The Federal Reserve collects market data from the dealers with which it has a trading relationship—the so-called primary dealers. The data cover transactions, positions, financing, and settlement activities in U.S. Treasury securities, agency debt securities, mortgage-backed securities (MBS), and corporate debt securities. The data are consolidated and released publicly by the Fed and used by market analysts to monitor market conditions and the activities of primary dealers.

Some market analysts claim that the financing data show a substantial rise in dealer leverage in recent years. A common argument is that low short-term interest rates have spurred a large number of carry trades—transactions in which market participants borrow money on a short-term basis and invest in higher yielding long-term assets. Such trades increase investors’ exposure to interest rate changes. A consequent concern is that rises in interest rates could lead to large trading losses, a rapid deleveraging, and high volatility.1

This edition of Current Issues examines what the primary dealer financing data can tell us about dealer leverage. We begin by reviewing why and how dealers engage in financing activities. We then discuss which financing data are reported to the Fed, which data the Fed releases, and what the data mean. From these data, we calculate and evaluate two measures of dealer leverage and discuss the relationship between our preferred leverage measure—net financing—and dealer risk taking. Lastly, we analyze the relationship between net financing, interest rates, and interest rate volatility.

Our analysis of the financing data indicates that dealer borrowing involving fixed-income securities has grown only modestly in recent years. Moreover, because the rise is not associated with increased net positions, there is little evidence to suggest greater risk taking. Nonetheless, we do find some evidence to support the idea that dealer leverage is related to interest rates and volatility. In particular, financing tends to decline both before and after increases in interest rates and volatility.

How and Why Financing Occurs
Financing activities provide the funds and securities needed by market participants when they trade securities. Primary dealers buy and sell fixed-income securities for several reasons: to make markets for their customers, to take speculative positions, and to hedge positions in derivatives and other fixed-income securities.

A dealer taking a long position by buying a security must pay the seller. If the dealer does not pay out of available funds, it has to finance the position by borrowing. Dealers typically do this by providing securities they already own as collateral for the loan. For example, a dealer financing the purchase of a Treasury security could borrow funds from a corporate treasurer while providing another fixed-income security as collateral.

Conversely, a dealer taking a short position by selling a security it does not own must deliver the security to the buyer. A short position is financed by borrowing the security while providing cash or other securities as collateral. For instance, a dealer selling short a Treasury note could borrow the note from a mutual fund and provide cash as collateral.

Several types of transactions are used to finance long and short positions, but all are essentially equivalent to the collateralized borrowings described above. In a repurchase agreement (repo), for example, an investor sells a security while agreeing to buy it back at a higher price on a future date. From the perspective of the funds lender—which buys a security while agreeing to resell it—such agreements are called reverse repos.

In a securities lending transaction, an investor lends a security while accepting another security or cash as collateral. If cash is offered as collateral, then the securities lender pays the borrower a cash collateral fee. When another security is offered as collateral, the securities borrower pays the lender a fee for the benefit of borrowing a specific security.

What Is Reported and by Whom
Market data are reported to the Federal Reserve Bank of New York by the primary government securities dealers. The dealers represent an important but limited subset of the fixed-income market. Moreover, dealer reporting entities may not include all trading activities of the larger organizations of which they are a part. Furthermore, additions to and withdrawals from the list of primary dealers as well as internal reorganizations change the dealer population over time.2

Dealers report (via form FR 2004) their transactions, positions, financing, and settlement activities in U.S. Treasury securities, agency debt securities, MBS, and corporate debt securities.3 Reporting occurs weekly, as of the close of business each Wednesday, for broad categories of securities.4 Data are submitted by 4:00 p.m. the following business day, usually Thursday, through the Federal Reserve’s Internet Electronic Submission System.

The Fed publicly releases summary data each Thursday at 4:15 p.m., one week after the data are collected.5 The data are aggregated across all dealers and are available only for broad categories of securities; individual issue data and individual dealer data are not made public. Table 1 shows the financing data reported for May 19, 2004, from the release of May 27.

Dealer Financing Data
Dealer financing is reported on a gross basis, distinguishing between "securities in" and "securities out" for each asset class. Securities in refer to securities received by a dealer in a financing arrangement, whereas securities out refer to securities delivered by a dealer. For instance, if a dealer enters into a repo, in which it borrows funds and provides securities as collateral, it would report securities out. If a dealer enters into a security-for-security transaction, it would report both securities in and securities out.

In addition, repos and reverse repos are reported across all asset classes combined. These memo items are subsets of the broader securities-in and securities-out figures, and reveal the share of dealer financing conducted through these particular transactions. On May 19, 2004, outstanding repos totaled $2,743 billion, accounting for 89 percent of the $3,076 billion in securities out across all asset classes (calculated from Table 1). In contrast, reverse repos of $2,022 billion represented only 69 percent of the $2,929 billion in securities in.

The financing data are also broken down by the length of the financing arrangement. Overnight and continuing (or open) agreements refer to financing activities that are either for one business day or that can be terminated on demand by either party, but otherwise continue indefinitely. Term agreements refer to financing activities that have an original specified length of more than one business day.

The financing figures reported are the actual funds paid or received. In the case of a repo, for example, a dealer reports the funds received on the settlement of the repo, and not the value of the pledged securities. When only securities are exchanged, the dealer reports the market value of the pledged securities.

Periodic Changes in the Dealer Reports
The Fed considers changes to the dealer reports every three years to improve market monitoring while limiting the reporting burden. The last major set of changes commenced with the reports for the week ending July 4, 2001. Asset classes covered by the reports were expanded to include corporate debt securities. Moreover, the financing data became more detailed by providing information by asset class.

In other ways, the financing data became less detailed in July 2001. Until then, dealers reported financing by the type of transaction employed, and not just total financing and repo financing. The earlier data are instructive, as they reveal that securities borrowing transactions account for the overwhelming majority of securities-in transactions that are not reverse repos. Thus, when dealers borrow funds, they rely largely on repos, but when they borrow securities, they rely on both reverse repos and securities borrowing transactions. Most of the securities borrowed are against cash rather than against other securities.6

Interpreting Dealer Financing Data

Net Repo Financing
Some market observers contend that the financing data show a marked rise in dealer leverage in recent years. This interpretation is typically based on an analysis limited to the repo data. In particular, net repo financing—the net amount of funds primary dealers borrow through fixed-income security repos—is calculated as repos minus reverse repos. The measure has indeed increased strongly in recent years (Chart 1), and was $722 billion on May 19, 2004 (calculated from Table 1).

However, net repo financing is an incomplete and potentially misleading measure of dealer leverage. First and foremost, it does not account for transactions that are essentially equivalent to repos but not reported as such. In particular, securities lending transactions perform the same economic function as repos but are not reported as repos in the financing data. Another drawback is that the asset classes covered by the measure have changed over time, most recently being expanded to include corporate debt securities in July 2001.

Net Financing
A measure of dealer leverage that addresses some of the drawbacks of the net repo financing measure is net financing. Net financing, calculated as securities out minus securities in, gauges the net amount of funds primary dealers borrow through all fixed-income security financing transactions. The measure thus encompasses all financing transactions reported by the dealers. Moreover, because securities in and securities out are reported by asset class, net financing can be computed to exclude corporates so as to generate a time series consistent with respect to the covered asset classes.7

On May 19, 2004, overall net financing was $147 billion (Table 1, combined securities out minus combined securities in), considerably less than net repo financing. In addition, net financing has consistently been far less than net repo financing for at least the past decade (Chart 1). Much of the increase in net repo financing in recent years can thus be explained by the measure’s exclusion of transactions that are essentially equivalent to repos. Moreover, an examination of net financing including and excluding corporates shows that some of the increase in net repo financing since July 2001 is likely explained by the inclusion of corporates since that time.

Although net financing is consistently far below net repo financing, it is apparent that net financing did indeed increase in recent years. In particular, net financing rose from mid-2000 to mid-2003, a period in which the federal funds target rate dropped from 6.50 percent to 1.00 percent. Net financing then fell sharply in spring 2004 amid growing expectations of a near-term increase in the funds rate.

Financing, Leverage, and Risk Taking
While net financing accurately measures the funds borrowed by primary dealers through fixed-income security financing transactions, it is itself an imperfect gauge of dealer leverage. In particular, it does not take into account the capital of primary dealers. Net financing divided by total assets (net financing plus capital) would be a better leverage measure, but the capital of the primary dealer units is generally not observed.

It is also worth noting that leverage captures only one dimension of risk taking. Leverage is related to risk taking because, for a given set of positions, higher leverage puts more capital at risk. However, just as leverage changes over time, so does the riskiness of positions, often making the net effect on capital at risk ambiguous. For example, a dealer could increase its leverage while decreasing the riskiness of its positions, causing its capital at risk to rise, fall, or remain the same.

Another way in which leverage and risk taking diverge is through speculative positions that do not affect net financing. For example, if a dealer buys a thirty-year bond and sells a three-month bill, the net effect on financing may be zero. An additional way is through forward, futures, and options positions, which for the most part do not require financing. As a result, if a dealer buys and finances, say, a ten-year note and sells a ten-year-note futures contract, its net financing will increase even though the two positions might largely offset one another in terms of risk exposure.

Primary dealer positions data include forward positions and thus provide evidence on dealer risk taking beyond the financing data.8 Net positions (in Treasuries, agencies, and mortgage-backed securities) tended to track net financing between mid-1994 and mid-2000, but the measures diverged markedly afterward (Chart 2). In fact, net positions have shown no long-term increase in recent years, suggesting that the upswing in net financing that has occurred is not attributable to an increase in net positions and is thus not clearly associated with greater risk taking. (It is worth noting, however, that the net positions measure is an imperfect indicator of risk taking for many of the reasons that net financing is; the relationship between these two measures is discussed further in the appendix.)

Dealer Financing, Interest Rates, and Interest Rate Volatility
Some market analysts have suggested that low short-term interest rates have spurred a large number of carry trades in recent years, thereby increasing investors’ exposure to interest rate changes. A resultant concern is that increases in interest rates could lead to large trading losses, a rapid deleveraging, and high volatility. By comparison, our analysis suggests that dealer borrowing has grown only modestly, and that the increase is not associated with greater position taking. Nonetheless, it is worth examining the extent to which our preferred leverage measure—net financing—is related to interest rates and interest rate volatility.

We analyze the relationship between net financing, interest rates, and interest rate volatility using correlation and regression techniques. The variables we employ, in addition to net financing, are the fed funds target rate (the primary tool of monetary policy), the three-month/fed funds spread (a predictor of future monetary policy), the ten-year/three-month spread (a predictor of the business cycle [Estrella and Mishkin 1996]), Moody’s Baa/Aaa spread (an indicator of credit quality), and implied interest rate volatility (a proxy for expected future volatility). Summary statistics for the variables are presented in Table 2.

Concurrent Relationships among Dealer Financing, Interest Rates,
and Interest Rate Volatility
Implied interest rate volatility is the only variable significantly correlated with net financing in terms of weekly changes, and the relationship is not especially strong economically or statistically (Table 3). One explanation for the negative relationship is that increased volatility elevates the amount of dealer capital at risk, causing dealers to reduce their net positions because of capital constraints.9 A second explanation has causality going the other way, with dealer speculation stabilizing security prices.10 Additional evidence we present later is consistent with both hypotheses.

Note that our statistical analysis is undertaken in terms of changes, and not levels, because the variables exhibit time trends over the sample period.11 An analysis in terms of levels would produce results that are quite different, and that might seem strong, but that are in fact spurious and misleading. For example, the fed funds target rate is highly negatively correlated with net financing in terms of levels, but the more appropriate tabulation in terms of changes reveals essentially no correlation.

Dynamic Relationships among Dealer Financing, Interest Rates,
and Interest Rate Volatility
To capture the dynamic relationships among the variables, we regress weekly changes in the variables on lagged weekly changes in the other variables, as well as on their own lagged values. Because the independent variables in the regressions are lagged, we can interpret the coefficients as predictive, that is, a change today in an independent variable predicts a future change in the dependent variable. The coefficients quantify average effects, based on past data, and do not show that one variable causes another.12

We find that increases in the fed funds target rate and in the three-month/fed funds spread precede decreases in net financing (Table 4, row 1). The coefficient of -0.70 for the fed funds rate, for example, implies that a one-time increase in the funds rate of 1 basis point is followed, on average, by a decrease in net financing of $700 million over thirteen weeks, holding other variables constant. Increases in the fed funds rate, or expectations thereof, may be inducing dealers to reduce their exposure to rising rates, leading to a decline in net financing.

Increases in implied interest rate volatility are also found to precede decreases in net financing. The coefficient of -1.78 for implied interest rate volatility (Table 4, row 1) suggests that a rise in volatility of 1 basis point predicts a decline in net financing of $1.78 billion. This finding is consistent with the argument that dealers face capital constraints. As volatility increases, so does the riskiness of positions, causing dealers to unwind positions and reduce net financing.

Not only do interest rates and volatility help to explain net financing, but net financing helps to explain interest rates and volatility (Table 4, column 1). An increase in net financing of $1 billion is expected to be followed by a decrease of 0.34 basis points in the fed funds rate over the next thirteen weeks and a fall of 0.32 basis points in implied interest rate volatility, although neither relationship is especially strong statistically. The interest rate finding is consistent with the argument that dealers reduce net financing in anticipation of an increase in the fed funds rate; the volatility finding supports the hypothesis that dealer positions tend to stabilize asset prices.

Net financing is not significantly related to either the ten-year/three-month spread, which predicts recessions, or the Baa/Aaa spread, which tends to increase during recessions. It therefore appears to be unrelated to expectations about changes in the business cycle.

Conclusion
Data reported to the Federal Reserve cover primary dealer financing activities involving U.S. Treasury securities, agency debt securities, mortgage-backed securities, and corporate debt securities. While some market analysts argue that the data show a considerable rise in dealer leverage in recent years, our analysis indicates that dealer borrowing involving fixed-income securities has grown only modestly. Moreover, the increase that has occurred is not associated with increased net positions, so there is little evidence to suggest greater risk taking by primary dealers.

We do, however, find evidence to support the idea that dealer leverage is related to interest rates and volatility. In particular, increases in interest rates and volatility precede decreases in financing while decreases in financing precede increases in interest rates and volatility. We leave it to future research to explore these relationships more fully and to ascertain the reasons behind them.