of this gold found its way into the Federal Reserve banks through the "reversing the pump" scheme and as reserve

TABLE 3
NET GOLD FLOWS TO THE UNITED STATES, 1915-17
                         Net Gold
          Period        (million)
1914   First quarter      $2.8
       Second quarter     -56.1
       Third quarter      -64.5
       Fourth quarter     -47.5
1915   First quarter       42.6
       Second quarter      94.8
       Third quarter      115.6
       Fourth quarter     167.5
1916   First quarter      -3.9
       Second quarter     124.4
       Third quarter      167.9
       Fourth quarter     241.7
1917             First quarter      241.5
       Second quarter      34.2
  Source: Banking and Monetary Statistics, 1914-41, p. 536.

deposits from member banks. Member bank reserve deposits grew from $261 million at the end of 1914 to $750 million on March 5, 1917, mostly due to the reserve pay-in schedule adopted by the 1913 Act (Federal Reserve Bulletin, July 1, 1917, p. 484). The Treasury also received some of this gold as more duties were paid on exports. Most of the Treasury gold found its way into general circulation through the issue of gold coin and certificates. From November 1914 to May 1917, gold coin and gold certificates increased from $1,579 million to $2,499 million (Federal Reserve Bulletin, July 1, 1917, p. 562). Over this period, total money in circulation increased from $3,715 million to $4,745 million, indicating that the Treasury accounted for about 89.3 percent of this change. The issue of Federal Reserve notes and the contraction of National Bank notes accounted for most of the remainder.

The seasonal behavior of interest rates in the United States may have been significantly affected by these gold flows. Clark (1986) suggests this possibility, but states: "In short, unless there was a perfect synchronization of liquidity effects and upward seasonal interest rate pressures, some vestiges of the seasonal pattern would still be evident (p. 116)." (15) There was, in fact, a close association between quarterly gold flows an the usual late summer and early autumn pressure on interest rates. As Table 3 shows, there was a two-fold increase in net gold imports in the second half of 1915 and a more than tripling in the second half of 1916. Although this observation is not proof that the gold flows are sufficient to reduce the seasonal behavior of interest rates, it supports that possibility.

More conclusive evidence is available from the behavior of call loan rates between 1914 and 1917. The call rate is the interest charged for loans collateralized by stock traded on the New York Stock Exchange. Call rate data are available weekly beginning in January 1890. (16) A regression using weekly call rates ([R.sub.t]) and net gold import ([N.sub.t]) shows the relationship between call rates and net gold imports: (17)

[R.sub.t] = monthly dummies + 0.828[R.sub.t-1] - 0.0022[N.sub.t] - 0.0072[N.sub.t-1], (1). (4.36) (-0.70) (-2.62)

where call rates the measured in percent, net gold imports are measured in millions of dollars, and the figures in parentheses are t-ratios. The adjusted R-squared is .77 and the sample begins on May 7, 1915, the earliest date for which the weekly gold figures are available from the Federal Reserve Bulletin, and ends on June 15, 1917. The regression equation implies that net gold imports in the current and previous week had a negative effect on interest rates during the period before the United States entered the war. Because net gold imports were larger in the second half of the year, as shown in Table 3, the dampening effect on interest rates would be higher for the period.

4. ADDITIONAL EMPIRICAL EVIDENCE

Three issues remain to be resolved regarding the effects of the Federal Reserve on the behavior of interest rates. The first is the question of whether the 1917 amendments created a structural change in the behavior of interest rates. Because they did alter the operating and note issue behavior of the Federal Reserve, it is expected that they would affect the stochastic structure of nominal interest rates. The weekly interest rate on call money loans at the New York Stock Exchange between 1890 and 1933 is used to investigate this proposition. [18]

TABLE 4
MIDDLE SWITCHPOINT USING WEEKLY DATA FOR THE INTEREST RATE
MODEL:
[R.sub.t] [alpha] + [beta][R.sub.t-1] + [epsilon.sub.t]
Month in 1917  -Log L   Posterior Odds Ratio
    March 9    3599.2          0.011
   March 16    3598.5          0.023
   March 23    3597.9          0.044
   March 30    3597.2     0.084
    April 6    3596.6          0.153
   April 13    3595.9          0.299
   April 20    3595.5          0.493
   April 27    3594.9          0.797
      May 4    3595.1          0.728
     May 11    3594.8          0.913
     May 18    3595.5          0.484
     May 25    3594.9          0.867
     June 1    3594.7          1.000
     June 8    3595.2          0.652
    June 15    3595.8          0.370
    June 22    3596.5          0.184
    June 29    3595.9          0.325
     July 6    3597.3          0.076
    July 13    3596.7          0.144
    July 20    3602.9          0.000
    July 27    3608.6          0.000
   August 3    3607.9          0.000
  August 10    3607.9          0.000
  August 17    3607.7          0.000
  August 24    3607.1          0.000
  NOTEs: Log is the log of the likelihood function.  The
posterior odds ratio is calculated assuming a diffuse
prior.  The log of the likelihood
function without any switchpoints is equal to -4218.5.

Following Mankiw, Miron, and Weil (1987) and Fishe and wohar (1990). an AR(1) model of interest rates is estimated as part of a switching regression. (19) Unlike these other studies, however, the switching program used here searches for multiple switchpoints. Three switchpoints are found for the call rate series: January 24, 1908, June 1, 1917, and January 8, 1930. (20) Table 4 shows the likelihood value and posterior odds ratio for observations in the neigborhood of the middle switchpoint. Each of these switchpoints represents an important change in monetary or financial policy. In January 1908, most major banks reinstated cash payments to depositors, which were suspended after the Knickerbocker Trust crisis of October 1907. In Jnuary 1930, the economy was beginning its slide into the Great Depression with significant liquidation of bank credit following the October Stock market crash. Finally, the middle switchpoint occrs exactly when the Congress was debating the Federal Reserve Act amendments of 1917 and only three weeks before they actually became law.

The second question of interest is whether the seasonal behavior of interest rates and Federal Reserve notes was affected by the 1917 amendments. This issue is

TABLE 5
SIGNIFICANCE TESTS OF DUMMY VARIABLES IN THE AR(1) MODEL
            Period             F-Statistic     Sample Size
Calls Rates:
  January 1890-December 1933     4.24 (*2)        2295
  January 1890-January 1908      3.16 (*2)         939
  January 1908-June 1917         3.42 (*2)         488
  January 1908-July 1914         5.82 (*2)         342
     July 1914-June 1917         1.20              151
     July 1917-January 1930      0.89              806
     July 1914-January 1930      1.03              657
  January 1930-December 1933     1.61              206
FR Notes Issued:
November 1914-December 1929     25.37 (*2)         788
November 1914-June 1917          1.41              133
    June 1917-December 1929     26.22 (*2)         654
  (*2) significant at the 1 percent level.

investigated using two approaches. The first approach examines whether the inclusion of monthly dummy variables improves the performance of the simple AR(1) model estimated above; that is, the null hypothesis is that the monthly dummies are not significantly different from each other. Table 5 summarizes the results of this exercise, using various sample periods for both call rates and Federal Reserve notes. (21) The series on Federal Reserve notes is from the Federal Reserve Bulletin, 1915-1929. (22) Because the Bulletin was not published until May 1915, the Annual Report, Comptroller of the Currency, 1915, was used to complete tghe series between 1914 and 1915. The F-statistic reported in Table 5 is a test of whether the dummy variables are significantly different from one another. The call rate series shows that the monthly dummies offer some explanatory power before July 1914, suggesting that the seasonal behavior of interest rates changed after this period. The notes issued series, estimated as an AR(1) model, shows that there is a strong dependence on monthly effects after 1917, but no corresponding dependence between November 1914 and June 1917. The evidence from Table 5 then implies that call rates lost their seasonal behavior after 1914, but that Federal Reserve notes did not begin their seasonal movement until after June 1917.

The second approach examines the autocorrelation coefficients for the call rate and notes issued series over various periods. These results are shown in Tables 6 and 7. The first column under each time period lists the autocorrelation coefficients for the series in level form and second column lists the autocorrelation coefficients after the series has been differenced and adjusted for first-order serial correlation. If a series is AR(1) with autoregressive coefficient p, then first differences after multiplying the lagged series by p will create a white noise filtered. If the

[TABULAR DATA OMITTED]

series is AR(1) but has a seasonal component in its error structure, first differences as described will remove the autogressive structure but, generally, retain the seasonal structure. The fifty-second lag in the "filtered" column of each of these tables is the important autocorrelation coefficient. If the fifty-second lag or nearby lags are statistically significant (two times the estimated std. error), then there is evidence of a seasonal factor affecting the series.

[TABULAR DATA OMITTED]

The call rate results in Table 6 show only a minor seasonal factor, at the fifty-first lag, in the 1890 to 1908 period. They show more evidence of seasonality in the 1908 to 1917 period, at the fifty-second lag, and somewhat weaker evidence for the 1917 to 1929 period. Except for the 1917 to 1929 period, where there is evidence of a small indeterministic seasonal component, these results are consistent with those shown in Table 5. (23) The notes issued results are much clearer. There is a strong seasonal component between 1917 and 1929, but no evidence of such an effect in the 1914 1917 period, which is what the dummy variable regressions found. The behavior of Federal Reserve notes issued is also illustrated by Figure 1, where there is a pronounced seasonal component toward the end of each year beginning after the passage of the 1917 amendments.

The third question of interest is whether the Federal Reserve pursued a policy of smoothing the seasonal behavior of nominal interest rates between 1914 and 1917 using its endowment of resources. The fact that notes issued did not follow a seasonal pattern does not prove that the Federal Reserve was ineffective before the amendments of 1917. The Federal Reserve was endowed with significant resources--mostly member bank reserves--in the form of lawful money. It could have chosen to use these resources, instead of notes, to smooth the seasonal behavior of interedst rates. If this were the case, then the investment policy of the Federal Reserve would be seasonal between 1914 and 1917.

To test whether investments were seasonal, weekly data on discounts, open market holdings, purchases of government securities, and member bank reserves were collected from the Federal Reserve Bulletin for the period beginning November 20, 1914, and ending June 15, 1917. The Federal Reserve's level of participation in the market is measured by the level of security purchases, member bank reserves, and by computing the ratio of security purchases to gross resources, where gross resources are equal to total assets plus gold held by Federal Reserve agents. The gold holdings of the agents are added to total assets because the gold was given to the agents as collateral for notes issued, so the gold was ariginally an asset of the district banks. If security purchases or this ratio tended to increase or member bank reserves tended to decrease in the fall, then the Federal Reserve might have a claim to smoothing the seasonal behavior of interest rates.

Table 8 reports the results of regressing security purchases, member bank reserves, and the securities to gross resources ratio on monthly dummy variables and a time trend, after correcting for first-order autocorrelation. The coefficients in these regressions show only a slight tendency for Federal Reserve policy to be more active during the late summer and early autumn. This tendency is less pronounced for the security purchases and member bank reserves regressions. As a group, the monthly dummies in the security purchases ([F.sub.11,122] = 1.189), member bank reserves ([F.sub.11,122] = 1.499), and securities to gross resources ([F.sub11,122] = 1.601) regressions are not significantly different from one another. There is not enough evidence then to claim that the Federal Reserve implemented a policy of smoothing the seasonal behavior of interest rates by purchasing securities using its endowment of reserves.

[TABULAR DATA OMITTED]

5. CONCLUSION

The behavior of call money rates suggests that a change in the structure of interest rates coincided nearly perfectly with the passage of the 1917 amendments to the Federal Reserve Act of 1913. A closer investigation of the institutional behavior of the Federal Reserve reveals that these amendments were necessary for the successful financing of the First and subsequent Liberty bond campaigns. Before the amendments, the Federal Reserve acted like a sponge, absorbing more gold than it issued in notes. After the passage of the amendments, the Federal Reserve pursued an active policy of discounting commercial paper and fifteen-day notes of member banks with new Federal Reserve notes, which were issued without the high collateral requirements imposed by the 1913 Act. The seasonal factor in call money rates was significantly reduced, but not totally eliminated, after the 1917 mendments.

These findings are part of a broader controversy: Does the Federal Reserve have the capacity to affect real economic activity? The results suggest that the Federal Reserve, when given the operating leverage, can affect nominal interest rates and their seasonal structure. There are real effects then only to the extent that the seasonal structure of interest rates is a function of real economic activity, a question that has yet to be fully addressed.

LITERATURE CITED

Andrew, A. Piatt. Statistics for the United States: 1867-1909. National Monetary Commission, S. Doc 570, 61st Congress, 2d session, 1910.

Barsky, Robert B., N. Gregory Mankiw, Jeffery A. Miron, and David N. Weil. "The Worldwide Change in the Behavior of Interest Rates and Prices in 1914." European Economic Review 32 (June 1988).

Board of Governors of the Federal Reserve System. Annual Report. Washington: USGPO, 1914-15.

Board of Governors of the Federal Reserve System. Banking and Monetary Statistics: 1914-1941. Washington: USGPO, 1976.

Brown, William A. The International Gold Standard Reinterpreted: 1914-1934. New York: National Bureau of Economic Research, 1940.

Bureau of the Census. Historical Statistics of the United States: Colonial Times to 1970. Bicentennial Edition, 1975.

Cahill, James C. Cahill's Consolidated Laws of New York. Chicago: Callaghan and Company, 1923.

Clark, Truman A. "Interest Rate Seasonals and the Federal Reserve." Journal of Political Economy 94 (February 1986), 76-125.

Comptroller of the Currency. Annual Report. Washington: USGPO, 1914-15.

Fishe, Raymond P. H., and Mark Wohar. "The Adjustment of Expectations to a Change in Regime: Comment." American Economic Review 80 (September 1990), 968-76.

Friedman, Milton, and Anna J. Schwartz. A Monetary History of the United States, 1867-1960. Princeton: Princeton University Press, 1963.

Lewis, George. "Changes in Federal Reserve Act Simplify Form of Weekly Bank Statement." Journal of the American Bankers Association (August 1917), 96-98.

Macaulay, Frederick R. The Movements of Interest Rates, Bond Yields and Stock Prices in the United States since 1856. New York: National Bureau of Economic Research, 1938.

Mankiw, N. Gregory, Jeffery A. Miron, and David N. Weil. "The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve." American Economic Review 77 (June 1987), 358-74.

Miron, Jeffery A. "Financial Panics, the Seasonality of the Nominal Interest Rate, and the Founding of the Fed." American Economic Review 76 (March 1986), 125-140.

Miron, Jeffery A. "The Founding of the Fed and the Destabilization of the Post-1914 Economy." Working Paper #2701, National Bureau of Economic Research, September 1988.

Romer, Christina D. "World War I and the Postwar Depression: A Reinterpretation Based on Alternative Estimates of GNP." Journal of Monetary Economics 22 (1988), 91-115.

Shiller, Robert. "Can the Fed Control Real Interest Rates?" In Rational Expectations and Economic Policy, edited by Stanley Fisher, pp. 117-68. Chicago: University of Chicago Press, 1980.

Wagel, Srinivas R. "Gold, Currency and the Standard of Value under War Conditions." Journal of the American Bankers Association (December 1917), 451-53.

Willis, Henry P. The Federal Reserve System: Legislation, Organization, and Operation. New York: The Ronald Press Company, 1923.

Willis, Henry P. The Theory and Practice of Central Banking. New York: Harper & Brothers, 1936.

Willis, Henry P., and William H. Steiner. Federal Reserve Banking Practice. New York: D. Appleton & Company, 1926.

(1) An alternative explanation for the decrease of interest rate seasonals after 1914 is that they were smoothed when several major European countries abandoned the gold standard. Although Germany, Austra-Hungary, Russia, and France suspended specie payments, the United States and Great Britain, the two countries that show an unambiguous decline in interest rate seasonals according to Clark (1986), did not suspend gold payments until late 1917. In addition, this explanation is suspect because the seasonality of interest rates did not reappear when the postwar gold standard was reintroduced in 1925.

(2) For discount rate policy during 1914, see Annual Report, Federal Reserve Boad, 1914, Exhibit M. For prevailing market rates, see Commercial and Financial Chronicle, Bank and Quotation Section, December 5, 1914, p. 18.

(3) Secretary McAdoo gave the district banks just three weeks notice when he advanced the opening date from early 1915 to mid-November, 1914, so many operating rules were not fully developed. There was no policy on how to conduct open market operations, for example, until the latter half of December 1914 when district banks were "given authority to purchase Government bonds within the limits of prudence, as they might see fit" (Annual Report, Federal Reserve Board, 1914, p. 16). Even then, the district banks did not purchase any government bonds until June 1915 (Annual Report, Comptroller of the Currency, 1915, I, p. 137).

(4) Material for this discussion is drawn from the Federal Reserve Act of 1913 and a series of editorials published by the Commercial and Financial Chronicle on July 3 1915 (p. 3), July 10, 1915 (pp. 70-71), July 17, 1915 (pp. 158-60), and August 7, 1915 (pp. 398-401).

(5) In a letter to the Commercial and Financial Chronicle, dated August 2, 1915, Governor Strong states:

The relations which are gradually developing between the reserve banks and member banks naturally lead to the member banks regarding the Federal Reserve banks as the first source for supplies of currency as needed. As these calls are made, the Federal Reserve banks must elect whether they shall retain the gold under the plan now in operation and ship Federal Reserve notes. To illustrate what gradually may be accomplished, investigation recently disclosed that of the $1,070,000,000 of gold certificates now in circulation in the United States, $722,000,000 was held by national, State and Federal Reserve banks, leaving $348,000,000 of gold certificates in the pockets of the people of the United States. If for one-half, or even more of this gold, Federal Reserve notes can be substituted, and the gold mobilized in the Federal Reserve banks, the banking system of this country will, for the first time, rest upon a foundation of gold such as we have long desired. Last Fall, this country witnessed a rather humiliating exhibition when it became necessary for the Federal Reserve Board and a committee of bankers to invite and even urge contributions of gold from over 800 national banks, in order that a means might be available for the citizens of this country to pay maturing indebtedness to Europe, when exchange could not be purchased (reprinted in Commercial and Financial Chronicle, August 7, 1915, pp. 412-13).

(6) Brown (1940), pp. 27-47, discusses the mobilization of gold and silver resources in the central banks of Europe after the outbreak of the war.

(7) Willis and Steiner (1926), pp. 91-100, discuss the agents' role in the Federal Reserve System.

(8) This assumes that the 35 percent gold reserve against deposits specified in the 1913 Act is not a binding constraint on note expansion.

(9) As Lewis (1917) notes, the 1917 amendments also lowered reserve requirements against demand deposits. This change would have released a sizable amount of funds for banks to lend had it not been accompanied by the provision that all reserves must be held in Federal Reserve banks. Previously, member banks had held some required reserves in their own vaults or with correspondent banks in either reserve or central reserve cities.

(10) Friedman and Schwartz (1963), p. 191, discuss this aspect of the 1917 Act.

(11) The rollover question is discussed by the Law Department of the Federal Reserve Board (Federal Reserve Bulletin, October 1, 1917, p. 765).

(12) The Federal Reserve Bulletin comments frequently on the use of fifteen-day discounts. In the June 1917 issue, for example, it is stated: "As a matter of fact, over 55 per cent of paper rediscounted (i.e., exclusive of collateral notes discounted with the Federal Reserve Banks) during April was paper in denominations of over $10,000, chiefly 15-day paper, presented by the larger banks with the view of securing additional funds to subscribe to government securities (p. 485)."

(13) The figures in Table 2 were culled from Federal Reserve Bulletins, 1917-18, and differ significantly from those reported in Banking and Monetary Statistics, 1914-1941 (p. 340). Because of the detailed reporting and repeated discussion of the fifteen-day discounts in the Federal Reserve Bulletins during 1917 it is believed that the figures in Table 2 are the correct ones.

(14) For example, if a member bank presented a $10 million fifteen-day note, backed by Treasury certificates, bearing a yield of 3.25 percent for discount at 3 percent, the Treasury acting through the Federal Reserve would, in effect, be paying that bank $1,027.40 every time the note rolled over.

(15) Clark (1986), note 22, also argues that if the gold imports dampered the seasonal in U.S. interest rates, then there would be an increased seasonal effect observed in England, our main trading partner, and Europe. Because the seasonal disappeared in England after 1914, the same year as in the United States, he argues that gold imports alone cannot be the reason for the reduced seasonal in the United States. This view, however, ignores the liquidity- and inflation-producing events occurring in England and Europe after the onset of World War I. Wagel (1917) has estimated that England's per capita note production increased from .68 pounds to 5 pounds sterling between August 1, 1914, and November 1, 1917. Similar changes are noted for the European countries. He also estimates that $17.45 billion in new notes were created over this period by the various participants and allies involved with the war. Of this amount, only $650 million is attributed to the United States. A dosage of liquidity/inflation of this magnitude may have eliminated interet rate seasonals or made them undetectable in England and Europe during the intrawar

(16) The call rate data are available from the 1910 statistical report of the National Monetary Commission, compiled by A. Piatt Andrew (1910), which covers the period 1890 to 1909, and from the Bank & Quotation Section of the Commercial and Financial Chronicle, which covers the period up to 1933. These data are also reprinted in Banking and Monetary Statistics, 1914-41 (p. 452) for periods after 1919.

(17) This regression was estimated using both OLS and a procedure to correct for first-order autocorrelation. Although there was almost no difference in coefficient values, the Durbin "h" statistic (h = 2.47) indicated that the autoregressive procedure was more appropriate, and so these results are reported in equation (1). In both cases, some monthly dummies were individually significant (April, June, and November), but they were not jointly significantly different.

(18) Generally these are competitively determined rates, but as noted by Macaulay (1938, p. A339) from September 1917 through January 1919 the call money market was under the supervision and control of the New York Stock Exchange Money Committee. It is not clear, however, that the Money Committee had any effect on the market mechanism over this period. For the seventeen-month period prior to Money Committee control, monthly call rates averaged 2.88 percent with a standard deviation of .743 percent. For the 17 months the Money Committee exercised control, call rates averaged 5.05 percent with a standard deviation of .682 percent. Because other interest rates, notably commercial paper rates, also rose between these two periods, the Money Committee probably had little or no effect on the stochastic behavior of call rates.

(19) According to the Durbin-Watson statistic (E-W = 0.507), the series exhibits significant positive autocorrelation in level form between 1890 and 1933.

(20) The chi-squared test of significance for these switchpoints equals 1247.6, which is significant at the 1 percent level with six degrees of freedom.

(21) The inclusion of monthly dummy variables did not affect the location of the switchpoints found for the call rate series.

(22) The weekly note issue series published in the Federal Reserve Bulletin ends in December 1929 and does not resume again until the mid-1930s.

(23) As measured by the relative size of the autocorrelation coefficients, the indeterministic seasonal is 62.5 percent of that observed between 1908 and 1917.

RAYMOND P. H. FISHE is professor of economics at the University of Miami.