Credit policy as fiscal policy.
Lucas, Deborah
III.D. Fiscal Multipliers
The CBO (201 lb) defines a fiscal multiplier as the change in a
nation's economic output generated by each $ 1 of the budgetary
cost of a change in fiscal policy. It reports a range of fiscal
multipliers, reproduced in table 1, for various types of expenditures
authorized under the ARRA. The wide range reflects the conflicting
evidence in the literature on their size. Auerbach and Gorodnichenko
(2012) show that some of the size variation is a function of the
business cycle, with much larger multipliers in downturns that are at
the high end of the range reported by the CBO.
As discussed in section I.B, under the interpretation that stimulus
affects aggregate demand largely through the spending of hand-to-mouth
consumers or liquidity-constrained households, these multipliers apply
more naturally to incremental borrowing than to credit subsidies. The
CBO multipliers may in fact be conservative measures when applied to
incremental borrowing because the offset from savings may be smaller
than for government programs that distribute funds to some people who,
unlike borrowers, have no immediate desire to spend them. However, how
to map the different credit programs into the listed categories is not
obvious. Student loans, small business loans, and other loans probably
correspond most closely to "transfer payments to individuals,"
in that they put money directly into people's hands that is likely
to be spent fairly quickly. Mortgages are most closely related to the
extension of a first-time homebuyer credit, although the terms of this
program are quite different from those for mortgages, and many mortgage
borrowers are not first-time borrowers.
To capture the cyclical variation in multipliers at different
phases of the business cycle suggested by Auerbach and Gorodnichenko
(2012), higher multipliers are applied in the distressed scenarios than
for normal times. For the distressed scenario, the multipliers are set
to 2.0 for student loans, business loans, and other loans. This choice
of multipliers is toward the high end of CBO's range for transfers
to individuals, and slightly less than the multiplier suggested by
Auerbach and Gorodnichenko (2012) for total spending in a recession. For
normal conditions, the multipliers for these programs are set to 0.5,
which is slightly above that suggested by Auerbach and Gorodnichenko
(2012) for an expansion. For Fannie Mae and Freddie Mac, I assume much
smaller multipliers of 0.3 under distressed conditions and 0.2 under
normal conditions. For the other mortgage programs, the multipliers are
set to 0.4 and 0.3, respectively. These numbers fall in the low to
middle range that the CBO reports for the first-time homeowner credit.
The choice of small multipliers for mortgage programs reflects the fact
that refinancing accounted for about 73 percent of U.S. mortgage
originations in 2010 according to Freddie Mac's 2010 Annual Report.
(16) Refinancing frees up some cash for borrowing-constrained households
because it lowers monthly mortgage payments (both through a lower
interest rate and because principal is reamortized), but the increase in
free cash flow is much less than the principal amount refinanced. Most
purchase mortgages also can be expected to have a limited stimulus
effect because the money is spent on existing structures that are not
part of output. The effects of larger and smaller multiplier are
considered in the sensitivity analysis.
IV. Stimulus Estimates and Sensitivity Analysis
Recall from section I.B that incremental borrowing, [DELTA]B, which
is attributable to federal credit assistance net of crowding out, is
[DELTA]B = dA + S(dB/dS)-C. The first two terms are incremental
borrowing along the extensive and intensive margins, which can be easily
quantified based on the assumptions made in section III. Crowding out,
C, is set to 0, due to the slack in the financial system in 2010.
Applying the multipliers to AB for each program and summing up the
results yields the estimate of fiscal stimulus. Dividing the fiscal
stimulus by the sum of credit subsidies quantifies the bang for the buck
of credit subsidies.
IV.A. Subsidy Totals and Borrowing Increases along the Intensive
Margin
Table 2 summarizes the assumed subsidy rates and the estimates of
incremental demand along the intensive margin, S(dB/dS). Multiplying the
loan disbursements by the subsidy rates gives a subsidy cost for
traditional credit programs of $29.7 billion. (27) The GSE subsidies add
$40.9 billion to that, bringing the total estimated subsidies for 2010
to $70.6 billion.
The construction of the estimates of incremental borrowing for each
program on the intensive margin are shown in table 3, using the demand
elasticities, disbursements, and subsidy rates described for each
program in section III. In total, $107 billion in additional borrowing
is attributed to this channel in 2010, mostly from the housing programs.
Note that in an economy without credit market frictions, and under
normal market conditions, federal credit subsidies of this magnitude
would be expected to have modest effects on economic aggregates; $70.6
billion is only about 11 percent of the $666 billion in nondefense
discretionary spending, and 0.5 percent of GDP. In this case, the
subsidies would cause some redistribution of wealth from taxpayers to
borrowers. Demand would increase for subsidized loans, and there would
be some crowding out of unsubsidized loans. Because eligibility for
subsidies is linked to specific investments and increases the demand for
them, some or all of the subsidy would be absorbed in higher relative
factor prices (for example, subsidized mortgages encourage more housing
purchases, putting upward pressure on house prices).
IV.B. Borrowing along the Extensive Margin and Aggregate Stimulus
Effects
The computations of borrowing for each program along the extensive
margin and the computation of aggregate stimulus effects under the
normal and stress scenarios are shown in tables 4 and 5. Incremental
borrowing along the extensive margin is found by multiplying the assumed
share of incremental loan demand for each program by the actual
disbursement amounts. The sum of incremental loan demand along the
extensive and intensive margins for each program is multiplied by the
corresponding fiscal multiplier to estimate incremental output. Note
that the intensive margin effects reported are the same in both tables 4
and 5 because they were directly computed for 2010.
Under the assumption that the effects in 2010 were at the midpoint
of the two scenarios that show $587 billion in incremental output with
distressed conditions and $101 billion with normal conditions, the
conclusion is that federal credit programs generated an estimated $344
billion in incremental output. This additional output was at a cost of
$70.6 billion, which translates into a substantial $4.86 of stimulus per
$1 of taxpayer cost. By comparison, the CBO estimated that the ARRA
increased output by $392 billion, with an average multiplier on spending
of less than 1.5.
IV.C. Sensitivity Analysis
Clearly, the estimate of the stimulus effects of credit policies is
highly sensitive to the many assumptions that went into the
calculations, and the true value could be considerably more or less than
the $344 billion that has been presented as the most plausible point
estimate. However, the conclusion that federal credit policies have
provided a significant amount of fiscal stimulus in recent years would
be robust to a fairly wide range of parameter choices. The normal and
distressed scenarios in themselves may provide lower and upper bounds
for the range of plausible stimulus effects of falling between $101
billion and $587 billion. Although there is no way to assign
probabilities, the midpoint was chosen to represent the best estimate
because 2010 was an early year in the recovery, but credit markets
remained tight. The assumptions about mortgages and student loans are
the most critical because of the large size of these programs. To the
extent that student loans are the most credible source of stimulus and
that the private student loan market was highly distressed, the estimate
of $219 billion in stimulus from this program in distressed conditions
could be taken as a tighter lower bound. A range of $219 billion to $587
billion still leaves an uncertainty factor of roughly 3 from top to
bottom, but this is narrower than the fivefold ranges of multiplier
uncertainty shown in table 1 above. Given the uncertainties about
multiplier effects, it would be difficult to make a strong case for
further narrowing the range.
As noted by Gale (1991), tax-exempt borrowing also provides federal
credit subsidies via the tax code. He reports a subsidy rate of 19
percent for this assistance. The CBO's multiplier range for
transfers to state and local governments are similar to those for
transfers to individuals. Applying this subsidy rate and a mid-range
multiplier of 1 to 2010 long-term municipal issuance volume, which
totaled a record high $430 billion, translates into an additional
stimulus of $81.7 billion. The base case estimates are also
conservative, in that the subsidy estimates mostly exclude
administrative costs.
V. Discussion and Conclusions
A unique aspect of U.S. credit markets is the large presence of
ongoing government-backed direct loan and loan guarantee programs, most
notably Fannie Mae and Freddie Mac, the student loan programs, and the
FHA, but also more than 100 smaller programs. Collectively, these
activities provided credit subsidies of varying sizes on $1.6 trillion
of loans disbursed in 2010, and they relaxed credit-rationing
constraints on many borrowers. Taking into account the likely effects
these programs had on causing borrowing in that year to be higher than
what would have been extended privately in their absence, and applying a
multiplier to these incremental balances similar to those applied to
traditional government spending and tax policies, yields an estimate of
fiscal stimulus from these programs in 2010 of roughly $344 billion,
similar to the amount that the CBO attributes to the ARRA. This estimate
is in some ways conservative because it excludes other forms of credit
support such as tax breaks on municipal bonds, which would add an
estimated $82 billion to the stimulus, and omits administrative costs
from subsidy estimates. Although there is considerable uncertainty about
this point estimate, its size suggests that the effects of fiscal policy
cannot be fully understood without taking the stimulus effects of
federal credit programs into account. And it also suggests that
structural changes in the larger federal credit programs have potential
macroeconomic and fiscal policy implications.
A few points deserve emphasis. Although the uncertainty surrounding
the reported point estimates of the stimulus is considerable, under a
wide range of plausible parameterizations the estimated effects are
extremely large. Hence, rather than dismissing these effects because the
impact is difficult to measure precisely, it makes sense to continue to
look for better ways to measure this phenomenon. It is also important to
remember that although the focus here has been on federal credit
programs as a relatively low-cost source of fiscal stimulus and
automatic stabilization during a severe economic downturn, these
programs have significant costs during more normal times that also must
be considered in assessing their overall welfare effects. These costs
include the likelihood that overly lax federal credit policies,
particularly mortgage-related subsidies, were one of the exacerbating
causes of the 2007 financial crisis. A final caution is that the cost of
the stimulus, though low compared with traditional fiscal policy, is
significantly more than the budgetary cost of the programs that are
calculated under rules that cause the full economic costs of credit
extension to be underreported.
This analysis raises several fundamental questions. The first is
how these results should change our perceptions about the depth of the
downturn and the effectiveness of other types of fiscal and monetary
stimuli. If the high-end estimates presented here are correct, then one
might conclude that either the economy was in worse shape than most
economists thought, or conventional fiscal and monetary stimuli had less
effect than some had previously estimated. Addressing this issue in
depth is beyond the scope of this analysis. My view is that there
remains a great deal of disagreement among prominent economists about
how effective either fiscal or monetary policy was, or what would have
transpired in the absence of those policies. This professional
uncertainty is reflected, for instance, in the wide range of multipliers
reported in table 1 that were derived from surveying the literature. My
preferred estimate of additional output of $344 billion in 2010,
although large in comparison with the ARRA, is small compared with the
$14.66 trillion in GDP and deficit of $1.3 trillion. Hence, although the
findings here should shift the perception of the total amount of the
fiscal stimulus that was provided, it need not significantly shift
one's prior beliefs about the severity of the downturn or the
effectiveness of other policies. With regard to monetary policy, the
analysis raises the intriguing question of whether the recovery of the
housing market benefited more from the loosening of borrowing
constraints via federal housing credit programs or from the Federal
Reserve's actions that lowered interest rates.
A further question is definitional, and concerns whether credit
policy, which up until now has acted as what might be called a shadow
stimulus, should be classified as fiscal policy, as monetary policy, or
as a third category of its own. The subsidies associated with credit
policies clearly are an expenditure of economic resources by the
government, and hence are fiscal in nature. The treatment of credit
subsidies in the federal budget as costs (albeit underestimated costs)
concurs with this view. At the same time, the channel through which the
subsidies translate into fiscal stimulus--by accommodating increased
borrowing, and thereby increased spending--is different than for other
fiscal policies. This difference was taken into account in the analysis
by applying estimates of fiscal multipliers to incremental borrowing
rather than directly to the subsidies. Nevertheless, because subsidy
provision is the root cause of the increase in aggregate demand, it
seems reasonable to consider these policies as part of fiscal policy,
broadly defined. Although the policies have some similarities with
actions the Federal Reserve took at about the same time through its
creation of emergency lending facilities, the case for treating the
programs as part of monetary policy is weak. It has been observed that
to the extent that the Federal Reserve was taking uncompensated credit
risk through its emergency facilities, its actions were fiscal and not
monetary. In fact, the Federal Reserve claimed that it was providing
liquidity and not taking credit risk. The CBO (2010b) for the most part
concurred; it found that because most of the risky Federal Reserve
facilities were backstopped by the Treasury through TARP (which was
counted as part of fiscal policy) or were otherwise protected against
credit losses, the fiscal costs of these facilities were in fact small.
By contrast, federal credit programs involve significant uncompensated
risk transfers and little liquidity provision.
There are also interesting areas for further research. An
intriguing question is to what extent similar channels for fiscal policy
through credit policy can be seen in other countries. In the case of
Europe, although governments affect credit by intervening heavily in the
banking system, there is much less reliance on U.S.-style government
credit programs. This raises the possibility that one reason for the
relatively strong U.S. recovery is that this channel for fiscal stimulus
is less available than in Europe.
ACKNOWLEDGMENTS I am grateful for suggestions from Alan Auerbach,
Janice Eberly, Bill Gale, and Damien Moore. I also would like to thank
seminar participants at Chulalongkorn University, the Brookings
Institution, the Becker Friedman Institute's 2014 Macro Financial
Modeling Conference, and the Federal Reserve Bank of Cleveland for their
comments on an early version of this paper.
References
Auerbach, Alan J., and Yuriy Gorodnichenko. 2012. "Measuring
the Output Responses to Fiscal Policy." American Economic Journal:
Economic Policy 4, no. 2: 1-27.
Bach, Laurent. 2014. "Are Small Businesses Worthy of Financial
Aid? Evidence from a French Targeted Credit Program." Review of
Finance 18, no. 3: 877-919.
Barnett, Jeffrey L., and Phillip M. Vidal. 2012. "State and
Local Government Finances Summary: 2010." Governments Division
Brief no. G10-ALFIN. Washington: U.S. Census Bureau,
http://www2.census.gov/govs/local/10_summaryreport.pdf
CBO (Congressional Budget Office). 2007. "Federal Financial
Guarantees under the Small Business Administration's 7(a)
Program." Washington. https://www. cbo.gov/publication/19220
--. 2010a. "The Budget and Economic Outlook: An Update."
Washington. https://www.cbo.gov/publication/21670
--. 2010b. "The Budgetary Impact and Subsidy Cost of the
Federal Reserve's Activities during the Financial Crisis."
Washington, https://www.cbo.gov/ publication/21491
--. 2010c. "CBO's Budgetary Treatment of Fannie Mae and
Freddie Mac." Washington, https://www.cbo.gov/publication/41887
--. 2010d. "Costs and Policy Options for Federal Student Loan
Programs." Washington, https://www.cbo.gov/publication/21018
--. 2011a. "Accounting for FHA's Single-Family Insurance
Program on a Fair-Value Basis." Letter to the Honorable Paul Ryan.
Washington. https://www. cbo.gov/publ ication/41445
--. 2011b. "Estimated Impact of the American Recovery and
Reinvestment Act on Employment and Economic Output from January 2011
through March 2011." Washington,
https://www.cbo.gov/publication/41461
--. 2011c. "Federal Loan Guarantees for the Construction of
Nuclear Power Plants." Washington,
https://www.cbo.gov/publication/41510
--. 2012. "Fair-Value Accounting for Federal Credit
Programs." Washington. https://www.cbo.gov/publication/43027
--. 2014. "Fair-Value Estimates of the Cost of Selected
Federal Credit Programs for 2015 to 2024." Washington,
https://www.cbo.gov/publication/45383
De Andrade, Flavio, and Deborah Lucas. 2009. "Why Do
Guaranteed SBA Loans Cost Borrowers So Much?" Working paper.
http://www.kellogg.northwestern.
edu/faculty/deandrade/deandrade_lucas_sba2009.pdf
DeFusco, Anthony A., and Andrew Paciorek. 2014. "The Interest
Rate Elasticity of Mortgage Demand: Evidence from Bunching at the
Conforming Loan Limit." Finance and Economics Discussion Series no.
2014-11. Washington: Board of Governors of the Federal Reserve System.
Elliott, Douglas J. 2011. Uncle Sam in Pinstripes: Evaluating U.S.
Federal Credit Programs. Brookings Institution Press.
Financial Management Service. 2011. Treasury Bulletin, March.
Washington: U.S. Department of the Treasury,
https://www.fiscal.treasury.gov/fsreports/rpt/ treasBulletin/b2011_1.pdf
Gale, William G. 1990. "Federal Lending and the Market for
Credit." Journal of Public Economics 42, no. 2: 177-93.
--. 1991. "Economic Effects of Federal Credit Programs."
American Economic Review 81, no. 1: 133-52.
Hull, John, Mirela Predescu, and Alan White. 2005. "Bond
Prices, Default Probabilities and Risk Premiums." Journal of Credit
Risk 1, no. 2: 53-60.
Jaffee, Dwight M., and Thomas Russell. 1976. "Imperfect
Information, Uncertainty, and Credit Rationing." Quarterly Journal
of Economics 90, no. 4: 651-66.
Jeske, Karsten, Dirk Krueger, and Kurt Mitman. 2013. "Housing,
Mortgage Bailout Guarantees and the Macro Economy." Journal of
Monetary Economics 60 no. 8: 917-35.
Lacker, Jeffrey M. 1994. "Does Adverse Selection Justify
Government Intervention in Loan Markets?" Economic Quarterly 80,
no. 1: 61-95.
La Porta, Rachel, Florencio Lopez-de-Silanes, and Andrei Shleifer.
2002. "Government Ownership of Banks." Journal of Finance 57,
no. 1: 265-301.
Lucas, Deborah. 2012. "Valuation of Government Policies and
Projects." Annual Review of Financial Economics 4: 39-58.
--. 2014. "Evaluating the Government as a Source of Systemic
Risk." Journal of Financial Perspectives 2, no. 3: Article 4.
Lucas, Deborah, and Damien Moore. 2010. "Guaranteed versus
Direct Lending: The Case of Student Loans." In Measuring and
Managing Federal Financial Risk, edited by Deborah Lucas. University of
Chicago Press.
Lucas, Deborah, and Marvin Phaup. 2010. "The Cost of Risk to
the Government and Its Implications for Federal Budgeting." In
Measuring and Managing Federal Financial Risk, edited by Deborah Lucas.
University of Chicago Press.
Munnell, Alicia H., Jean-Pierre Aubry, and Dan Muldoon. 2008.
"The Financial Crisis and Private Defined Benefit Plans."
Issue in Brief no. 8-18. Chestnut Hill, Mass.: Boston College, Center
for Retirement Research.
OMB (Office of Management and Budget). 2010. "Budget of the
U.S. Government: Federal Credit Supplement, Fiscal Year 2011."
Washington: White House.
https://www.gpo.gov/fdsys/pkg/BUDGET-2011-FCS/pdf/BUDGET 2011-FCS.pdf
--. 2011. "Budget of the U.S. Government: Analytical
Perspectives, Fiscal Year 2012." Washington: White House.
https://www.gpo.gov/fdsys/pkg/BUDGET 2012-PER/pdf/BUDGET-2012-PER.pdf
--. 2015. "Budget of the U.S. Government: Analytical
Perspectives, Fiscal Year 2016." Washington: White House.
https://www.gpo.gov/fdsys/pkg/BUDGET2016-PER/pdf/BUDGET-2016-PER.pdf
Passmore, Wayne, Shane M. Sherlund, and Gillian Burgess. 2005.
"The Effect of Housing Government-Sponsored Enterprises on Mortgage
Rates." Real Estate Economics 33, no. 3: 427-63.
Remy, Mitchell, Deborah Lucas, and Damien Moore. 2011. "An
Evaluation of Large-Scale Mortgage Refinancing Programs." Working
Paper no. 2011-4. Washington: Congressional Budget Office.
Rothschild, Michael, and Joseph Stiglitz. 1976. "Equilibrium
in Competitive Insurance Markets: An Essay on the Economics of Imperfect
Information." Quarterly Journal of Economics 90, no. 4: 629-49.
Smith, Bruce. 1983. "Limited Information, Credit Rationing,
and Optimal Government Lending Policy." American Economic Review
73, no. 3: 305-18.
Smith, Bruce D., and Michael J. Stutzer. 1989. "Credit
Rationing and Government Loan Programs: A Welfare Analysis." Real
Estate Economics 17, no. 2: 177-93.
Stiglitz, Joseph E., and Andrew Weiss. 1981. "Credit Rationing
in Markets with Imperfect Information." American Economic Review
71, no. 3: 393-410.
Whalen, Charles J., and Felix Reichling. 2015. "The Fiscal
Multiplier and Economic Policy Analysis in the United States."
Working Paper no. 2015-02. Washington: Congressional Budget Office.
Williamson, Stephen D. 1994. "Do Informational Frictions
Justify Federal Credit Programs?" Journal of Money, Credit and
Banking 26, no. 3, pt. 2: 523-44.
Comments and Discussion
COMMENT BY
ALAN J. AUERBACH During the Great Recession, the struggle to come
up with effective countercyclical policies while confronting significant
financial market disruption and the zero lower bound led to innovations
in monetary policy that in some cases tested the boundaries between
monetary and fiscal policy. In her paper, Deborah Lucas offers us
another prospective element of an expanded view of countercyclical
fiscal policy: the use of government credit subsidies to increase
private sector borrowing at both the intensive and extensive margins.
Lucas argues that credit policy played a significant role in lessening
the severity of the Great Recession, and that the output multipliers of
U.S. government credit policy are much larger than standard fiscal
policy multipliers, especially in periods of recession.
This is an interesting and thought-provoking paper that should
stimulate further research. My comments all relate to its striking
implication that very large multipliers should make credit policy a
major element of countercyclical policy.
WHAT DOES THE MULTIPLIER multiply? A critical step in Lucas's
multiplier calculation is her assumption that multipliers--taken from
the fiscal policy literature relating output changes to changes in
government taxes, transfers, and direct purchases--should be applied to
the additional borrowing per unit of subsidy. Thus, because the subsidy
amount is just a fraction of the induced increase in borrowing, there is
a first-stage multiplier before the standard multiplier is applied.
Lucas argues that for the liquidity-constrained, access to additional
borrowing should have a similar impact as the receipt of grants,
although she does make adjustments for programs and circumstances where
the refinancing of existing loans is important.
To decide whether this approach makes sense, it is useful to
clarify how the multiplier is being defined. In the recent literature
from which Lucas draws, one may find a variety of definitions--including
the impact effect on output of a shock to, say, government spending,
d[Y.sub.t]/d[G.sub.t]; the cumulative effect on output over some
horizon, d([[summation].sup.T.sub.i=0][Y.sub.t+i])/d[G.sub.t]; the peak
effect on output over some horizon, [MATHEMATICAL EXPRESSION NOT
REPRODUCIBLE IN ASCII]; and the cumulative effect on output relative to
the initial spending shock plus the cumulative government spending
induced by the initial shock over the same period,
d([[summation].sup.T.sub.i=0][Y.sub.t+i])/d([[summation].sup.T.sub.i=0][G.sub.t+i]). Whether it makes much of a difference which approach one
chooses depends, among other things, on the nature of the policy
shock--for example, whether it is temporary or long-lived. In the case
of the borrowing shocks that Lucas considers, there is effectively a
reversal of policy in subsequent years, as increased borrowing induces
increased payments of principal and interest. Thus, one would expect a
smaller cumulative impact relative to the initial shock than would be
the case for typical fiscal policy interventions. However, to the extent
that repayments are small relative to initial borrowing over the
relevant horizon (which Lucas indicates is the case) and borrowers are
liquidity constrained, the eventual reversal of shocks does not seem
quantitatively important. The bigger issue, in my view, is the
assumption that output responds to the initial increase in borrowing
rather than to the present value embedded subsidy or some other measure
of policy. On this issue, a key question is who is doing the initial
borrowing; and the paper appears to offer more than one answer. The
theoretical model that Lucas develops to explain why small subsidies can
lead to large increases in borrowing, as depicted in figure 1 of her
paper, indicates that the expansion at the extensive margin comes from
low-risk borrowers being enticed into the market. But the narrative
surrounding the empirical section strongly suggests that much of the
expansion reflects lending to individuals who have no other access to
credit. This story is more consistent with her multiplier assumptions,
but not with her theory. My sense is that the latter perspective is more
what Lucas has in mind, which supports her multiplier assumption but
also raises other issues, such as how estimated subsidies should respond
to expansions of lending.
Overall, I agree that cash up front should also matter, not just
the present value subsidy; but I am not entirely convinced that the
impact on aggregate activity should be as large as is implied by
applying fiscal multipliers to the amount of new borrowing.
Are the multipliers themselves reasonable? Lucas relies heavily on
the empirical multiplier literature, particularly that from the
Congressional Budget Office (2011), in assigning multipliers to the
different credit programs, in good times and bad. Based on my own work
cited in her paper, I am of course very sympathetic to using larger
multipliers when the economy is in a distressed period than during a
normal period, the two cases she distinguishes in her analysis. The
biggest impact of this cyclical variation comes through student loans,
using multipliers based on treating them as similar to transfer payments
to individuals. Although students resemble recipients of transfer
payments in an important respect--they have low current incomes and
assets--I am unsure whether this similarity itself justifies the
multiplier assumption. In particular, student loans cover tuition, and
payments to universities are quite different than payments to private
producers. Given states' balanced budget requirements, reductions
in public university revenues could have large effects on aggregate
demand, but not necessarily right away. Also, one would expect access to
student loans to have a big impact on labor force participation among
younger workers, much bigger than among many groups receiving transfer
payments, notably the elderly. This, too, could reduce the relative
impact on aggregate activity of an increase in student loans.
I do not mean to be overly critical of the assumptions Lucas makes
here, given the lack of empirical evidence regarding the relevant
multipliers. But existing empirical estimates of multipliers are subject
to large standard errors, especially for particular components of the
federal budget, for which there are fewer reliable estimates.
Translating these estimates into multipliers for credit programs
involves another big layer of uncertainty. So we need to assume large
confidence intervals around the paper's multiplier estimates,
leaving aside the distinction between normal and distressed times.
SUBSIDIES AND CREDIT EXPANSION An important part of the
paper's analysis involves the estimation of subsidies for a range
of government credit programs. Lucas argues for a more rigorous approach
than is commonly used by the government itself, and which typically
yields larger estimated subsidy costs. But applying her methodology to
the current exercise still requires assumptions and judgment, because
2010--the year on which she focuses--was a very atypical one in credit
markets. She does one set of calculations for the subsidy rates for
different programs, which are held fixed across the two
scenarios--distressed and normal--that she considers in her simulations.
But one might expect the composition of borrowers and the embedded
subsidy per dollar of new loans to vary across the two scenarios. For
example, if riskier borrowers enter, then the subsidy rate should be
higher.
If the subsidy rate does not increase substantially in the
distressed scenario, even though the federal government is accounting
for a substantially larger expansion of credit, this suggests that an
important function of government credit programs in bad times is the
provision of liquidity, rather than subsidies. Lucas distinguishes
between these two functions at the end of the paper, in contrasting the
effects of credit programs with the actions of the Federal Reserve
during the Great Recession, but I am not convinced that credit
subsidies, rather than simply the provision of credit, were important
during this period. Of course, the central role of financial market
disruption distinguishes the Great Recession from serious earlier
postwar recessions, and Lucas notes that there was little cyclical
variation in credit program disbursements in these earlier periods. But
if credit expansion follows the normal scenario in a typical recession,
then the amount of additional stimulus provided by credit subsidies
would also be small, even if the multiplier is large, for there would
not be much additional credit.
I am not clear from her paper how Lucas envisions the relationship
between subsidies and credit expansion. On the intensive margin, she
uses assumed elasticities to translate subsidy rates into credit
expansion; but on the extensive margin, she simply estimates the amount
of additional credit that is attributable to the programs. It is not
clear whether this expansion relates to the subsidy rates or to other
elements of the programs, such as the provision of liquidity. Thus, her
estimates of the program's bang for the buck--the additional output
per $1 in program subsidies, $4.86 per $ 1 in taxpayer cost--cannot
necessarily be thought of as a marginal effect that tells us how much
additional output we will get for an additional $1 subsidy.
CREDIT PROGRAM DESIGN AND STABILIZATION POLICY The tax
system's function in providing an automatic stabilizer is
important, and it can be substantially affected by changes in the tax
structure (Auerbach and Feenberg 2000; Kniesner and Ziliak 2002). For
example, the traditional tax reform approach of lowering tax rates and
broadening the tax base reduces the strength of automatic stabilizers.
With few exceptions (McKay and Reis 2013), however, little thought has
been given to how the stabilization function should guide tax policy
design. The same critique applies to government credit programs, now
that Lucas has called our attention to their potential role in
countercyclical policy. How should such concerns shape the design and
reform of government credit programs?
The question of how large subsidies should be, given the
countercyclical objective, is a hard one to address based on this paper,
for though it specifies the relationship between subsidies and borrowing
at the intensive margin (summarized in table 3 of the paper), it does
not do so for borrowing at the extensive margin. It may well be that the
subsidy rate, within a plausible range, has a relatively small impact on
cyclical variation. This may be fortunate, in a way, given that subsidy
rates seem to vary across programs in a manner that has no apparent
rationale and may be dictated by political rather than economic
objectives. For example, the program with the highest subsidy
rate--student loans--has the lowest response elasticity on the intensive
margin. However, scaling the amount of available credit to economic
conditions might help a lot, although perhaps much more in recessions
like the most recent one than in the typical recession. Also, as the
paper emphasizes, credit expansion can have both potential benefits and
economic costs, such as increased moral hazard and a distorted
allocation of capital. Here again, a crucial issue is who is doing the
additional borrowing when credit expands. Just as with the debate over
financial bailouts and financial regulation, we have an opportunity to
reform credit programs before the next recession, aiming to not only
improve the stabilization function that Lucas has highlighted but also
to do so without unnecessarily exacerbating the economic costs of
associated market distortions.
REFERENCES FOR THE AUERBACH COMMENT
Auerbach, Alan J., and Daniel R. Feenberg. 2000. "The
Significance of Federal Taxes as Automatic Stabilizers." Journal of
Economic Perspectives 14, no. 3: 37-56.
Congressional Budget Office. 2011. "Estimated Impact of the
American Recovery and Reinvestment Act on Employment and Economic Output
from January 2011 through March 2011." Washington,
https://www.cbo.gov/publication/41461
Kniesner, Thomas J., and James R Ziliak. 2002. "Tax Reform and
Automatic Stabilization." American Economic Review 92, no. 3:
590-612.
McKay, Alisdair, and Ricardo Reis. 2013. "The Role of
Automatic Stabilizers in the U.S. Business Cycle." Working Paper
no. 19000. Cambridge, Mass.: National Bureau of Economic Research.
COMMENT BY
WILLIAM G. GALE By the end of 2010, outstanding federal debt stood
at about $9 trillion. At the same time, as Deborah Lucas notes in her
paper, the government's outstanding loans and loan guarantees,
combined with the mortgages held or guaranteed by Fannie Mae and Freddie
Mac, equaled about $8 trillion, and other government-backed debt
obligations include deposit insurance, pension insurance, a variety of
implicit guarantees, the Troubled Asset Relief Program, and several
Federal Reserve programs. Looking at the sheer size of federal borrowing
versus federal lending is not an apples-to-apples comparison, of course,
but it suffices to make the basic point that federal lending has been
studied very little relative to other areas of federal activity. (1)
Lucas's paper is a welcome and intriguing exception and will,
I hope, spur further research on this topic. The paper, which might be
alternatively titled "The Accidental Stimulus," provides
illustrative calculations of the impact of federal credit subsidies on
output during good times and bad. Essentially, it is a "proof of
concept" that federal lending could have boosted the economy
significantly during and after the Great Recession, and thus provided an
automatic stabilizer and a stimulus that previous research and policy
discussions have not recognized.
The main line of argument is straightforward. Federal lending has
effects through two channels. It provides loans that would not have
otherwise been made by the private sector--the extensive margin. And, by
offering loans at subsidized rates, it increases the quantity of loans
demanded--the intensive margin. Summing the intensive and extensive
margin effects generates an estimate of the additional lending due to
the programs, to which program-specific multipliers are then applied to
generate the net increase in output. The increase in output is divided
by the subsidy value of new loan activity (not loan volume) to obtain
bang-for-the-buck estimates that are comparable to those that are
typically estimated for federal spending and tax changes.
Lucas estimates that federal credit policies, as they existed in
2010, provided stimulus effects totaling between $101 billion and $587
billion (between 0.7 and 4.0 percent of GDP). This is obviously a very
wide range of possible effects, but one that is consistent with the fact
that the key parameters--the extensive margin effects and the
multipliers--are extremely difficult to pin down. The extensive margin
effects are elusive due to a lack of evidence. And the multipliers are
hard to calculate both because there is uncertainty about the size of
the multipliers that apply to spending programs and because it is not
clear how appropriate it is to apply those multipliers that have been
estimated for spending programs to credit programs. Loans need to be
paid back, so the loan multiplier for a given sector of the economy may
be smaller than the spending multipliers. Conversely, a government loan
might crowd in a project that also has other loans and equity investment
associated with it, so the loan multipliers could be larger than the
spending multipliers.
Although the paper recognizes that there may be crowding out, the
quantitative calculations assume that there is none. In principle,
crowding out can occur in several ways. First, in order to lend an
additional $1, the government must first borrow it (holding other tax
and spending policies constant). This borrowing may somewhat reduce the
supply of credit to private investors, unless the supply of credit is
perfectly elastic or is not being utilized to provide loans to begin
with--conditions that Lucas argues apply to 2010. Second, when the first
type of crowding out exists, borrowing by one target group for federal
lending could crowd out borrowing by other target groups (Gale 1991).
This depends crucially on who the marginal borrower is in the market.
Suppose there are unsubsidized borrowers and two target groups that
receive government subsidies. If the government increases subsidies to
one target group, this group's demand will increase, and the
resulting increased flow of funds to this group may come at the expense
of loans to the other target group.
The assumption that there is no crowding out may be reasonable in
the bad times scenario, where the credit market collapses; but in the
good times scenario, it may be a stretch. If crowding out is more
extensive in good times than in bad ones, as seems plausible, there are
at least two implications. First, the $101 billion output stimulus
estimate during good times is overstated (and the true stimulus effect
could be zero or negative, depending on the extent of crowding out).
Second, the automatic stabilizer effect of federal credit would be even
larger than what Lucas estimates. Under Lucas's assumptions, as the
economy falls into a tailspin, not only would the extensive margin and
multiplier effects rise (as she already posits) but the crowding out
parameter would fall, too, adding a third channel for stimulus.
The theoretical model developed in the paper exists to provide
motivation, not to generate empirical specifications. The model has
several main findings. First, extensive margin effects can be large;
small changes in subsidies can generate big responses in outcomes. This
is a typical result for models of credit or insurance markets with
imperfect information (Rothschild and Stiglitz 1976; Stiglitz and Weiss
1981; Mankiw 1986; Gale 1990a, 1990b). But there is little evidence
pertinent to this issue. It would be helpful to have examples or
evidence on this subject, given this feature's ubiquity in
imperfect-information models of lending and insurance.
Second, in Lucas's model, federal credit crowds in safer
borrowers. This suggests that expansions of federal lending would be
associated with declines in default rates. My impression is that the
data do not support this implication. For example, the vast expansion of
federal housing loan guarantees that helped lead to the 2007-08
financial crisis and the Great Recession also led to an increase in
defaults, not a decline. The idea that subsidies crowd in safer
borrowers is a feature of the models developed by Michael Rothschild and
Joseph Stiglitz (1976), but not necessarily of other approaches (Smith
and Stutzer 1989; Gale 1990a). In Mankiw (1986), safer or riskier
borrowers could be crowded in by subsidies, depending on the parameter
values. In Gale (1990b), subsidies crowd in riskier borrower groups.
Determining whether the marginal borrower under federal credit programs
is safer or riskier than average seems like a very important direction
for new research.
More broadly, the paper suggests a need to rethink the economics of
the Great Recession and recovery. If Lucas's results are
correct--especially her bad times scenario--then either the economy
would have been in much worse shape in 2010 than people thought if there
had not been a stimulus, or the other fiscal and monetary stimulus
efforts provided were much less effective than people think. This
creates a bit of a puzzle, however, because one would expect a fiscal
stimulus to be most helpful in precisely the type of economy where
credit subsidies are helpful--when there are borrowing constraints,
incomplete markets, and other types of friction.
Another interesting issue is how credit subsidies would work in a
general equilibrium model that (i) generates business cycles and (ii)
has credit markets that shut down in bad times. It is plausible to think
of some aspects of credit policy as fiscal policy; for example, lending
at a subsidized rate seems quite similar to the spending subsidies or
tax wedges that are traditionally analyzed in public finance. It is also
plausible to categorize some aspects of credit policy as monetary
policy; during the Great Recession, the Federal Reserve set up
facilities to lend to banks so they could in turn lend to particular
target groups, which does not seem that different, in economic terms,
from the government lending directly to those groups (Kohn 2010). It is
less important, however, whether credit policy is characterized as
fiscal or monetary policy in those models, and more important to
understand the channels through which credit policies work and the
offsets that are created elsewhere in the economy. On the basis of
Lucas's intriguing findings, this seems like a fruitful direction
for future research.
REFERENCES FOR THE GALE COMMENT
Bosworth, Barry R, Andrew S. Carron, and Elisabeth H. Rhyne. 1987.
The Economics of Federal Credit Programs. Brookings Institution Press.
Elliott. Douglas J. 2011. Uncle Sam in Pinstripes: Evaluating U.S.
Federal Credit Programs. Brookings Institution Press.
Gale, William G. 1990a. Collateral, Rationing, and Government
Intervention in Credit Markets. In Asymmetric Information, Corporate
Finance, and Investment, edited by R. Glenn Hubbard. University of
Chicago Press.
--. 1990b. "Federal Lending and the Market for Credit."
Journal of Public Economics 42, no. 2: 177-93.
--. 1991. "Economic Effects of Federal Credit Programs."
American Economic Review 81, no. 1: 133-52.
Kohn, Donald L. 2010. "The Federal Reserve's Policy
Actions during the Financial Crisis and Lessons for the Future."
Speech given at Carleton University, Ottawa, May 13.
https://www.federalreserve.gov/newsevents/speech/kohn20100513a.htm
Mankiw, N. Gregory. 1986. "The Allocation of Credit and
Financial Collapse." Quarterly Journal of Economics 101, no. 3:
455-70.
Rothschild, Michael, and Joseph Stiglitz. 1976. "Equilibrium
in Competitive Insurance Markets: An Essay on the Economics of Imperfect
Information." Quarterly Journal of Economics 90, no. 4: 629-49.
Smith, Bruce D., and Michael J. Stutzer. 1989. "Credit
Rationing and Government Loan Programs: A Welfare Analysis." Real
Estate Economics 17, no. 2: 177-93.
Stiglitz, Joseph E., and Andrew Weiss. 1981. "Credit Rationing
in Markets with Imperfect Information." American Economic Review
71, no. 3: 393-410.
(1.) For broad discussion of the relevant issues, see Bosworth,
Carron, and Rhyne (1987) and Elliott (2011).
GENERAL DISCUSSION Robert Hall spoke first, asserting that the
central question of the paper was the extent to which government credit
programs affected consumption, and that it is useful to look at this
question from the perspective of what happened to consumption over the
relevant period. He argued that, during the period from about 2001 to
the beginning of 2007, consumers played a Ponzi game. That is,
consumption was sufficiently high that it was financed in part by
continuous borrowing. Or to put it differently, borrowing was greater
than the amount of interest paid on outstanding household debt. Then,
starting in late 2008, 2009, and 2010, in particular, an enormous
squeeze took place on consumption. Paradoxically, he noted, this was
seen as a big increase in household saving in the national income
accounts, despite there being strong evidence of a credit squeeze on the
household. The squeeze would have been a lot worse if it had not been
for the fact that people were able to use government lending programs to
some extent to offset the squeeze that was taking place on the
household, and that reduction in consumption would have been even
larger. The question is: How much larger? This question is basically one
of how much leakage there was.
Hall also spoke to the question of fiscal multipliers. It is widely
agreed upon in the literature that multipliers are substantially higher
at the zero lower bound. (This is because normally there is an interest
rate offset operating through the Taylor rule that offsets a spending
expansion. With a zero lower bound, this offset does not occur, implying
a high multiplier.) Hall disagreed with the view held by some that only
distinctively Keynesian models have multipliers. On the contrary, any
kind of macroeconomic model involving a government policy that affects
spending will result in some sort of fiscal multiplier. As a skeptic
about some of the features of Keynesian models, Hall still believes in
multipliers.
Justin Wolfers made three brief points. First, he argued that the
large amount of uncertainty surrounding the magnitudes of Lucas's
estimates was a feature of the model, not a bug. During his remarks,
discussant Bill Gale had noted that the net effect ranged from 0.7 to 4
percent of GDP. Wolfers encouraged everyone to be candid about how
little is really known about this topic. Second, he noted that it seems
quite straightforward that the view of the world outlined by Lucas has
very clear empirical implications about whose consumption should go up,
and whose did not. A household with student loans, for instance, would
clearly benefit more than a household with no student loans. Third and
finally, he expressed concern that Lucas's
"bang-for-the-buck" framework seemed nonobvious, and wondered
if she could do more to motivate the cost calculation.
David Romer, in relation to Wolfers's first point, argued for
actually making the range of possible estimates wider. According to the
paper, it seems that about one-half of the effects are coming from
housing loans, and about one-third are coming from student loans. He
wondered about the extent to which the multiplier channels for these
programs differed from those of conventional fiscal policy. For
instance, suppose that when someone takes out a student loan, he or she
goes to college instead of working. Does the fact that the person is not
working lead to a reduction in GDP? Does the person going to college
make it easier for someone else to get a job? What are the GDP effects
when the person graduates? And so on. Likewise, for housing, much of the
first round of spending presumably goes toward buying existing homes,
meaning that the funds that are being lent end up in the pockets of the
person who just sold the home, who is presumably no longer liquidity
constrained. The channels, therefore, again sound completely different
from normal multipliers. Romer admitted to not having a good handle on
what the precise issues were, but wondered if Lucas could do more to
expand upon them.
Romer also mentioned the possibility of unusual effects tied to the
nitty-gritty of particular programs. He referenced a recent paper by
Marco Di Maggio, Amir Kermani, and Christopher Palmer, in which the
authors discuss some unexpected effects of the refinancing provisions of
the Home Affordable Refinance Program. (1) Rather than cash-out
refinance, the authors found evidence of cash-in refinance: People were
scraping together money from various sources to get their loan-to-value
ratios down to 80 percent. It is therefore possible, Romer noted, that
in some cases, the government's helping people out by allowing them
to refinance actually reduced consumption (a negative multiplier). All
this is to say, he concluded, that some complicated government programs
make it very difficult to have much confidence in our ability to
estimate multipliers with any precision, and he congratulated Lucas for
at least trying something.
Related to Romer's comments, William Brainard encouraged Lucas
to spend more time addressing housing specifically. When housing prices
crashed, with prices in many places well below replacement cost, there
was very little reason for new house construction. Individuals were
likely to use new loans either to refinance or buy an existing house,
and they were not likely to contract the construction of a new house;
nor was there likely to be much speculative building. Multipliers, he
concluded, might be very different when housing prices are depressed
than when housing prices are high.
Matthew Shapiro suggested that government's expansion of
credit, which requires borrowers to repay, might actually be less
effective in boosting economic activity than standard government
rebates, which are not required to be paid back. Multipliers reported by
the Congressional Budget Office typically assume a policy experiment in
which the economy crashes, the government sends out rebate checks, and
policymakers analyze consumers' marginal propensities to consume.
One could imagine that as a stimulus policy, instead of giving rebate
checks, the government gave out credit cards. It is not unreasonable to
assume a lower multiplier for the credit cards (which would have to be
paid back) than for the rebate checks (which would not have to be paid
back). What underlies the size of the multiplier is whether or not
people are liquidity constrained; that kind of thinking, Shapiro argued,
should support the notion of lower multipliers. Furthermore, if a
recession is accompanied by people being in a panic from having borrowed
too much, having policies that make it easier to borrow might actually
not be that effective because people are just not in a borrowing mode.
People who have the ability to borrow probably do not need help with
credit, and the people who are constrained might actually want to borrow
less. According to Shapiro, when auto sales were collapsing in November
2008, it was not that people wanted car loans but could not get them, it
was that they really did not want loans during that period. Freeing up
credit, therefore, probably would not have caused a huge stimulus.
Ricardo Reis wanted to push further on the analogy between the
sorts of credit policies outlined in the paper and monetary policy more
generally. In monetary policy, a credit subsidy extended to a bank
lowers the cost at which the bank funds itself in order to make a loan,
which is presumed to lead to more loans, stimulating economic activity.
The credit channel that Lucas isolates in the paper reminded Reis of the
bank lending channel in monetary policy. There has been a lot of work in
the last few years trying to establish by how much exactly credit
subsidies lead to an increase in lending. Whether or not the credit
policies in the paper qualify as monetary or fiscal, there does seem to
be a lot of overlap between the bank lending channel of monetary policy.
Perhaps, he suggested, the fairly reliable estimates produced in this
literature could help inform Lucas's analysis.
Where the literature has struggled, Reis noted, is in linking
increased lending to real economic activity. That is, how much does the
extra $1 lent lead to an increase in output? What the literature has
found--and Reis reiterated that the estimates are not very accurate--is
that most multipliers are not way above 1. That is, going from lending
into output does not look like a transfer multiplier or a purchase
multiplier, or at least the numbers do not seem to quite match up. Reis
also noted that what is different about credit subsidies relative to the
bank channel of monetary policy is that the credit subsidies are much
more targeted, whereas monetary policy would lower the cost of lending
of all kinds of loans. It is certainly plausible, and potentially even
likely, that because credit subsidies are targeted they would indeed
have much stronger effects.
Martin Baily said the paper did not fully capture the extent to
which the 2008 financial crisis caused a semicollapse of the
infrastructure of lending, particularly mortgage lending. Many of the
banks were capital constrained, with regulators breathing down their
necks telling them to reduce risk and tighten lending standards. Many
households did not want to buy a house because housing prices had been
declining. The government took over Fannie Mae and Freddie Mac and had
them continue issuing mortgage loans and, more important, continue
refinancing existing loans. Securitization had collapsed; nobody wanted
to be in the business of securitization, and there was a belief that it
would take some time to repair private credit markets and allow them to
operate normally. Credit policies at that time, he suggested, were
playing a different role than credit subsidies in normal times.
Phillip Swagel agreed with Reis that the ability to target credit
subsidies was vital during the crisis because, as Baily had noted, it
would have been very difficult to repair the credit markets and allow
them to operate normally. In housing, for example, private label
securitization had ended by late 2008, and banks were not eager to do
balance sheet lending; the rescue of Fannie Mae and Freddie Mac in
September 2008 involving targeted credit subsidies for housing was
essential to supporting continued economic activity even as other parts
of credit markets locked up. Institutional arrangements in other areas
made it similarly important to be able to target credit subsidies such
as credit for auto dealers--without these so-called floor plan loans,
the institutional arrangements in the auto retail sector would have had
a seriously negative impact on auto sales. This was especially the case
in late 2008 and early 2009, when credit markets were particularly
strained.
In the specific case of General Motors, there were institutional
arrangements that Swagel believed turned out to be very important. He
described that when someone buys a car from a dealer, the dealer
finances the car, which is why he wants it off the lot as quickly as
possible. The type of floor plan loan, according to Swagel, was not
initially included in the Federal Reserve's Term Asset-Backed
Securities Loan Facility (TALF) program. Karen Pence, an adviser at the
Board of Governors of the Federal Reserve System, interjected
enthusiastically, "That's not right!" Moderator Janice
Eberly, not wanting to hold her back, let Pence continue. As someone who
worked on the TALF program, Pence assured everyone that they were made
very aware that the floor plan financing was a big problem. Rather, the
issue was that the automakers could not achieve AAA ratings for their
floor plans, and the Federal Reserve was not allowed to take anything
that was not rated AAA. She added that the Federal Reserve actually
worked very well with staff at the Treasury Department to try to find a
way around that.
Wendy Edelberg, representing the Congressional Budget Office,
wanted to clarify the apparent conflation of two different ways that
multipliers can vary over the business cycle. The multipliers used in
Lucas's paper represented the total aggregate fiscal multiplier,
the full range that the Congressional Budget Office used when the
economy was at the zero lower bound. This total aggregate fiscal
multiplier, she explained, is actually a combination of two things: the
impulse to aggregate demand, coupled with the way in which it churns
through the economy using a demand multiplier. Standard practice at the
Congressional Budget Office is to use the same demand multiplier for any
given impulse to aggregate demand. The consensus in the room seemed to
be that both components of the total multiplier--the impulse and the
churning--could vary over the business cycle. Clearly, the churning can
vary over the business cycle, she noted, due to the response of monetary
policy. The consensus seemed to also be that the impulse to aggregate
demand could vary, for instance, because of how liquidity constrained
borrowers are, points raised earlier by Romer and Shapiro. Nonetheless,
Edelberg wanted to stress the point that it seems useful to separate the
two components of the total aggregate fiscal multiplier.
Karen Dynan noted that, given the importance of housing in
Lucas's results, past results are no guarantee of future
performance. One important channel through which the housing stimulus
came in the recent recovery was people being able to refinance,
unlocking cash through lower monthly payments. A lot of that
refinancing, she added, could not have been done without the
complementary programs that allowed people with underwater homes to
refinance. Dynan noted that the next time this sort of housing crisis
occurs, there will probably be a different housing finance regime in
place, so it is worth thinking about the points made in this paper when
talking about how to design that system.
On the uncertainty surrounding the estimates of the multipliers
brought up by Gale, Wolfers, and others, Lucas disagreed with
Gale's implication that there is no evidence to say whether the
numbers are right or wrong. She also suggested that, as a profession,
economists should pay attention to things where the mean effect is known
to be very large, despite there being a high degree of uncertainty. She
stressed that it is very important to talk about things that are hard to
quantify when they are likely to be large.
On the question of her bang-for-the-buck calculation, Lucas agreed
with remarks made by Gale that the calculation is not marginal; nor is
the same multiplier likely to apply to the next episode, as Dynan had
suggested. She only meant to suggest that this is a potentially powerful
mechanism, and large in this instance. She emphasized that the Great
Recession was a particularly severe financial crisis, and the next
recession and past recessions might not look the same.
On the discussion of whether or not credit policy was analogous in
some ways to monetary policy, a point pushed by Reis, Lucas sought to
clarify her point of view. She outlined that monetary policy helped
credit markets by providing what the Federal Reserve calls "greater
liquidity." The distinction she sought to make between liquidity
and subsidy follows the party line of the Federal Reserve, namely, that
providing liquidity does not require taking on credit risk, the
canonical example being that when a bank comes to the Federal
Reserve's discount window, there is plenty of collateral, and so
there is very little credit risk involved. The bottom line is, although
the Federal Reserve did provide small credit subsidies--a point Lucas
conceded members of the Federal Reserve might not agree with her on--the
Troubled Asset Relief Program basically covered the riskier of the
Federal Reserve's programs. She was unwilling to concede that the
credit policies outlined in her paper resembled anything close to
traditional monetary policy.
On the question of whether her model was literally applicable to
the situation in 2010, Lucas was sympathetic with the view that the
story might not be so simple, but she still thought that credit
subsidies and the channels identified in the model were the essence of
why credit programs create a fiscal stimulus. There were all kinds of
complicated reasons for credit markets being disrupted, but at the heart
of it, the government for the most part extended credit opportunities
categorically to otherwise-constrained borrowers--for example, all
students are allowed to take out student loans, regardless of their
circumstances--which is a big difference from the way that private
markets operate.
Lucas joked that she was saddened by having to defend the concept
of fiscal multipliers. She appreciated Hall's sympathy, and his
unbridled defense of multipliers. She had considered the issues raised,
the range of the multipliers, and the choices about how much the
movement of extensive margins reflected those considerations. Lucas
defended her choices of relatively small extensive margins for most of
the housing programs. Some other papers, she noted, have argued that
Fannie Mae and Freddie Mac do not provide much incremental credit during
good times. In contrast, during the 2008 financial crisis, Lucas
believes they did provide incremental credit that increased aggregate
demand. For student loans, she argued for a high incremental effect,
since there is a general understanding that most students are heavily
constrained.
Specifically on the issues relating to housing--brought up by no
less than Romer, Brainard, Baily, and Dynan--Lucas noted that she had
put a lot of effort into making reasonable assumptions. She was well
aware of the fact that a lot of the credit subsidies do go toward
refinancing. The cash-out refinancing, mentioned by Dynan, is one
mechanism for mortgages to create a stimulus. Purchase loans, even if
they are used to buy existing structures, also create a stimulus because
when people buy new houses they spend a lot of money on other things,
such as new furniture and appliances. However, Lucas's assumptions
of relatively small multipliers for housing programs reflect the fact
that much of the money does go to purchasing existing structures or to
replace existing debt.
Shapiro had brought up the question of how much it mattered that
loans, as opposed to rebates, had to be repaid. Lucas noted that a lot
of government loans are fairly long term. Government loans are typically
not analogous to a credit card, where one has to pay back the balance
relatively quickly. Student loans, for instance, have a horizon of 20 to
30 years; likewise, mortgages can range anywhere from 15 to 30 years.
The fact that one has to repay the loans eventually, Lucas believes, is
not that important.
Others, notably Romer and Brainard, had argued that credit
multipliers would tend to be smaller than traditional multipliers. Lucas
disagreed, and argued that credit multipliers might actually be bigger
than regular multipliers. If someone receives a tax rebate, for
instance, that person may actually save a portion of it. A loan,
conversely, is very costly to take out, so it is much more likely that
the borrower will spend all that money. The only issue is what the money
is spent on and whether what it is spent on is going to have a big
multiplier effect or not. Therefore, one should not think of credit as
naturally having a smaller multiplier effect than other spending.
(1.) Marco Di Maggio, Amir Kermani, and Christopher Palmer,
"Unconventional Monetary Policy and the Allocation of Credit,"
Columbia Business School Research Paper no. 16-1 (March 2016).
DEBORAH LUCAS
Massachusetts Institute of Technology
(1.) Elliott (2011) provides a history and more complete discussion
of federal credit programs.
(2.) Related theoretical analyses include those by Jaffee and
Russell (1976), Stiglitz and Weiss (1981), Smith (1983), Smith and
Stutzer (1989), Gale (1990), Lacker (1994), and Williamson (1994).
(3.) This assumption economizes on notation and is without a loss
of generality.
(4.) A related problem for the government arises when it sets a
uniform interest rate that is above the market rate for identifiably
safe borrowers, and those borrowers are picked off by private lenders
that can profitably underprice the government, as has happened at
certain times with federal student loans.
(5.) In a frictionless market, a subsidized borrower could turn
around and sell the loan for an amount equal to the credit subsidy, but
in practice it is prohibitively costly for borrowers to monetize the
subsidy.
(6.) This excludes programs classified as emergency lending
associated with the financial crisis.
(7.) See Barnett and Vidal (2012) and Financial Management Service
(2011).
(8.) The online appendixes for this and all other papers in this
volume may be found at the Brookings Papers web page,
www.brookings.edu/bpea, under "Past Editions."
(9.) In the interest of full disclosure, I was a coauthor or
reviewer of all the CBO studies referenced except for the ones on fiscal
multipliers and ARRA stimulus.
(10.) The volumes that are relevant to the subsidy calculations
include refinanced loans even if the previous mortgage also carried a
federal guarantee. This is because the models used to predict guarantee
costs treat refinancing events as precluding further defaults.
(11.) The valuation exercise employed a Monte Carlo model of
mortgage cash flows, together with the prices ot PMI guarantees, GSE
guarantees, and of mortgage-backed securities, to infer risk-neutral
prices that could then be used to value FHA guarantees.
(12.) All FCRA estimates are from the OMB (2011).
(13.) During that time the government also purchased guaranteed
loans from lenders. Those purchases do not create new credit subsidies
for borrowers and those loans are excluded from the reported subsidy
estimates.
(14.) The methodology was similar in both analyses, but the
reported subsidy rates are somewhat different, in part because of the
time periods considered. However, the subsidy rates reported by the CBO
(2010d) are more applicable to this analysis because they take into
account the mix of loan types and interest rate conditions in 2010.
(15.) A complete description of the valuation model is given by
Lucas and Moore (2010).
(16.) The estimate is referred to in the report as a market-value
estimate, but it is conceptually equivalent to what is described as a
fair-value estimate in later CBO publications.
(17.) The data, which are reported by the federal agency running
the program, are of mixed quality and in some cases are clearly
incorrect. In these cases, the risk charge added to the FCRA subsidy
rate is the average of the risk charges across all the smaller programs.
(18.) The present value of fees is assumed to be unaffected by the
discount rate, which is only correct for upfront fees. However, the data
on periodic fees are not reliable, which is why the effect of
differential risk adjustment for fees is not calculated. Neglecting the
difference is likely to have a very small effect on the total subsidy
estimates.
(19.) The CBO (2011c) provides a detailed example of this approach
for nuclear construction loan guarantees.
(20.) The risk charge plus the OMB's FCRA estimate is used
instead of the rough fair-value estimates because the FCRA estimates are
generally based on more complete information about cash flows and their
timing.
(21.) Jeske, Krueger, and Mitman (2013) are an exception; they
propose a structural model to assess the effects of the GSEs.
(22.) Indirect support for this assumption comes from the finding
that GSE pricing is only slightly more favorable than that on comparable
private-label mortgages (Passmore, Sherlund, and Burgess 2005).
(23.) To the extent that this may overstate the extent to which FHA
borrowers would be constrained during a crisis, note that FHA lending is
shown below to add only $67 billion in stimulus, and that shading that
number down would not change the main conclusions.
(24.) For an estimate of the effects of GSE and FHA program rules
on refinancing activity during this period, also see Remy, Lucas, and
Moore (2011).
(25.) De Andrade and Lucas (2009) find that SBA subsidies may
benefit banks more than small businesses. However, Bach (2014) provides
evidence that the effects of a French targeted credit program
significantly reduced credit constraints.
(26.) In the Annual Report, see table 8 on p. 66
(http://www.freddiemac.com/investors/er/ pdf/10k_022411.pdf).
(27.) By contrast, a total subsidy cost of--$11.7 billion (that is,
savings) was reported in the federal budget.
Table 1. Ranges for U.S. Fiscal Multipliers
Estimated multiplier
Type of activity Low estimate High estimate
Purchases of goods and services by 0.5 2.5
the federal government
Transfer payments to state and local 0.4 2.2
governments for infrastructure
Transfer payments to state and local
governments for other purposes 0.4 1.8
Transfer payments to individuals 0.4 2.1
One-time payments to retirees 0.2 1.0
Two-year tax cuts for lower- and 0.3 1.5
middle-income people
One-year tax cut for higher-income 0.1 0.6
people
Extension of the first-time homebuyer 0.2 0.8
credit
Corporate tax provisions primarily 0 0.4
affecting cash flow
Source: CBO (2011b).
Table 2. Summary of Fair-Value Subsidy
Estimates for Federally Assisted Credit, 2010
Fair-value
Fair-value subsidy
Loan volume subsidy value
(billions rate (billions
Category Agency of dollars) (percent) of dollars)
Housing Federal Housing
Administration 319 2.5 8.0
Housing Department of
Veterans
Affairs 63 3.2 2.0
Housing Rural Housing
Service 17 4.4 0.7
Student loans Department of
(guaranteed) Education 20 16.0 3.1
Student loans Department of
(direct) Education 85 13.0 11.0
Business Small Business
Administration 17 6.5 1.1
Other Various 64 6.0 3.8
traditional
Subtotal (a) 585 29.7
Housing Fannie Mae and
Freddie Mac 1,011 4.1 40.9
Total 1,596 70.6
Sources: Author's calculations; OMB (2011).
(a.) The sum of disbursements is lower than the total in the
OMB's "Analytical Perspectives" because the Treasury's TARP and
mortgage-backed securities transactions and the Department of
Education's purchases of seasoned student loans are excluded.
Table 3. Calculation of Incremental Borrowing
along the Intensive Margin, 2010
Loan
volume
(billions
Category Agency of dollars) Elasticity
Housing Federal Housing
Administration 319 1.8
Housing Department
of Veterans
Affairs and
Rural Housing
Service 80 1.8
Student loans Department of
Education 105 0.65
Business Small Business
Administration 17 0.8
Other Various 64 0.8
traditional
Subtotal (a) 585
Housing Fannie Mae and
Freddie Mac 1,011 1.8
Total 1,596
Incremental
Subsidy borrowing
rate (billions of
Category Agency (percent) dollars)
Housing Federal Housing
Administration 2.5 14.3
Housing Department
of Veterans
Affairs and
Rural Housing
Service 3.5 5.0
Student loans Department of
Education 14.0 9.6
Business Small Business
Administration 6.2 0.8
Other Various 6.0 3.1
traditional
Subtotal (a) 32.8
Housing Fannie Mae and
Freddie Mac 4.1 74.6
Total 107.4
Sources: Author's calculations; OMB (2011).
(a.) Loan volume is lower than the total disbursements in the
OMB's "Analytical Perspectives" because the Treasury's TARP and
mortgage-backed securities transactions and the Department of
Education's purchases of seasoned student loans are excluded.
Table 4. Incremental Output in a Normal Period
2010 loan
volume
(billions of Constrained
Category Agency dollars) share
Housing Federal Housing
Administration 319 0.10
Housing Department
of Veterans
Affairs and
Rural Housing
Service 80 0.10
Student loans Department of
Education 105 0.75
Business Small Business
Administration 17 0.75
Other traditional Various 64 0.50
Subtotal (a) 584
Housing Fannie Mae and
Freddie Mac 1,011 0.00
Total 1,595
Incremental Incremental
loan volume loan volume
along along
extensive intensive
margin in margin in 2010
normal times (billions of
(billions of dollars)
Category Agency dollars)
Housing Federal Housing
Administration 31.9 14.3
Housing Department
of Veterans
Affairs and
Rural Housing
Service 8.0 5.0
Student loans Department of
Education 78.8 9.6
Business Small Business
Administration 12.5 0.8
Other traditional Various 31.9 3.1
Subtotal (a) 163 33
Housing Fannie Mae and
Freddie Mac 0.0 74.6
Total 163 107
Incremental
output
Multiplier in (billions of
normal times dollars)
Category Agency
Housing Federal Housing
Administration 0.3 13.9
Housing Department
of Veterans
Affairs and
Rural Housing
Service 0.3 3.9
Student loans Department of
Education 0.5 44.2
Business Small Business
Administration 0.5 6.6
Other traditional Various 0.5 17.5
Subtotal (a) 86
Housing Fannie Mae and
Freddie Mac 0.2 15
Total 101
Sources: Author's calculations; OMB (2011).
(a.) Loan volume is lower than the total disbursements in the
OMB's "Analytical Perspectives" because the Treasury's TARP and
mortgage-backed securities transactions and the Department of
Education's purchases of seasoned student loans are excluded.
Table 5. Incremental Output in a Distressed Period
2010 loan
volume
(billions of Constrained
Category Agency dollars) share
Housing Federal Housing
Administration 319 0.90
Housing Department
of Veterans
Affairs and
Rural Housing
Service 80 0.50
Student loans Department of
Education 105 0.95
Business Small Business
Administration 17 0.85
Other traditional Various 64 0.75
Subtotal (a) 584
Housing Fannie Mae and
Freddie Mac 1,011 0.25
Total 1,595
Incremental
loan volume
along Incremental
extensive loan volume
margin in along
distressed intensive
times margin in 2010
(billions of (billions of
Category Agency dollars) dollars)
Housing Federal Housing
Administration 286.8 14.3
Housing Department
of Veterans
Affairs and
Rural Housing
Service 40.0 5.0
Student loans Department of
Education 99.8 9.6
Business Small Business
Administration 14.1 0.8
Other traditional Various 47.8 3.1
Subtotal (a) 488 33
Housing Fannie Mae and
Freddie Mac 252.8 74.6
Total 741 107
Incremental
Multiplier output
in distressed (billions of
Category Agency times dollars)
Housing Federal Housing
Administration 0.4 120.5
Housing Department
of Veterans
Affairs and
Rural Housing
Service 0.4 18.0
Student loans Department of
Education 2.0 218.6
Business Small Business
Administration 2.0 29.9
Other traditional Various 2.0 101.7
Subtotal (a) 488.6
Housing Fannie Mae and
Freddie Mac 0.3 98
Total 587
Sources: Author's calculations; OMB (2011).
(a.) Loan volume is lower than the total disbursements in the
OMB's "Analytical Perspectives" because the Treasury's TARP and
mortgage-backed securities transactions and the Department of
Education's purchases of seasoned student loans are excluded.
