PRELIMINARY DISCUSSION DRAFT September 11–12, 2018
13
payments) and find a smaller multiplier of 0.5 to 1.0, although this reflects
heterogeneous estimates including a multiplier of -0.7 to -3.3 for education and police
spending but much higher multipliers like 2 to 2.3 for low-income support and 1.8 for
infrastructure and other grants. Dube, Kaplan, and Zipperer (2015) also use county-level
stimulus spending, finding slightly larger effects than Feyrer and Sacerdote (2011).
Translating state-level multipliers into national multipliers is not straightforward
because of many factors, most notably the spillovers from one state to another and other
general equilibrium effects (Nakamura and Steinsson 2014 and Farhi and Werning 2016).
Chodorow-Reich (forthcoming), however, argues that in many cases the policy-relevant
multiplier should assume no interest rate response and deficit financing, in which case
he shows that state-level cross-sectional multipliers are a rough lower bound on what the
relevant Keynesian national multiplier would be. He summarizes his analysis as finding
that his own results and other cross-sectional results “implies a no-monetary-policy-
response deficit-financed national multiplier of about 1.7 or above. This magnitude falls
at the very upper end of the range found in a recent review article based mostly on time
series evidence (Ramey 2011). Thus, cross-sectional multiplier studies suggest the
national multiplier can be larger than often assumed.”
A few other provisions of the stimulus have also been studied, although most of
these estimates do not meaningfully change the ex-ante parameters originally assumed
when the stimulus passed.
7
The one case where the evidence suggests a substantially
larger effect than originally assumed is for bonus depreciation. The CEA estimates
assumed a multiplier of 0.1 for business tax provisions, similar to CBO’s low estimates.
Zwick and Mahon (2017), however, studied the actual data and found large effects of
temporarily expanded equipment depreciation on investment, particularly for small
and financially constrained firms. This finding is consistent with some of the thinking
that went into designing these provisions in the first place, which were intended in part
as an interest free loan whose benefit would be related to the cash flow in the initial
years but whose cost was the much lower present value to the government. The original
bonus depreciation, for example, provided $50 billion of tax reductions in the first two
fiscal years as firms shifted depreciation allowances earlier but then most of this
money was recouped in future years, leaving the 10-year cost at only $7 billion. If only
10 percent of the $50 billion was spent as additional investment that would lead to a
multiplier of nearly 1 evaluated against the entire fiscal cost of the measure.
8
L
"M.)"FGG\"%)N$2)"$*'"V$S5*R"@/%S"1$+"=)%)"326'5)'"N+"1$%S)% ")2"$0:" CFG!KD"=./"0//S)'"$2"2.)"59?$-2"/*"
-/*369?25/*"$*'",$.9J",.$?5%/J"$*'",0)9%/'"CFG!FD"=./"%)05)'"/*"3)07I%)?/%2)'"36%&)+3J"=52."2.)"7/%9)%"
75*'5*R"0$%R)%"59?$-23"2.$*"2.)"0$2)%:"M.)"59?$-2"/7"2.)"-$3."7/%"-06*S)%3"?%/R%$9"53"')N$2)'"=52."V5$*"$*'"
,675"CFG!FD"$*'"#/)S32%$J"1600)%"$*'"@)32"CFG!LD"75*'5*R"39$00"2/"*)R$25&)")77)-23"=.50)"O%))*J"V)0`)%J"
1$%S)%"$*'"P/W$3"CFG!\D"75*'5*R"$"0$%R)%")77)-2a$02./6R."2.)"% )3 /0625/*"/7"2.53"')N$2)".$3"05220)")77)-2"/*"2.)"
)UI$*2)"%)36023"%)?/%2)'")$%05)%"N)-$63)"-$3.I7/%I-06*S)%3"=$3"/*0+"YK"N5005/*"/7"2.)"/&)%$00"Y!:Z"2%5005/*"
32596063:"_2.)%")&$06$25/*3"/7"3?)-575-"?%/R%$93"5*-06')"^)%R)%J"M6%*)%J"$*'"b=5-S"CFG!\D"75*'5*R"0$%R)"
)77)-23"/7"2.)"<5%32I259)"#/9)N6+)%3"M$U"B%)'52"$*'"B./'/%/=IP)5-.J"B/R05$*)3)J"$*'"($%$N$%N/6*53"
C7/%2.-/95*RD"75*'5*R"39$00)%"0$N/%"9$%S)2")77)-23"/7"2.)"6*)9?0/+9)*2"5*36%$*-)")U2)*35/*3:"
\
"M.)")U?$*35/*"/7"A)2"_?)%$25*R"Q/33)3"CA_Q3D"=$3"$03/"N$3)'"/*"$"35950$%"0/R5-J"N62">/N%5'R)"CFG![D"
75*'3"2.$2"9/3 2"/7"2. )3) "$' '525/ *$ 0"?$ +9 ) *23 "=) %)"3$ &) '"N +"75%93"%$ 2.) %"2.$ *"3? ) *2a$02./6R."2.53"'/)3"
*/2"%60)"/62"$''525/*$0"5*&)329)*2"5*"7626%)"+)$%3:"