Retail Sweep Programs and Monetary Asset Substitution
Barry E. Jones
State University of New York at Binghamton
Adrian R. Fleissig
California State University, Fullerton
Thomas Elger
Lund University, Sweden
Donald H. Dutkowsky*
Syracuse University
`
Abstract
This paper examines how retail sweep programs affect monetary asset substitution.
Estimates from the Fourier flexible form reveal that sweeping generates systematic and
sometimes large distortions in estimated bank depositor substitution elasticities.
JEL Classification Codes: E41, E52, G21
Keywords: Monetary Asset Substitution; Fourier; Retail Sweep Programs
*Department of Economics, Syracuse University, Syracuse, NY 13244. E-mail:
[email protected].
1.
Introduction
Retail sweep programs in the US have increased substantially since their introduction in
January 1994. Under retail sweeping, banks reclassify customer balances from checkable
deposits into savings deposits (Anderson 2002). Sweeping reduces required reserves as these
funds are reported as savings deposits. However it leaves customers perceived transactions
deposits unchanged, since they have unrestricted access to swept funds.
Retail sweeping has become a key issue in bank reform. The Federal Reserve has
repeatedly called for paying interest on bank reserves held as deposits (e.g. Feinman 1993,
Meyer 1998, Kohn 2004). Another reform, described in Bennett and Peristiani (2002), proposes
to remove reserve requirements entirely. All these writings criticize sweeping as a wasteful use
of bank resources that has arisen due to reserve avoidance.1
This paper investigates the effect of retail sweep programs on monetary asset
substitution. Asset substitution by bank depositors in response to changes in relative user costs
leads to movements in deposits, reserves, and the composition of reserves. Since this behavior
can affect the transmission of monetary policy and its effectiveness (e.g. Belongia and Ireland
2006), such estimates provide important policy information. By distorting the reported money
measures, however, retail sweeping leads to distorted estimates of depositor user cost elasticities.
We examine this distortion by comparing findings from monetary asset data adjusted for
retail sweep programs with the reported data. Hicks and Allen own price elasticities and
Morishima elasticities of substitution are computed from estimates of the semi-nonparametric
Fourier flexible form. This procedure allows elasticities to vary over the sample period and has
1Retail
sweep programs have also led to distortion in the M1 monetary aggregate (Anderson 2002) and the demand
for narrow money (Dutkowsky and Cynamon 2003).
1
been widely used to study asset substitution (Davis and Gauger 1996, Fisher and Fleissig 1997,
Drake, Fleissig and Swofford 2003).
2.
Reported and Sweep-Adjusted Data
Seasonally adjusted monetary data come from FRED. The set of monetary assets for the
reported data is
A1 = CUR+DD, currency (including travelers checks) + demand deposits;
A2 = OCD, other checkable deposits;
A3 = SAV, savings deposits;
A4 = STD, small time deposits;
A5 = MMMF, retail money market mutual funds.
Sweep-adjusted monetary assets are formed using estimates of the amount of funds swept
from DD and OCD. The cumulative sum of newly initiated retail sweep programs (CSWEEP) is
reported by Anderson (2002). Shares of swept funds from DD and OCD (SDD and SOCD) are
computed from data that decompose the cumulative amount of swept funds.2 Series for
estimated funds swept from DD and OCD are SWEEP_DD = SDD*CSWEEP and
SWEEP_OCD = SOCD*CSWEEP. Following Jones, Dutkowsky, and Elger (2005), the sweepadjusted assets subtract the swept funds from SAV and add them back to DD and OCD:
A1* = CUR+DD + SWEEP_DD;
A2* = OCD + SWEEP_OCD;
A3* = SAV – (SWEEP_DD + SWEEP_OCD).
Since STD and MMMF are unaffected by retail sweeping, A4* = A4 and A5* = A5. All assets
are converted into real terms using the personal consumption expenditure price index (P).
The user cost of the ith monetary asset (Barnett 1978) is given by i = P(R – Ri)/(1 + R),
where R and Ri are respectively the interest rate on the benchmark asset and on the ith asset. The
2We
thank Spence Hilton and Dennis Farley for cumulative swept funds data from DD and OCD for 1987:1-2004:1.
For 2004:2-2004:8, we set SDD and SOCD equal to their values in 2004:1.
2
6 month Treasury bill rate serves as R.3 Own rates for A2-A5 come from the Federal Reserve
Bank of St. Louis, and the own rate for A1 equals zero. Since sweeping is invisible to
depositors, we use the same user costs for the reported and sweep-adjusted data. The sample
consists of monthly observations for 1987:1-2004:8.
3. Elasticities and the Fourier Flexible Form
The Hicksian elasticity of substitution between the ith and jth assets, for i, j, = 1, 2, …, n,
is given by ijh   ln xih /  ln  j , where x ih denotes the Hicksian demand for the ith asset. The
more widely used Allen elasticity of substitution is directly related to the Hicksian measure
by  ija ijh / si , where si =  i xih / E x is the share of total expenditure (Ex) on the ith asset.
Hicksian and Allen elasticities are better suited to measure own price elasticities, where i = j.
With three or more assets, Blackorby and Russell (1989) show that the correct elasticity of
substitution is the Morishima elasticity MEij  si ( aji  iia ). MEij measures how the ratio of the ith
to the jth asset responds, holding utility constant, to a change in its relative user cost (i/j).
We obtain elasticities from estimating a demand system of monetary assets using the
Fourier flexible form. The Fourier provides a semi-nonparametric approximation to the
unknown data generating function and is defined as (Gallant 1981):
f (v,  )  u 0  b v 


A 
J

1
v Cv    u 0  2  u j cos( jk v)  w j sin( jk v)  ,
2
 1 
j 1

A
where C    u 0 k k ,  = {b, u0, uj, wj, for j = 1,2,…,J and  = 1,2,…,A}, and v is a vector
 1
of expenditure normalized user costs. A multi-index, k, denotes partial differentiation of the
utility function. The Fourier share equations:
3All
interest rates are annualized one month yields on a bond interest basis. To ensure non-negative user costs, we
add 200 basis points to the benchmark rate as in Fisher, Hudson, and Pradhan (1993).
3




A 
J

v i bi    u 0 v k   2  u j sin( jk  v)  w j cos( jk  v) k i v i
 1 
j 1

s i ( v,  ) 
,
A 
J

b v    u 0 v k   2  j u j sin( jk  v)  w j cos( jk  v) k  v
 1 
j 1

are estimated with additive residuals. We use a first-order vector autoregressive process as in
Berndt and Savin (1975) to correct for serial correlation. The estimation is performed in
International TSP 4.5 and all results are available upon request.
4.
Empirical Results
We begin by comparing Hicksian own price elasticities from the reported data to the
corresponding estimates from the sweep-adjusted data in Figures 1a-1c. Beginning in 1994,
elasticities for reported CUR+DD systematically exceed those of the sweep-adjusted data in
Figure 1a
Hicksian Own Price Elasticities
Currency plus Demand Deposits (A1 vs A1*)
-0.30
Sweep Adjusted (A1*)
Reported (A1)
-0.40
-0.50
-0.60
-0.70
-0.80
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
absolute magnitude (Figure 1a). The same result holds for the OCD elasticities (Figure 1b). In
contrast, elasticities for reported SAV are systematically lower in absolute magnitude than for
sweep-adjusted SAV (Figure 1c). Further evidence on the role of sweeping comes from own
price elasticities for A4 versus A4* and A5 versus A5* (not reported here). Nearly identical
estimates for the reported and corresponding sweep-adjusted measures occur over the entire
sample period, as neither STD nor MMMF are directly affected by retail sweeping.
4
Figure 1b
Hicksian Own Price Elasticities
Other Checkable Deposits (A2 versus A2*)
-0.65
Sweep Adjusted (A2*)
Reported (A2)
-0.75
-0.85
-0.95
-1.05
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
The estimated Allen own price elasticities generate larger differences than the Hicksian
measures, since the reported and sweep-adjusted assets have different expenditure shares for
transactions deposits and savings deposits. Since the numerators of the shares for A1* and A2*
include swept funds, they are larger than those for A1 and A2. Consequently, dividing the
Figure 1c
Hicksian Own Price Elasticities
Savings Deposits (A3 versus A3*)
-0.35
Sweep Adjusted (A3*)
Reported (A3)
-0.45
-0.55
-0.65
-0.75
-0.85
-0.95
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Hicksian elasticities by their respective shares results in a greater spread between the Allen own
elasticity for transactions deposits for the reported data and its sweep-adjusted counterpart,
beginning in 1994.
The differences can be substantial. By 2004:8, the spread is about 1.5 (-4.5 versus -3.0)
for A1 versus A1*, and close to 10 (-17.5 versus -7.5) for A2 versus A2*. At the same time, the
expenditure share for A3 is smaller than for A3*. Therefore, dividing the Hicksian elasticities by
5
the respective shares results in a larger difference between the Allen own price elasticities for A3
versus A3*, which by 2004:8 is about 0.5 (-1.0 versus -1.5).
Summary statistics for estimated Morishima elasticities over the sample period appear in
Table 1. Retail sweeping generally leads to greater elasticities of substitution and additional
variability. Means of the Morishima elasticities for the reported data exceed the sweep-adjusted
data in seventeen out of twenty cases. Higher standard deviations for the reported data occur in
all cases, more than double the sweep-adjusted values for ME12, ME31, ME42, ME43, and ME52.
Elasticity
ME12
ME13
ME14
ME15
Table 1
Morishima Elasticities of Substitution
Reported Data
Sweep-Adjusted
Mean
Std. Deviation
Mean
Std. Deviation
1.731
0.225
1.524
0.110
1.497
0.129
1.387
0.072
2.391
0.391
2.173
0.306
1.052
0.067
1.029
0.060
ME21
ME23
ME24
ME25
1.210
1.288
1.511
0.835
0.059
0.087
0.174
0.116
1.194
1.213
1.498
1.085
0.032
0.048
0.127
0.060
ME31
ME32
ME34
ME35
1.523
2.055
2.719
1.654
0.124
0.272
0.467
0.110
1.351
1.526
2.165
1.452
0.060
0.138
0.300
0.077
ME41
ME42
ME43
ME45
1.388
1.496
1.446
1.293
0.105
0.168
0.103
0.073
1.304
1.351
1.329
1.298
0.054
0.044
0.048
0.072
ME51
ME52
ME53
ME54
1.036
0.713
1.304
1.511
0.035
0.227
0.084
0.174
1.022
1.108
1.226
1.497
0.025
0.073
0.052
0.127
Figures 2a and 2b show graphs of ME31 and ME32, Morishima elasticities of substitution
between reported and sweep-adjusted SAV and each of the transactions assets due to a change in
the user cost of SAV. These elasticities most directly encompass sweeping behavior.
6
Figure 2a
Morishima Elasticity of Substitution :
Savings Deposits and Currency Plus Demand Deposits (ME31)
2
Sweep Adjusted
Reported
1.75
1.5
1.25
1
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Estimates for reported versus sweep-adjusted CUR+DD (Figure 2a) and OCD (Figure 2b) are
Figure 2b
Morishima Elasticity of Substitution:
Savings Deposits and Other Checkable Deposits
(ME32)
2.75
Sweep Adjusted
Reported
2.5
2.25
2
1.75
1.5
1.25
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
systematically higher for the reported data, with the gap widening after 1994. This result is again
consistent with retail sweeping, with banks as well as customers making portfolio reallocation.
5.
Conclusion
Systematic differences, and in some cases large discrepancies, in estimated own price
elasticities and Morishima elasticities occur due to retail sweep programs. Sweeping
overestimates depositor own price elasticities for transactions deposits and underestimates those
for savings deposits. In addition, estimated elasticities of substitution between savings deposits
and transactions deposits are systematically greater for the reported data. Retail sweeping also
generates greater variability in substitution among monetary assets.
7
Advocates for paying interest on reserves or removing reserve requirements emphasize
the inefficiency of sweeping. Our findings provide empirical evidence for other adverse effects
as well, and further support such reforms. Since both proposed reforms would obviate the need
for sweeping, distortion in estimates of asset substitution would be eliminated, thereby reflecting
portfolio decisions of depositors alone.
References
Anderson, R.G. (2002), “Federal Reserve Board Data on OCD Sweep Account Programs,”
http://research.stlouisfed.org/aggreg/swdata.html.
Barnett, W.A. (1978), "The User Cost of Money," Economics Letters, 1(2), 145-149.
Belongia, M.T. and P.N. Ireland (2006), “The Own-Price of Money and the Channels of
Monetary Transmission,” Journal of Money, Credit, and Banking 38(2), 429-445.
Bennett, P. and S. Peristiani (2002), “Are US Reserve Requirements Still Binding?” FRBNY
Economic Policy Review, 8, 53-68.
Berndt, E.R. and N.E. Savin (1975), "Estimation and Hypothesis Testing in Singular Equation
Systems with Autoregressive Disturbances," Econometrica, 43, 937-957.
Blackorby, C. and R.R. Russell (1989), "Will the Real Elasticity of Substitution Please Stand
Up?" American Economic Review, 79(4), 882-888.
Davis, G.C. and J. Gauger (1996), “Measuring Substitution in Monetary-Asset Demand
Systems,” Journal of Business and Economic Statistics, 14(2), 203-208.
Drake, L., A.R Fleissig, and Swofford (2003), "A Semi Nonparametric Approach to Neoclassical
Consumer Theory and Demand for UK Monetary Assets," Economica, 99-120.
Dutkowsky, D.H. and B.Z. Cynamon (2003), “Sweep Programs: The Fall of M1 and the Rebirth
of the Medium of Exchange,” Journal of Money, Credit, and Banking, 35(2), 263-279.
Feinman, J.N., (1993), “Reserve Requirements: History, Current Practice, and Potential
Reform,” Federal Reserve Bulletin, 569-589.
Fisher, D. and A.R. Fleissig (1997), "Monetary Aggregation and Demand for Assets," Journal of
Money, Credit and Banking, 29(4), 458-475.
Fisher, P., S. Hudson, and M. Pradhan (1993), “Divisia Measures of Money: An Appraisal of
Theory and Practice,” Bank of England Working Paper #9.
8
Gallant, A.R. (1981), "On the Bias in Flexible Functional Forms and an Essentially Unbiased
Form: The Fourier Flexible Form," Journal of Econometrics, 211-245.
Jones, B.E., D.H. Dutkowsky, and T.Elger (2005), “Sweep Programs and Optimal Monetary
Aggregation,” Journal of Banking and Finance, 29, 483-508.
Kohn, D.L. (2004), “Testimony of Governor Donald L. Kohn, Regulatory Reform Proposals,
Before the Committee on Banking, Housing, and Urban Affairs, US Senate,” 2004.
Meyer, L.H. (1998), “Testimony of Governor Laurence H. Meyer, The Payment of Interest on
Demand Deposits and Required Reserve Balances, Before the Committee on Banking,
Housing, and Urban Affairs, US Senate,” March 3, 1998.
9