The Long-Run Effects of Monetary Policy

The long-run effects of monetary policy ?
Òscar Jordà † Sanjay R. Singh ‡ Alan M. Taylor §
January 2020
Abstract
Is the effect of monetary policy on the productive capacity of the economy long lived?
Yes, in fact we find such impacts are significant and last for over a decade based on:
(1) merged data from two new international historical databases; (2) identification of
exogenous monetary policy using the macroeconomic trilemma; and (3) improved
econometric methods. Notably, the capital stock and total factor productivity (TFP)
exhibit hysteresis, but labor does not. Money is non-neutral for a much longer period
of time than is customarily assumed. A New Keynesian model with endogenous TFP
growth can reconcile all these empirical observations.
JEL classification codes: E01, E30, E32, E44, E47, E51, F33, F42, F44.
Keywords: monetary policy, money neutrality, hysteresis, trilemma, instrumental vari-

ables, local projections.
? We are thankful to Gadi Barlevy, Susanto Basu, James Cloyne, Stéphane Dupraz, John Fernald, Jordi Galı́,
Yuriy Gorodnichenko, Pierre-Olivier Gourinchas, Valerie Ramey, Ina Simonovska, Andrea Tambalotti, and
many seminar and conference participants at the Barcelona Summer Forum, the NBER Summer Institute IFM
Program Meeting, the SED Annual Meeting, the Federal Reserve Banks of Richmond, San Francisco, St. Louis,
and Board of Governors, the Midwest Macro Conference, Claremont McKenna College, the Swiss National
Bank, the Norges Bank, UC Davis, University of Wisconsin-Madison, Universität Zürich, and Vanderbilt
University, who provided very helpful comments and suggestions. Antonin Bergeaud graciously shared
detailed data from the long-term productivity database created with Gilbert Cette and Rémy Lecat at the
Banque de France. All errors are ours. The views expressed herein are solely those of the authors and do not
necessarily represent the views of the Federal Reserve Bank of San Francisco or the Federal Reserve System.
† Federal Reserve Bank of San Francisco and Department of Economics, University of California, Davis
([email protected]; [email protected]).
‡ Department of Economics, University of California, Davis ([email protected]).
§ Department of Economics and Graduate School of Management, University of California, Davis; NBER;
and CEPR ([email protected]).
Electronic copy available at: https://ssrn.com/abstract=3529773

Are there circumstances in which changes in aggregate demand can have an
appreciable, persistent effect on aggregate supply?
— Yellen (2016)
1. Introduction
What is the effect of monetary policy on the long-run productive capacity of the economy?
At least since Hume (1752), macroeconomics has largely operated under the assumption
that money is neutral in the long-run, and a vast literature spanning centuries has gradually
built the case (see, e.g., King and Watson, 1997, for a review). Lately, however, many have
begun to question this monetary canon. Barro (2013), for example, provides evidence that
high levels of inflation result in a loss in the rate of economic growth. Work by Gopinath,
Kalemli-Özcan, Karabarbounis, and Villegas-Sanchez (2017) links interest rates to the level
of productivity, whereas more recently, Benigno and Fornaro (2018) link low interest rates
with the rate of growth of productivity.
We argue that any decisive investigation of monetary neutrality must rest on three
pillars. First, we are looking at long-run outcomes and so we need a sample based on longer
time series data and, if possible, on a wide panel of countries to obtain more statistical
power. Second, it is essential to identify exogenous movements in the interest rate to obtain
a reliable measure of monetary effects and to avoid confounding. Third, as we will show
below, the empirical method used can make a big difference. Common approaches are
designed to maximize short-horizon fit. It is therefore important to use methods that are
instead better suited to characterize behavior over longer spans of time. We discuss how
we build on each of these three pillars next.
Starting first with the data pillar, we rely on two new macro-history databases spanning
125 years and 17 advanced economies. First, we use the data in Jordà, Schularick, and
Taylor (2017), available at www.macrohistory.net/data. This database contains major
macroeconomic aggregates, such as output, interest rates, credit, and so on. Second, in
order to decompose output into its components, we bring in the data from Bergeaud,
Cette, and Lecat (2016) and available at http://www.longtermproductivity.com.1 Their
data include observations on investment in machines and buildings, number of employees,
and hours worked. Using these variables, we can then construct measures of total factor
productivity (TFP), as we show later.
With regard to identification, the second pillar, we exploit the trilemma of international
finance (see, e.g., Obstfeld, Shambaugh, and Taylor, 2004, 2005; Shambaugh, 2004). The
1 We are particularly thankful to Antonin Bergeaud for sharing some of the disaggregated series from
their database that we used to construct our own series of adjusted TFP.

basic idea is that when a country pegs its currency to some base currency, but allows free
movement of capital across its borders, it effectively loses full control over its own domestic
interest rate: a correlation in home and base interest rates is induced, which is one when
the peg is hard and mobility is perfect, but generally less than one otherwise. Suitably
constructed shocks to the base interest rate are then exogeous to the home country (Jordà,
Schularick, and Taylor, 2016, 2019). Of course, this mechanism depends on a variety of
other factors that we examine through the lens of an extended, stylized Mundell-Fleming
framework (see, e.g., Gourinchas, 2018). This extension allows for contagion via financial
linkages, deviations from uncovered interest rate parity, and partial control over monetary
policy through small exchange rate interventions. Using this framework, we examine a
rich array of spillover mechanisms that could threaten the exclusion restriction typical of
instrumental variable assumptions. By making appropriate adjustments based on these
economic mechanisms and novel econometric methods, we show that our results are robust
to potential violations of the exclusion restriction (Conley, Hansen, and Rossi, 2012; van
Kippersluis and Rietveld, 2018).
The third and final pillar of our analysis has to do with the econometric approach. We
use local projections (Jordà, 2005) in order to get more accurate estimates of the impulse
response function (IRF) at longer horizons. As we show formally, as long as the truncation
lag in local projections is chosen to grow with the sample size (at a particular rate that we
make specific below), local projections estimate the impulse response consistently at any
horizon. Other procedures commonly used to estimate impulse responses do not have this
property (see, e.g., Lewis and Reinsel, 1985; Kuersteiner, 2005).
Supported by these three pillars, we show that, surprisingly, monetary policy affects TFP,
capital accumulation, and the productive capacity of the economy for a very long time. In
response to an exogenous monetary shock, output declines and even twelve years out it has
not returned to its pre-shock trend. Next, we investigate the source of this hysteresis and
find that capital and TFP experience similar trajectories to output. In contrast, total hours
worked (both hours per worker and number of workers) return quickly to the original
trend. This finding is distinct from the usual labor hysteresis mechanism emphasized in
the literature (see, e.g., Blanchard and Summers, 1986; Galı́, 2015a; Blanchard, 2018). These
results have important implications for how we think about standard models of monetary
economies.
Before providing a framework to make sense of these results, we first discuss the
robustness of our findings. In particular, there are obvious differences in the monetary
arrangements before and after WW2. Think of the pre-WW1 Gold Standard versus the
Bretton Woods era, for example. Moreover, the Global Financial Crisis, second only to the

Great Depression, could be affecting our findings unduly. Thus we examine several sample
splits of the data. Fernald (2007), Bergeaud, Cette, and Lecat (2016) and Gordon (2016)
document trend breaks in TFP for a variety of countries in our sample. These breaks could
affect our estimates and therefore we re-estimate our models using breaks determined with
the structural break testing procedure by Bai and Perron (1998).
In any instrumental variable analysis, it is important to assess the exclusion restriction.
Absent overidentifying restrictions to formally test the validity of the instrument, we resort
to a variety of alternative approaches. First, we allow for spillovers motivated by our
Mundell-Fleming framework and captured by fluctuations in base-country as well as global
GDP, current account balances, and nominal exchange rates. In particular, we control for
past, current, and future values of these variables. Including future values of variables is
unusual, but is meant to stack the odds against our findings. We also account for unknown
spillovers, following Conley, Hansen, and Rossi (2012) and van Kippersluis and Rietveld
(2018), by exploiting different subpopulations present in our sample.
There have been numerous influential papers written based on post-WW2 U.S. data,
which naturally form a useful benchmark for our analysis (see, e.g., Ramey, 2016; Nakamura
and Steinsson, 2018). In particular, we examine whether monetary shocks identified as in
Romer and Romer (2004) produce the same results when applied to the data in Fernald
(2014) on utilization adjusted TFP, capital, and total hours worked. We find similar results to
those based on our sample and identification approach, with some noteworthy differences
that we shall discuss. This exhaustive list of robustness checks give us confidence in our
findings.
In order to make economic sense of our findings, we extend a workhorse quantitative
macroeconomic model (Christiano, Eichenbaum, and Evans, 2005; Smets and Wouters, 2007)
with endogenous hysteresis effects. We choose a parametrically convenient process for
endogenous TFP growth that borrows from the formulation in Stadler (1990). We estimate
the hysteresis elasticity directly from the cumulative responses of TFP and output using a
two-step, classical minimum distance procedure. Our estimated elasticity is similar to that
assumed in DeLong and Summers (2012) and comfortably falls within a tight probability
interval.
In our model, a contractionary monetary policy shock lowers output temporarily
producing a slowdown in TFP growth. Under a standard Taylor rule, this slowdown in TFP
growth accumulates to yield permanently lower trend levels of output and capital, while
labor returns to the stationary equilibrium quickly. The model can generate empirically
consistent medium-run effects on GDP while maintaining other conventional textbook
results in monetary economics (as in, e.g., Galı́, 2015b).

These hysteresis effects naturally matter for how we build models of monetary economies
and design optimal monetary policy. A recent literature provides micro-foundations for
endogenous hysteresis effects via TFP growth (Fatas, 2000; Barlevy, 2004; Anzoategui,
Comin, Gertler, and Martinez, 2019; Bianchi, Kung, and Morales, 2019). Our findings and
framework incorporate a natural reduced-form representation of these mechanisms.2
Whilst we confirm the reduced-form implications of these models, a fuller investigation
into the empirical basis for appropriate micro-foundations would require an entirely new
paper, and most likely additional micro evidence, that we must leave for future research.
For example, the design of an optimal rule would depend on the type of frictions assumed.
In an environment where the only friction comes from nominal rigidities, a strict inflation-
targeting rule, for example, could deliver optimal allocations. As an illustration of how one
might modify the policy rule in our environment, we show that a hysteresis-augmented
policy rule generates a milder dip and quicker recovery in output at a modest cost to
inflation stabilization.
Our paper is related to the seminal work by Cerra and Saxena (2008) who documented
that output losses in the aftermath of economic and political crises in low-income and
emerging economies are highly persistent. More recently, researchers have also documented
similar effects following financial crises for advanced economies (see e.g., Reinhart and
Rogoff, 2009; Schularick and Taylor, 2012). Fatás and Summers (2018) document strong
hysteresis effects of fiscal consolidations for advanced economies following the global
financial crisis.
Over the much longer run, we document a causal effect of monetary policy shocks
in a historical panel data for advanced economies. There is mixed evidence of long-run
non-neutrality proposition in monetary economics as already mentioned above (King and
Watson, 1997). In the structural VAR literature, Bernanke and Mihov (1998) fail to reject
long-run neutrality in their identified responses to monetary policy shocks, which Mankiw
(2001) interprets as evidence of long-run non-neutrality due to the imprecision of the
estimates. More recently, Moran and Queraltó (2018) use three-equation VAR models
to emphasize an empirical link between TFP growth and monetary policy shocks. The
persistent effects of monetary policy bear implications for measurement of potential output
(see, e.g., Reifschneider, Wascher, and Wilcox, 2015; Coibion, Gorodnichenko, and Ulate,
2017). Relatedly, these echo the important early findings by Evans (1992) who cast doubt
on the exogeneity of TFP shocks, a key tenet of all mainstream real-business cycle and
dynamic stochastic general equilibrium (DSGE) models.
2 Foran alternate modeling of hysteresis, we refer the reader to Farmer and Nicolò (2018); Farmer and
Platonov (2019) for models of unemployment hysteresis.

2. Data and series construction
The empirical features motivating our analysis rest on two major international and historical
databases.
Data on macro aggregates and financial variables, including assumptions on exchange
rate regimes and capital controls, can be found in www.macrohistory.net/data. This
database covers 17 advanced economies reaching back to 1870 at annual frequency. Detailed
descriptions of the sources of the variables contained therein, their properties, and other
ancillary information are discussed in Jordà, Schularick, and Taylor (2017) and Jordà,
Schularick, and Taylor (2019), as well as references therein. Importantly, we will rely on the
trilemma instrument discussed in Jordà, Schularick, and Taylor (2016), and more recently
Jordà, Schularick, and Taylor (2019), as the source of exogenous variation in interest rates.
The instrument construction details will become clearer in the next section.
The second important source of data relies on the work by Bergeaud, Cette, and Lecat
(2016) and available at http://www.longtermproductivity.com. This historical database
adds to our main database observations on capital stock (machines and buildings), hours
worked, and number of employees, and the Solow residuals (raw TFP). In addition, we
construct time-varying capital and labor utilization corrected series using the procedure
discussed in Imbs (1999) with the raw data from Bergeaud, Cette, and Lecat (2016) to
construct our own series of utilization-adjusted TFP. We went back to the original sources
so as to filter out cyclical variation in input utilization rates in the context of a richer
production function that allows for factor hoarding. We explain the details of this correction
in the appendix in Section B.3
3. Identification
The trilemma of international finance gives a theoretically justified source of exogenous

variation in interest rates for open pegs, as Jordà, Schularick, and Taylor (2019) showed. The
logic is straightforward: under a peg, short-term rates in two countries will be arbitraged.
With strict interest parity, rates are exactly correlated and arbitrage is perfect, but this is not
necessary for our approach. Even with frictions or imperfect arbitrage, a non-zero interest
rate correlation is induced which is enough for identification purposes, as we show.
Formally, consider a setting with two countries: a large foreign economy that permits its
3 Our construction of productivity assumes misallocation related-wedges are absent. We have not yet
found the data to take into account markups in our productivity estimation. See Basu and Fernald (2002) and
Syverson (2011) for extensive discussions on what determines productivity.

exchange rate to float, and a smaller home economy that pegs its exchange rate to the large,
base economy. Under some conditions—to be made explicit momentarily—the interest rate
in the base provides a source of exogenous variation, i.e., an instrumental variable, for the
interest rate at home.
We divide the sample into three subpopulations of country-year observations. The bases
will refer to those economies whose currencies serve as the anchor for the subpopulation of
pegging economies, labeled as the pegs. Other economies, the floats, allow their exchange to
be freely determined by the market, and will not peg to any base.
Base and peg country codings can be found in Jordà, Schularick, and Taylor (2019, Table
1 and Appendix A), and are based on older, established definitions (Obstfeld, Shambaugh,
and Taylor, 2004, 2005; Shambaugh, 2004; Ilzetzki, Reinhart, and Rogoff, 2019). A country
j is defined to be in a peg at time t and denoted with the dummy variable q j,t = 1 if
it maintained a peg to its base at dates t − 1 and t. (This conservative definition serves
to eliminate the possibility of opportunistic pegging, and turns out that transitions from
floating to pegging and vice versa represent less than 5% of the sample, the average peg
lasting over 20 years. Interestingly, pegs are, on average, more open than floats.4 )
Based on this discussion, we construct the instrument as follows. Let ∆i j,t denote
changes in country j’s short-term nominal interest rate at date t, let ∆ib( j,t),t denote the
change in short term interest rate of country j’s base country b( j, t), and ∆i
b b( j,t),t denote the
predictable component explained by a variety of base country macroeconomic aggregates.
Loosely speaking, think of it as the rate that would be predicted by a policy rule.5 However,
since countries in a given year may not be perfectly open to capital flows, we then scale
the base shock, adjusting for capital mobility using the capital openness index of Quinn,
Schindler, and Toyoda (2011) and denoted k j,t ∈ [0, 1].
The resulting trilemma instrument is therefore defined as z j,t ≡ k j,t (∆ib( j,t),t − ∆i
b b( j,t),t ).
Identification thus works for the subpopulation with q j,t = 1, the pegs. But the subpopula-
tion with q j,t = 0, the floats, can also provide useful information to assess the exclusion
restriction using the methods described in Conley, Hansen, and Rossi (2012) and van
Kippersluis and Rietveld (2018), as we show below.
4 Inthe full sample, the index k averages 0.87 for pegs (with a standard deviation of 0.21) and 0.70 for floats
(with standard deviation 0.31). After WW2 there is essentially no difference between them. The average is
0.76 for pegs and 0.74 for floats with a standard deviation of 0.24 and 0.30 respectively. See Jordà, Schularick,
and Taylor (2019) for further details on the construction of the instrument.
5 The list of controls used to construct ∆i
b b( j,t),t include log real GDP; log real consumption per capita; log
real investment per capita; log consumer price index; short-term interest rate (usually a 3-month government
security); long-term interest rate (usually a 5-year government security); log real house prices; log real
stock prices; and the credit to GDP ratio. The variables enter in first differences except for interest rates.
Contemporaneous terms (except for the left-hand side variable) and two lags are included.

3.1. Justifying the instrument with an open economy model
Although the arbitrage mechanism behind the trilemma is easily grasped, in this section we
investigate the economic underpinnings of our identification strategy more formally with a
variant of the well known Mundell-Fleming-Dornbusch model. In particular, we incorporate
the extensions to the model discussed in Blanchard (2016) and Gourinchas (2018), which
embed various financial spillover mechanisms. Intuition is best communicated in a static
setting, which can be micro-founded with pre-fixed prices and wages as in Obstfeld and
Rogoff (1995). However, the appendix (see Section C) extends the model to a dynamic
setting with rational expectations as in Gali and Monacelli (2005) to provide more formal
justification.
Specifically, consider a framework made of two countries: a small domestic economy
and a large foreign economy, which we can call the United States, for now. Foreign (U.S.)
variables are denoted with an asterisk. Assume prices are fixed.
Given interest rates, the following equations describe the setup:
Y = A + NX ,
A = ξ − ci − f E ,
NX = a(Y ∗ − Y ) + bE ,
Y ∗ = A∗ = ξ ∗ − c∗ i ∗ ,
E = d(i∗ − i ) + gi∗ + χ ,
where a, b, c, c∗ , d, f , g, χ ≥ 0. Domestic output Y is equal to the sum of domestic absorption

A and net exports NX. Domestic absorption depends on an aggregate demand shifter
ξ, and negatively on the domestic (policy) nominal interest rate i. f denotes financial
spillovers through the exchange rate (e.g., balance sheet exposure of domestic producers in
a dollarized world).6 If f ≥ 0, then a depreciation of the exchange rate E hurts absorption.
Net exports depends positively on U.S. output Y ∗ , negatively on domestic output Y,
and positively on the exchange rate. U.S. output is determined in similar fashion except
that the U.S. is considered a large country, so it is treated as a closed economy. Finally, the
exchange rate depends on the difference between domestic and U.S. interest rates and on
a risk-premium shock. The term g is intended to capture risk-premium effects associated
with U.S. monetary policy.7
6 Jiang, Krishnamurthy, and Lustig (2019) provide a micro-foundation to generate these spillovers associ-
ated with the global financial cycle (Rey, 2015).
7 Itskhoki and Mukhin (2019) argue that such risk-premia violations of UIP are smaller under exchange
rate pegs, i.e., g is smaller.

In order to make the connection between the instrument as we defined it earlier and
this stylized model, we now think of ∆i∗ as the instrument z j,t ≡ k j,t (∆ib( j,t),t − ∆i
b b( j,t),t )
described earlier. The next section explores the benchmark setting of the trilemma to derive
the basic intuition.
The textbook specification with hard pegs
Under the assumption that f = g = χ = 0, many interesting channels are switched off and
the model just introduced reduces to the textbook Mundell-Fleming-Dornbusch version.
Consider what happens when the U.S. changes its interest rate, ∆R∗ . Since g = 0, to
maintain the peg it must be that ∆i = ∆i∗ . The one-to-one change in the home interest rate
has a direct effect on domestic absorption given by −c∆i.
However, notice that changes in the U.S. rate affect U.S. absorption and in turn net
exports. Piecing things together:
∆Y = ∆A + ∆NX ,
c c∗ a
∆Y = − ∆i − ∆i∗ .
1+a 1+a
As is clear from the expression, ∆i∗ affects domestic output directly (and not just through
∆i), resulting in a violation of the exclusion restriction central to instrumental variable
estimation. However, note that this violation is easily resolved by including net exports
as a control, or even just base country output, something we do later in the estimation.
Moreover, in this simple static model, all effects are contemporaneous. However, in practice
the feedback loop of higher U.S. interest rates to lower net exports to lower output will take
place gradually, in large part alleviating the exclusion restriction violation.
Financial spillovers with soft pegs
Consider now a more general setting with financial spillovers, that is, g > 0 and f > 0
and a soft peg. That is, the central bank may adjust using interest rates and allow some
movement of the exchange rate.
This will affect the pass through of U.S. interest rates to domestic rates since now:
1 d+g ∗
∆i = ∆e + ∆i ,
d d
where ∆E ∈ ±∆e refers to some band within which the exchange rate is allowed to
fluctuate.

The effect on output from changes in U.S. interest rates is very similar, but with an
added term:
c c∗ a
∆Y = ∆i − ∆i∗ + (b − f )∆e .
1+a 1+a
Under a hard peg policy, with ∆e = 0, an increase in U.S. interest rates boosts home
interest rates but it no longer does so one-to-one, as explained earlier. Partial flexibility in
exchange rates under a soft peg, with |∆e| > 0, gives some further monetary autonomy
to the home economy, and reduces the pass-through to home interest rates, all else equal.
This additional flexibility in exchange rates, however, results in other financial and trade
spillovers due to dependence of domestic absorption and net exports on the exchange rate
as shown by the term (b − f )∆e.
Summarizing our discussion, it is important to recognize that exogenous variation in
interest rates (induced either through the trilemma mechanism as just discussed, or through
alternative channels) has effects through domestic absorption and through net exports.
This secondary channel, if not properly controlled for, generates violations of the exclusion
restriction.
Importantly, because most studies of monetary economies take a closed economy
perspective, these international channels are not factored in. For example, this would be
true for the commonly used Romer and Romer (2004) shocks.8 This is quite possibly an
under-appreciated source of bias when estimating the effects of monetary policy. Below we
develop methods to quantify and correct for this sort of bias.
4. Measuring long-run treatment effects in finite samples
The statistical approach used to measure the propagation of a monetary experiment over
long horizons can make a big difference in the impulse response estimates obtained in
a finite sample. The impulse response representation of the data is the infinite moving
average representation or MA(∞). Thus, it is natural to make this assumption in thinking
about the data generating process since it relates directly to the object of interest. If in
addition the MA(∞) is invertible, it will have a corresponding infinite vector autoregressive
representation or VAR(∞).
8 lnclosed economy identification of monetary shocks (for e.g. Romer and Romer, 2004), expectations of
path of economic variables is factored in through Fed staff’s forecasts. While we do not use similar forecasts,
we believe our identification is less likely to be affected by concerns of systematic changes in base country
rates. One, we control for base country GDP growth and inflation rate when constructing our instrument.
Two, we also control for future path of base country GDP growth when estimating the impulse responses.
Three, at a broader level, it is unlikely that a base country systematically changes interest rate taking a peg’s
economic conditions into account.

In finite samples, there are two approaches to estimate impulse responses. One is
to truncate the VAR(∞) to VAR( p) and derive impulse responses from the truncated
model. The other is to estimate impulse responses with local projections with a similar lag
truncation. The estimates at long horizons turn out to be more precisely estimated with
local projections, as we now explain.
Truncated local projections and a VAR( p) can be thought of as members of the class
of semi-parametric sieve estimators (see, e.g., Grenander, 1981; Geman and Hwang, 1982;
Bühlmann, 1995). The thought experiment to justify the asymptotic analysis consists of
letting the truncation lag in both approaches to increase with the sample size at the same
rate, to be made explicit below.
Briefly, we show that impulse responses estimated with finite lag VARs are guaranteed
to be consistent up to the truncation lag, under some assumptions. Unless the truncated
VAR represents the data generating process (DGP) itself, this results in a quantifiable and
potentially large source of bias for horizons beyond the truncation lag. In contrast, local
projections do not have this problem.
The rest of the section provides formal arguments justifying this and other statements.
Many derivations are relegated to the Appendix for expedient reading. In addition, we
provide a simple Monte Carlo experiment to illustrate the differences between these two
methods more tangibly.
4.1. Setup
Our top level starting point is to assume that an m-dimensional vector valued process {yt }
is generated by the MA(∞) given by
∞
yt = ∑ B0t− j , (1)
j =0
where yt = (y1t , . . . , ymt )0 , B0 = Im , and {t } are a sequence of i.i.d. random vectors with
mean zero and positive definite covariance matrix Σ.9 We omit deterministic terms (such
as the constant and time trends) for convenience, but without loss of generality. Setting
aside identification to simplify the presentation of the main ideas, note that Equation 1 is
the impulse response representation of {yt }.
Let || Bj ||2 = tr ( B0j Bj ), i.e., the Froebenius norm. Further assume that ∑∞
j=0 || B j || < ∞,
and det{ B(z)} 6= 0 for |z| ≤ 1. These are the conditions in the Wold decomposition
9 Thisis the assumption made, for example, in, e.g., Lewis and Reinsel (1985) and Kuersteiner (2005).
However, Gonçalves and Kilian (2007) provide similar results for possibly conditionally heteroskedastic
martingale difference sequences.
10

theorem (Wold, 1938) by which any stationary vector process can be decomposed into its
deterministic components and Equation 1. Note that the assumption of stationarity is not
required for local projections (Jordà, 2005; Plagborg-Møller and Wolf, 2018).
4.2. Impulse responses from a VAR(p)

Under the stated assumptions, Equation 1 is invertible and can be expressed as the following
infinite order vector autoregression or VAR(∞) process
∞
yt = ∑ A j yt − j + t , (2)
j =1
where ∑∞ ∞
j=1 || A j || < ∞ and A ( z ) = I − ∑ j=1 A j z = B ( z )
j −1 satisfies det{ A ( z )} 6 = 0 for
|z| ≤ 1.
Assuming the data are generated by a process such as Equation 1 is quite general and it
will allow us to investigate instances when the monetary shocks have long-run effects. Note
that invertibility is not a necessary condition for local projections, as discussed in Stock and
Watson (2018) and Plagborg-Møller and Wolf (2018).
Using this setting and under a variety of general assumptions, Lewis and Reinsel
(1985), Kuersteiner (2005), Gonçalves and Kilian (2007) and references therein establish the
consistency and asymptotic normality of the VAR(p) coefficient matrices A j for j = 1, . . . , p.
However, two recurring assumptions merit further discussion. One is the assumption that
p → ∞ as T → ∞ at a rate p3 /T → 0. That is, p is allowed to grow with the sample
size, but not too fast. As remarked by Kuersteiner (2005), this rate does not provide much
guidance as to how to choose p and hence, he provides an optimal procedure to determine
p in finite samples.
The second and more relevant assumption often made is that p is chosen as a function
of T such that
∞
1
T 1/2
∑ || A j || → 0 as p, T → ∞, (3)
j = p +1
to ensure consistency of Â1 , . . . , Â p .

In turn, this means that the impulse response function will be consistently estimated
up to horizon h. To see this, recall the well-known recursion (Durbin, 1959) relating
moving-average and autoregressive coefficient matrices,
Bh = Ah + Ah−1 B1 + · · · + A1 Bh−1 . (4)
Given this recursion and direct application of the continuous mapping theorem, it is easy
11

to see that consistency of Â1 , . . . , Â p means that
p
|| B̂j − Bj || −
→0 as p, T → ∞ for j = 1, . . . , p. (5)
However, consistency of Bh is not guaranteed for h > p. The source of the inconsistency
can be quantified to be
A p+1 Bh−( p+1) + · · · + Ah−1 B1 + Ah ,
so that the sign of the inconsistency is, a priori, unknown.

Note that, all along, we have assumed that the truncation lag is optimally chosen—
though in practical applications, it is likely that practitioners will chose to small a truncation
lag so that
∞
1
T 1/2
∑ A j → c 6= 0 ,
j = p +1
in which case, the inconsistency due to misspecification of the model is of order T −1/2 .
Setting aside this problem, a simple example illustrates inconsistency resulting from optimal
truncation alone.
Consider, for expositional purposes, truncation at p = 1 and for now assume that this is
sufficient to ensure consistency of Â1 . Using expression Equation 4, we next show both
the impulse response coefficients of the VAR(∞) (in the first column of Equation 6 below)
against the relationships resulting from the truncated VAR(1) (presented in the second
column of Equation 6 below), specifically
 


 B1 = A1 

 B1 = A1

 

B = A21 + A2 B = A21

 

 2  2

 

VAR(∞) → B3 = A31 + 2A1 A2 + A3 , VAR(1) → B3 = A31 . (6)

 

B4 = A41 + 3A21 A2 + 2A1 A3 + A4 B4 = A41

 


 


 

· · ·
 · · ·

Clearly, extending the lag-length of the VAR from p to H (where H is the longest
horizon displayed for the impulse response function) would solve the problem (assuming
the remaining conditions in the theorem are met). In practice, and due to limited sample
sizes, practitioners tend to favor relatively short lag-lengths, which potentially exposes them
to undesirably inconsistent estimates of the impulse response even when the coefficients of
the VAR are consistently estimated.
12

4.3. Impulse responses from local projections
Under the same assumptions stated earlier, local projections have better consistency proper-
ties than truncated VARs, in general. Consider the local projection
yt+h = Ah,1 yt−1 + · · · + Ah,p yt− p + ut+h . (7)
The assumptions that ensure consistency of the Â j for j = 1, . . . , p also ensure that
p
|| Âh,1 − Bh+1 || −
→0 as p, T → ∞ , (8)
as shown in Lusompa (2019). The appendix provides a broad overview of the main
arguments for this result. In particular, even if the DGP is a VAR(∞) and the local
projection only includes up to p lags, the LP provides a direct estimate (rather than a
nonlinear transformation of estimated parameters) of the impulse response coefficient
matrix for any h. Moreover, the appendix shows that consistency of the truncated local
projection is achieved under the exact same conditions as consistency of the VAR coefficient
matrices for the first p lags.
What does this all mean in practice? Under general conditions that would lead re-
searchers to consistently estimate a finite-order VAR given a sample of data, the impulse
response coefficients corresponding to horizons that exceed the truncation lag of the VAR
would be expected to exhibit bias. This bias can be quantified theoretically though in
practice it will depend on several factors. The bias arises because the truncation of the VAR
exposes the calculation of the impulse response coefficient to omitted terms that may not be
small, specially when the data are persistent. In those cases, of course, a higher truncation
lag would be advisable. In practice, researchers tend to choose relatively parsimonious VAR
specifications due to the limited sample sizes and the parameter estimation load required in
a VAR. Moreover, Bayesian methods are unlikely to result in better outcomes as shrinkage
to the priors commonly used will make the VARs more efficient, but possibly make the
bias even worse for impulse responses.
Local projections have several advantages. Because they provide a direct estimate of
the impulse response coefficient of interest (rather than obtaining the response coefficients
as the result of a nonlinear transformation of estimated parameters), truncation of the
lag-length is not as big a concern. Moreover, because local projections can be estimated
equation by equation, the parametric load is reduced. In settings where the data are
persistent, such as ours, and where interest is in characterizing impulse responses at long
horizons, local projections are more appealing.
13

4.4. Impulse response small-sample biases: Monte Carlo evidence
To further illustrate the potential biases that can arise in practice, the next section provides
a simple Monte Carlo exercise to illustrate the main ideas. The DGP is an MA(10) whose
coefficients are generated from a Gaussian basis function. Specifically, we set
10
yt = µ + ∑ b j ut− j + ut ; ut ∼ N (0, 0.5) ;
j =1
b∗j
( 2 )
j−β

bj = ; b∗j = α exp − , j = 1, . . . , 40 .
∑ j b∗j δ
with α = 1, β = 6, and δ = 12. Using these parameter choices, the maximum impact is
normalized to have size 1, which occurs 6 periods after impact, and tapers back to 0 with a
√
half-live given by 12 ln 2 ≈ 10. The resulting b∗j are then scaled so that the long run impact
is normalized to 1, resulting in the b j parameters that we ultimately use in the simulation.
The experiment then consists in simulating the MA(10) using Gaussian shocks N (0, 0.5).
The constant term is set to µ = 1. We initialize the process with 500 draws that are
then discarded to avoid initialization issues. We consider two sample sizes, 150 and 300
observations. The experiments are repeated 1,000 times and the average over these Monte
Carlo replications is then reported. Hence, we estimate impulse responses for this MA(10)
process using a VAR(4) and also by local projections with 4 lags, and denoted LP(4). Note
that both of these models are misspecified since a lag length of 10 would be appropriate.
The top row of Figure 1 displays the raw impulse responses and the bottom row displays
the cumulated responses, which are the responses later used in our analysis. Moreover and
without loss of generality, we flip the sign of the responses to match the empirical responses
below. Several results are worth noting. First, it is clear that the impulse responses estimated
with local projections fit the true impulse responses much better than those from a VAR.
Because the MA(10) is designed to have a maximum effect that is delayed by 6 periods
after the shock, the VAR(4) has a hard time picking the shape of the response. This is not
alleviated by increasing the sample size to 300. As a result, the cumulated responses can be
far off, as shown in the figure.
The bottom row of Figure 1 reports the cumulated responses corresponding to the
impulse responses calculated in the top row. Because the cumulated impact is normalized
to 1, it is easy to assess the bias resulting from the small sample estimates. Regardless of
whether one uses the smaller or the large sample size, the VAR(4) based response reduces
the long-run impact by about 40% from the true measure. This bias is less than half for the
LP(4), even though the LP has the same lag length as the VAR.
14

Figure 1: Evaluating VAR versus LP responses in finite samples: A Monte Carlo experiment
Impulse response Impulse response

Sample Size: 150 Sample Size: 300
.05
.05
0
0
-.05
-.05
-.1
-.1
-.15
-.15
0 5 10 15 20 0 5 10 15 20
Horizon Horizon
Cumulated response Cumulated response

Sample Size: 150 Sample Size: 300
0
0
-.2
-.2
-.4
-.4
-.6
-.6
-.8
-.8
-1
-1
0 5 10 15 20 0 5 10 15 20
Horizon Horizon
True VAR(4) LP(4)
Notes: Impulse responses in the top row calculated with a VAR(4) and local projections with four lags or LP(4). Bottom row displays
cumulated responses. Sample size with 150 observations displayed in the left-hand column, sample size with 300 observations displayed
on the right-hand column. Averages over 1,000 Monte Carlo replications. See text.
15

5. Empirical approach: Monetary shocks have long-lived effects
The basic empirical approach relies on local projections (Jordà, 2005) estimated with
instrumental variables (LPIV). Several applications of these methods are available in the
literature, though a more general discussion of the method can be found in Ramey (2016),
Stock and Watson (2018) and Jordà, Schularick, and Taylor (2019).
Based on the latter, we estimate in particular the impulse responses
y j,t+h − yi,t−1 = αi,h + ∆i

b j,t β h + x j,t γh + yb( j,t,t+h) λh + v j,t+h , (9)
for h = 0, 1, . . . , H; i = 1, . . . , N; t = t0 , . . . , T, where y j,t+h is the outcome variable for

country j observed h periods from today, αi,h are country fixed effects at horizon h, ∆i b j,t refers
to the instrumented change in the short-term government bond (3-months in duration),
our stand-in for the policy rate which we instrument with z j,t , the trilemma instrument as
discussed earlier; and x j,t collects all additional controls including lags of the outcome and
interest rates, as well as lagged values of other macro aggregates.10 Moreover, we control for
global business cycle effects through a global real GDP control variable to parsimoniously
soak up time effects. To tie our hands as much as possible against finding any long-run
effects of monetary policy, we also control for future values of real GDP in the base country
at each horizon h, denoted by yb( j,t,t+h) . Motivated by the identification discussion in
Section 3.1, these set of controls also account for potentially anticipated information on the
base-country business cycle. We estimate Equation 9 with instrumental variable methods
and report cluster robust standard errors.
Table 1 reports the first-stage regression of the pegging country’s short term interest rate
∆i j,t on the instrument z j,t with controls x j,t , country fixed effects and (robust) clustered
standard errors. The t-statistic is well above 3 for full sample and post-WW2 samples
illustrating that it is not a weak instrument. We refer the reader to Jordà, Schularick, and
Taylor (2019) for detailed discussion on the instrument and proceed henceforth assuming
the reader is on board regarding instrument relevance and strength.
10 The list of domestic macro-financial controls used include log real GDP; log real consumption per
capita; log real investment per capita; log consumer price index; short-term interest rate (usually a 3-month
government security); long-term interest rate (usually a 5-year government security); log real house prices;
log real stock prices; and the credit to GDP ratio. The variables enter in first differences except for interest
rates. Contemporaneous terms (except for the left-hand side variable) and two lags are included. We control
for contemporaneous values of other macro-financial variables for two purposes a) base rate movements
might be predictable by current home macro-conditions, and b) we wanted to impose restrictions in the spirit
of Cholesky ordering whereby real GDP is ordered at the top.
16

Table 1: Trilemma instrument: First stage evidence
pegs (q = 1) All years PreWW2 PostWW2

z j,t 0.52∗∗∗ 0.35∗∗ 0.56∗∗∗
t-statistic [8.62] [2.05] [8.97]
Obs 672 148 524
Notes: ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Full sample: 1870–2015 excluding WW1: 1914–1919 and WW2: 1939–1947. Pre-WW2
sample: 1870–1938 (excluding 1914–1919). Post WW2 sample: 1948–2015. These regressions include country fixed effects as well as up
to two lags of the first difference in log real GDP, log real consumption, investment to GDP ratio, credit to GDP, short and long-term
government rates, log real house prices, log real stock prices, and CPI inflation. In addition we include world GDP growth to capture
global cycles. See text.
5.1. Main results

The main story is illustrated by the response of real GDP to a shock to domestic interest rates
using the trilemma instrument. Before we show the main results, we highlight the value of
our instrumental variable with a comparison of the response of real GDP to a shock in the
short-term domestic interest rate calculated using selection-on-observables identification
versus identification with the trilemma instrument. This is shown in Table 2. The table
reports coefficient estimates of the impulse response calculated with each identification
approach for the full and post-WW2 samples. LP-OLS refers to identification via selection,
LP-IV to the trilemma instrument identification. The samples are restricted to pegging
economies to match the samples in both cases.11
Table 2 is organized as follows. We provide the coefficient estimates by row, with a
test of the null hypothesis that LP-OLS and LP-IV estimates are equlal for the full and
post-WW2 samples. The differences between identification schemes could not be starker:
LP-IV estimates are economically and statistically significant, and the LP-IV response is
considerably larger at all horizons.
We display these results graphically in Figure 2. Regardless of the sample used, a 1
percentage point increase in domestic short-term interest rates has sizable and long-lasting
effects on GDP. In the full sample, GDP declines by 3.69 percent over 12 years. A similar
effect is found when we restrict the sample to post-WW2. The drop 12 years after impact is
3.49 percent. This is a far cry from traditional notions of long-run neutrality found in the
literature.
What is the source of this persistent decline? We decompose GDP into its components,
namely, hours worked (employees times number of hours per employee); capital stock
(measured capital in machines and buildings); and the Solow residual (using a Cobb-
Douglas production function) labeled as total factor productivity (TFP). (Using the Imbs
11 The plots and inference are robust to using real GDP per capita. See Table A1
17

Table 2: LP-OLS vs. LP-IV. Attenuation bias of real GDP responses to interest rates. Trilemma instrument.
Matched samples
Responses of real GDP at years 0 to 10 (100 × log change from year 0 baseline).
(a) Full Sample OLS-IV (b) Post-WW2 OLS-IV
Year LP-OLS LP-IV p-value LP-OLS LP-IV p-value
(1) (2) (3) (4) (5) (6)
h=0 0.07∗∗ 0.01 0.50 0.03 0.08 0.44
(0.03) (0.08) (0.02) (0.07)
h=2 -0.31∗∗ -1.37∗∗∗ 0.00 -0.29∗∗∗ -1.20∗∗∗ 0.00

(0.14) (0.34) (0.11) (0.31)
h=4 -0.22 -1.89∗∗∗ 0.00 -0.15 -1.24∗∗∗ 0.01

(0.23) (0.48) (0.20) (0.40)
h=6 -0.12 -2.67∗∗∗ 0.00 -0.07 -2.55∗∗∗ 0.00

(0.36) (0.59) (0.29) (0.59)
h=8 -0.34 -4.83∗∗∗ 0.00 0.08 -2.93∗∗∗ 0.00

(0.35) (1.03) (0.35) (0.71)
h = 10 0.13 -4.22∗∗∗ 0.00 0.34 -2.80∗∗∗ 0.00

(0.37) (1.03) (0.42) (0.80)
h = 12 0.35 -3.82∗∗∗ 0.00 0.56 -3.31∗∗∗ 0.00

(0.40) (0.89) (0.43) (0.92)
KP weak IV 46.71 50.56
H0 : LATE = 0 0.00 0.00 0.00 0.00
Observations 835 765 655 585
Notes: ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Cluster robust standard errors in parentheses.Full sample: 1890–2015 excluding WW1:
1914–1919 and WW2: 1939–1947. Post WW2 sample: 1948–2015. Matched sample indicates LP-OLS sample matches the sample used to
obtain LP-IV estimates. KP weak IV refers to the Kleibergen-Paap test for weak instruments. H0 : LATE = 0 refers to the p-value of the
test of the null hypothesis that the coefficients for h = 0, ..., 10 are jointly zero for a given subpopulation. OLS = IV shows the p-value
for the Hausmann test of the null that OLS estimates equal IV estimates. See text.
(1999) correction, in Appendix Section B we further decompose the Solow residual into
factor-utilization plus a residual utilization-adjusted TFP.)
Figure 3 displays the responses of each of these components to the same shock to the
domestic short-term interest rate instrumented with the trilemma, both for the full and the
post-WW2 samples. Figure 3a displays the responses of total hours worked, capital and
raw TFP without error bands to provide a clearer sense of the dynamic paths. Figure 3b
displays each of the components with the one and two standard deviation error bands. In
Appendix Section D.2, we provide corresponding figure for post-WW2 sample.
Several features deserve mention. Figure 3 shows that there are similar declines in
capital and raw TFP whereas total hours worked exhibits a much flatter pattern. Because
18

Figure 2: Baseline response to 100 bps trilemma shock: Real GDP
(a) Full sample: 1890–2015 (b) Post-WW2 sample: 1948–2015

2
2
0
0
Percent
Percent
-2
-2
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
IV OLS IV OLS
Notes: Response to a 100 bps shock in domestic interest rate instrumented with the trilemma. Responses for pegging economies. Full
sample: 1890–2015 (World Wars excluded). Post-WW2 sample: 1948–2015.LP-IV estimates displayed as a solid blue line and and 1 S.D.
and 2 S.D. confidence bands constructed using cluster-robust standard errors. See text.
Figure 3: Baseline response to 100 bps trilemma shock: Real GDP and components
(a) (b)
real GDP total hours
0
0
Percent
Percent
-2
-2
0
-4
-4
-6
-6
-2
0 4 8 12 0 4 8 12
Percent
Year Year
capital stock TFP
-4
0
Percent
Percent
-2
-2
-6
-4
-4
0 4 8 12
Year
-6
-6
GDP TFP K L 0 4 8 12 0 4 8 12
Year Year
sample: 1890–2015 (World Wars excluded). LP-IV estimates displayed as a thick line and and 1 S.D. and 2 S.D. confidence bands. See
text.
19

Figure 4: Baseline response to 100 bps trilemma shock: a) Short term nominal interest rate, b) real interest
rate, and c) real interest rate multiplier
(a) nominal interest rate (b) real interest rate (c) real interest rate multiplier
Short-term nominal interest rate Short term real interest rate Real interest rate multiplier
3
1.5
0
1
-.2
.5
Percent
1
0
-.4
0
-.5
-1
-.6
-1
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Notes: Response to a 100 bps shock in domestic interest rate instrumented with the trilemma. a) Responses of short term nominal interest
rate for pegging economies. b) Responses of short term real interest rate for pegging economies. Inflation expectations constructed
from the impulse response of consumer price level index to the same trilemma shock. c) Response of cumulative change in real GDP
divided by cumulative change in short term nominal interest rate for pegging economies. Sample: 1890–2015 (World Wars excluded).
LP-IV estimates displayed as a solid blue line and and 1 S.D. and 2 S.D. confidence bands. See text.
capital enters the production function with a smaller weight, it should be clear from the
figure that most of the decline in GDP is explained by TFP variable, and then capital, with
total hours worked mostly flat.
To be sure, the response of hours worked conforms well with the textbook response
to a monetary shock. Total hours fall in the short-run, but then recover quickly and
remain mostly flat. Capital accumulation also follows textbook dynamics in the short-
run. The response is initially muted but builds up over time. But unlike a textbook new
Keynesian model (Galı́, 2015b), capital does not appear to recover even 12 years after the
shock. Similarly, TFP falls gradually rather than suddenly. Over time, the decline in TFP
accelerates, ending at a level 3.32 percent lower in full sample by year 12 relative to year 0.
One explanation for the long-lasting effects of the monetary shock could be that domestic
interest rates remain elevated for a long period of time as well. In other words, persistence
is generated by a delayed response in interest rates. A simple check of this proposition can
be done in two steps.
Figure 4 shows that the short-term real interest rate does indeed take approximately 8
years to return to zero deviation, while the nominal interest rate returns to zero deviation
after 4 years. The response of nominal interest rate is typical of what has been reported
often in the literature (see, e.g., Christiano, Eichenbaum, and Evans, 1999; Ramey, 2016).
Secondly, we can calculate the responses of the main variables normalized by the response
of interest rates over time to sterilize the dynamics of interest rates themselves. This is no
different than calculating a multiplier (see, e.g., Ramey, 2016; Ramey and Zubairy, 2018).
20

Thus, in Figure 4c we show the ratio of the cumulative change in GDP to the area under
the real interest rate path in Figure 4b. In a given period, the difference in the level of real
interest rate relative to the counterfactual path measures the tightness of monetary policy.
We consider these cumulative gaps as a measure of overall monetary policy tightness (the
area under the solid line in Figure 4b). By year 12, the multiplier is -0.4 in the full sample.
In post-WW2 sample, this multiplier is similar (See Section D.2 in the appendix). These
numbers indicate that the monetary policy shocks, measured as cumulative response of
real interest rate, are somewhat larger but comparable with those obtained in the literature
using VAR and Romer and Romer (2004) shocks for the U.S.12
6. Robustness and discussion
Our baseline specification included lags and current values of global GDP growth and base
country GDP growth, and future values of base country GDP growth. This rich specification
served multiple purposes. Global shocks that caused bases to change interest rates are
controlled for during instrument construction, as well through use of these controls, in IRF
estimation. Comparison with OLS estimates, which control for contemporaneous home
economy macro-variables (as in a Cholesky ordering), further allay some concerns on
systematic structural breaks in GDP or TFP growth picked up as regime shifts over decades.
We now discuss further robustness checks to ensure that the persistent effects we
identified are not misattributed to monetary policy shocks.
6.1. Spillover correction for the trilemma instrument

A violation of the exclusion restriction could occur if base rates affect home outcomes
through channels other than movements in home rates or the spillovers highlighted in
Section 3.1. Additional influences via such channels are sometimes referred to as spillover
effects. These could occur if base rates proxy for factors common to all countries. That
said, these factors would have to persist despite having included global GDP to soak
up such business cycle variation. In addition to the spillover control strategy used in
our baseline specification, we now assess such spillover effects more formally, using two
separate approaches: a) a control function approach developed in Jordà, Schularick, and
Taylor (2019), and b) controlling for foreign variables at each horizon to remove any spillover
channels from the interest rate channel.
12 In their online appendix (Table A.1), motivated by the findings of Coibion (2012), Nakamura and
Steinsson (2018) report the multiplier of monthly industrial production to monetary shocks over 36 months to
lie between 1 and 2 using the local projections specification with VAR shocks and Romer-Romer shocks.
21

Synthetic control function approach
In this section we use a direct spillover correction based on Conley, Hansen, and Rossi
(2012); van Kippersluis and Rietveld (2018) that takes advantage of the subpopulation of
floats (for which our instrument is theoretically invalid). The solution borrows from Jordà,
Schularick, and Taylor (2019), where the reader can find the detailed derivations. Here we
simply sketch the main ideas with a simple example.
Suppose the equation to be estimated is:
∆y = ∆i β + zφ + v ,
where ∆y is the outcome variable, ∆i is an endogenous variable, and z is an instrumental

variable that violates the exclusion restriction as long as φ 6= 0. However, we may assume
that E(z v) = 0, as is common in regression analysis when z is exogenous. Although for
the subpopulations of floats z is not a valid instrument, it can be exploited to provide an
estimate of the spillover effect φ using an OLS estimate of the previous regression and the
first stage (invalid) regression
∆i = z b + η .
Without loss of generality, we can assume that β = λφ where λ is an unknown parameter

than can be calibrated through economic arguments since it is simply a statement of how
large is the spillover effect relative to the effect of the endogenous variable. We choose a
conservative set of values for λ ∈ [1, 8]. In our case, the endogenous variable is the change
in the home nominal interest rate and the instrument is, essentially, the surprise in the base
country rate. As a result, one can show that
(φ̂OLS + b̂ β̂OLS )
φ̂(λ) = ,
1 + λ b̂
and hence estimate by instrumental variables, on the subpopulation of pegs, the following
spillover corrected regression
(∆y − zφ̂(λ)) = ∆i β + v + z(φ̂(λ) − φ(λ)) ,
where it is clear that the error term is the sum of two terms whose mean converges to zero
asymptotically. Figure 5 shows our spillover-corrected estimates of response of output to a
100 bps monetary policy shock. A light-green shaded area with dashed border shows the
spillover corrections. Direct spillover corrections have negligible effects on our estimates.
22

Figure 5: Response to 100 bps trilemma shock with spillover corrections: Real GDP
(a) Full sample: 1890–2015 (b) Post-WW2 sample: 1948–2015

2
2
0
0
Percent
Percent
-2
-2
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
IV OLS IV spillover corrected IV OLS IV spillover corrected
sample: 1890–2015 (World Wars excluded). Post-WW2 sample: 1948–2015. LP-OLS estimates displayed as a dashed red line, LP-IV
estimates displayed as a solid blue line and 1 S.D. and 2 S.D. confidence bands, LP-IV spillover corrected estimates displayed as a light
green shaded area with dashed border, using λ ∈ [1, 8]. See text and Jordà, Schularick, and Taylor (2019).
Controlling for base country GDP growth, current account and exchange rate
A second approach that attempts to provide validity to the exclusion restriction is directly
controlling for a primary channel through which the spillover effects may originate. A
reduction in demand my arise from other trading partners when the tightening affects other
countries as well. Or the interest rate channel induced contraction in pegging economy,
by reducing demand for other countries’ output. may be subject to spillbacks. If the
transmission is primarily driven by increased trade, controlling for global GDP growth
can potentially absorb these effects allowing us to remove domestic demand effects with
respect to international spillbacks.
A monetary tightening in the base country may reduce the demand for goods from the
pegging economy. This effect would amplify the effect of the trilemma shock on home
output. Another implication from the Mundell-Fleming-Dornbusch model is that there
are financial spillovers that may amplify the effects through the exchange rate channel. To
account for these effects, at each horizon h, we control for global GDP growth rate, base
country’s GDP growth rate, exchange rate of the pegging economy with respect to the USD
and the current account of the peg. Since we do not have exchange rate data with respect
to other countries, we indirectly control for those spillovers using the current account of
the peg country at each horizon.
23

Figure 6: Response to 100 bps trilemma shock with additional controls: Real GDP
Controls: base GDP, global GDP, CA and XRD 5 breaks in TFP
(a) Open economy model based controls (b) Structural breaks in TFP growth
Real GDP Real GDP
2
2
0
0
-2
-2
Percent
Percent
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
sample: 1890–2015 (World Wars excluded). LP-IV estimates displayed as a solid blue line and 1 S.D. and 2 S.D. confidence bands. See
text.
Figure 6a plots the IRFs to trilemma identified shock. Controlling for variables motivated
by Section 3.1 does not affect our main result - monetary shocks have a very persistent
effect on real GDP.
6.2. Accounting for structural breaks

Fernald (2014) and Gordon (2016) have convincingly argued that there are structural breaks
in TFP growth in the U.S. economic trajectory. It is plausible that there are structural breaks
in other economies’ TFP growth rates. If these structural breaks implying slowdown in TFP
growth occur around the identified monetary shocks, it could bias our results leading us
to attribute the persistent effects incorrectly to monetary shocks. To address this concern,
we first estimate five structural breaks in TFP growth and GDP growth for each country in
our sample using the UD-max statistic of Bai & Perron (1998). We report these estimated
structural break dates in the appendix D.7. Then in our baseline specification, we allow
output growth to lie in either of the five regimes at horizon zero as well as horizon h. Our
specification is conservative since we allow horizon h regime changes in the estimation
along with horizon 0 regimes.
Figure 6b plots the estimated impulse response when including structural breaks in TFP
growth. As evident, our results are robust to accounting for structural breaks.13
13 In the appendix D.5, we report the IRFs allowing for structural breaks in GDP growth.
24

6.3. Additional robustness exercises
We conduct variety of additional robustness exercises to check the validity of our results.
These are shown in the appendix and we describe them briefly now.
In Section D.3, we show the IRFs of real GDP to monetary shock when we introduce the
control variables in levels instead of first differences. We also show the responses when we
increase the number of lags in our estimation to five. Results are robust to either of these
substitutions.
In Section D.4, we show trajectory of various macroeconomic and financial variables to
the identified monetary shock. In particular, we trace out path of CPI-based measure of
price level, consumption per capita, investment per capita, long-term nominal rate, loans to
GDP ratio, real and nominal house price indices, and real and nominal stock price indices.
In Section E, we report the estimation results when the dependent variable is real GDP
per capita instead of real GDP. Our results are robust to not including data post-2007
(i.e., after the global financial crisis). Finally, while we used an unbalanced sample in our
baseline estimation (i.e., estimates of IRFs at shorter horizons are estimated with more data
than those at longer horizons), our results are robust to using a fixed estimation sample at
all horizons.
6.4. Is the U.S. different?

The U.S. for most of our historical sample is not a pegging economy (apart from the Gold
standard years). It is the quintessential base country for many economies in our sample.
For this reason, the trilemma instrument mechanically sets aside any information coming
from the U.S. during estimation. It is natural to wonder the extent to which U.S. data
conforms with the patterns presented so far.
In this section, we examine U.S. data post-WW2. This allows us to incorporate three
useful modifications. First, we use higher frequency quarterly data. Second, we use the
alternative utilization adjusted series for TFP constructed by Fernald (2014). Third, in order
to achieve identification we rely on the instrumental variable constructed by Romer and
Romer (2004) based on the Federal Reserve staff’s implied forecast errors for the policy rate,
and extended to recent years by Wieland and Yang (2016).
It turns out that results based on U.S. data largely confirm the results we reported in
the previous section for non-U.S. pegging economies. Figure 7 plots the the path of real
GDP along with its three components: total hours worked, capital, and utilization adjusted
TFP. The responses are qualitatively similar to those in the long-run panel, although the
amplitudes are more muted. Quarterly data is naturally noisier than yearly data, but
25

Figure 7: Baseline response to 100 bps Romer and Romer (2004) shock : U.S. postwar data
Real GDP (Util. Adj.) TFP index (Fernald 2014)
.5
.5
0
0
-.5
Percent
Percent
-.5
-1
-1.5
-1
-2
0 8 16 24 32 0 8 16 24 32
Quarter Quarter
IV OLS
Hours, bus sector Capital input

1
0
-.5
0
Percent
Percent
-1
-1
-1.5
-2
-2
0 8 16 24 32 0 8 16 24 32
Quarter Quarter
IV OLS
Notes: Response to a 100 bps shock in federal funds rate rate instrumented with policy forecast residuals (Romer and Romer, 2004;
Wieland and Yang, 2016). Responses of real GDP, utilization-adjusted TFP (Fernald, 2014), capital stock and hours worked for U.S.
economy. Quarterly sample: 1969-Q1: 2007-Q4. Quarterly data series are taken from Fernald (2014).
smoothing over a temporary recovery in GDP 4 to 5 years after the shock, real GDP ends
nearly one percent lower 8 years after impact. Utilization adjusted TFP and hours exhibit
a similar U-shaped pattern, with TFP nearly back to zero by year 8. Strikingly, capital
accumulation exhibits a protracted decline over the entire period, ending about 1.25 percent
lower after eight years. For comparison, we plot the various components on the same graph
in the appendix (see Section D.9).14
14 We estimate the following specification for the U.S. economy:
yt+h − yt−1 = αh + ∆i
b t β h + xt γ h + v t + h ; h = 0, 1, . . . , H; i = 1, . . . , N; t = t0 , . . . , T ;
where yt+h is the outcome variable at horizon h, ∆ib t is the instrumented change in the Federal Funds rate,
xt is the set of controls that includes contemporary and four lags of log real GDP, log CPI, and changes in
federal funds rate. We do not include the contemporary variable when it is same as the dependent variable.
We report robust standard errors.
26

The data for utilization-adjusted TFP series are based on sectoral data for the U.S. that
account for heterogeneity across workers and types of capital. Fernald (2014) notes that
there are various other corrections that are not conducted in the quarterly series due to the
lack of rich-industry level data.
The finding that monetary policy shocks can affect utilization-adjusted TFP echo the
Evans (1992)’s critique of using Solow residuals as productivity shocks in RBC models.
While the construction of quarterly utilization-adjusted TFP series is detailed and thorough,
our analysis suggests caution against using the quarterly-adjusted residuals as “pure” TFP
shocks (i.e., perfectly orthogonal to demand shocks).15,16
7. A model of hysteresis
Impulse responses calculated with standard methods that internally favor reversion to
the mean will tend to underestimate the value of the response at longer horizons. By
relying on local projections, we allow the data to more directly speak as to its long-run
properties. The evidence presented in the previous sections strongly indicate that these
long-run effects are important and require further investigation. In order to think through
a possible mechanism that explains our empirical findings, we augment a textbook New
Keynesian model with endogenous productivity growth in a stylized manner. The baseline
framework is that of a medium-scale DSGE model as in Christiano, Eichenbaum, and Evans
2005; Smets and Wouters 2007; or Justiniano, Primiceri, and Tambalotti 2013.
Households demand final consumption good and supply differentiated labor to a labor
union. Final consumption good is packaged by a perfectly competitive retailer using a
Dixit-Stiglitz aggregate of a continuum of intermediate goods. Each intermediate good
is produced by a monopolistically-competitive firm that sets prices in a Calvo fashion.
These firm hire labor unit packaged by labor unions that also set wages in a staggered
fashion. Government runs a balanced budget every period and central bank sets short
term nominal interest rate on a riskfree bond (in net zero supply) following a rule that
we describe shortly. Goods market and bond markets clear every period. We leave the
formal model to the appendix, and focus on the key departure that allows us to introduce
productivity hysteresis with a parametrically convenient process.
15 Ramey (2016) also documents that utilization adjusted TFP series fail Granger causality tests (See Table
11 in her paper). We are grateful to Valerie Ramey for alerting us to that analysis.
16 In the appendix D.10, we show that similar results are obtained with samples beginning until at least
1973Q2. However, these persistent effect results are not robust to considering further shorter samples for the
U.S. economy. This is consistent with the findings of Coibion, Gorodnichenko, and Ulate (2017).
27

7.1. Hysteresis effects
In order to be able to capture the empirical features describe in the previous sections, we
examine a richer specification of the low of motion for total factor productivity Zt than is
conventional. In particular, we assume that the law of motion for Zt is:

f ,t−1
log Zt = log Zt−1 + µt + η log Yt−1 /Yt−1 ,
where µt is the exogenous component of the TFP growth rate, that may be subject to trend
f ,t−1
shocks. Yt is aggregate output at time t. Yt−1 is the flexible price level of output in period
t − 1 conditional on Zt−1 , and will be referred to as the potential output at time t − 1. The
second component denotes the endogenous component of TFP growth, where η is the
elasticity of TFP growth rate with respect to fluctuations in output due to nominal rigidities.
We refer to this as the hysteresis elasticity (to be consistent with DeLong and Summers 2012).
The above law of motion allows business cycles to affect TFP growth rate only in
the presence of nominal rigidities or inadequate stabilization. For clarity, we employ
this parametric-convenient functional form for hysteresis.17 A micro-founded model of
innovation and productivity growth that yields this exact representation under monetary
policy shocks can be found in the recent literature embedding endogenous growth into
DSGE models (Bianchi, Kung, and Morales, 2019; Garga and Singh, 2016). The effects of
business cycles on TFP growth rate that are unrelated to nominal rigidities can be denoted
by time varying values of µt , which may depend on other shocks (markup shocks, stationary
TFP shocks, discount factor shocks, capital quality shocks etc.). For ease of exposition, we
only focus on the hysteresis effects induced by the presence of nominal rigidities and treat
µt as an exogenous series.
This functional dependence creates a role for hysteresis stabilization by central banks in
a reduced-form manner (Yellen, 2016). Long-run effects of monetary policy shocks depend
on the value of η. Theoretically, there is no a priori reason to expect η to be positive.
While a “cleansing” effect of recessions may induce counter-cyclicality, recessions may
reduce funding access to firms to conduct R&D, skill development, and learning-by-doing.
The sign on the cyclicality of TFP misallocation is also ambiguous and depends on the
assumptions in a model. To clarify, in this paper, we are only able to discuss the sign and
the magnitude of η in response to temporary monetary shocks. We will provide below
estimates for η in response to monetary shocks from our estimated empirical IRFs.
17 A similar setup was used by Stadler (1990) in his seminal work.
28

7.2. Government
The central bank follows a Taylor rule in setting the nominal interest rate it . It responds to
deviations in inflation, output and output growth rate from time-t natural allocations.


ρ R φπ !φy 1−ρR
1 + it 1 + i t −1  πt Yt mp
= f ,t
 et , (10)
1 + iss 1 + iss πss Yt
where iss is the steady state nominal interest rate, πt is gross inflation rate, πss is the steady
f ,t
state inflation target, Yt is the time-t natural output, ρ R determines interest-rate smoothing
mp
and et∼ N (0, σr ) is the monetary policy shock.
We assume government balances budget every period, where total lumpsum taxes
go into giving a production subsidy to intermediate good producers, a wage subsidy to
workers and other government spending.
7.3. Simulations
As the DSGE model is intentionally standard, we take parameters from the literature
Justiniano, Primiceri, and Tambalotti (2013). We report these in Table 3. The steady state
parameters imply (annualized) real interest rate of 2.40%, and an investment-GDP ratio of
17%. For the monetary shock process, we chose the following parameters σr = 0.02 and
ρ R = 0.8. The short term nominal interest rate increases by about 10 basis points on impact
in the three simulations that we report here.
The new parameter in our model, relative to the business cycles literature, is η: the
hysteresis elasticity. Table 4 reports the point estimates for η implies by the estimated
impulse responses. We use a two-step classical minimum distance approach to recover
η. In the first step, we estimate the IRF of TFP and real GDP to monetary policy shock.
Using the estimated coefficients (see Figure 3), we then estimate η as the ratio of the two
IRFs. Following the persistent drop in output after the Great Recession in the US, DeLong
and Summers (2012) infer that this parameter could be as high as 0.24. While our estimate
is on the higher side, there is considerably large confidence interval. In our calibration
henceforth, we use the value of 0.10 which is also within the assumptions of DeLong and
Summers (2012).
Figure 8 plots the model-implied impulse responses for output, capital stock, real
interest rate and inflation after a monetary policy shock. Solid blue line reports the IRFs
for endogenous growth model with η = 0.10, and dashed blue line reports IRFs for the
comparable exogenous growth benchmark i.e. η = 0. The IRFs for output and capital stock
29

Table 3: Parameters
(a) Steady-state parameters

β δk α µ
Discount Capital Capital Trend
factor depreciation rate share growth rate
0.999 0.025 0.28 2%
(b) Parameters characterizing the dynamics
ν λp λw θp θw h
Inverse Price s.s. Wage s.s. Price Calvo Wage Calvo (Internal)
Frisch elasticity markup markup probability probability habit
1.00 0.15 0.15 0.750 0.750 0.5
a00 (1) 1
a 0 (1)
S”(1) φπ φy 1− λg η
Capital Investment Taylor rule Taylor rule (normalized) Government Hysteresis
utilization cost adjustment cost inflation response output response spending share elasticity
4 2 1.50 0.125 0.20 0.05
Notes: The table shows the parameter values of the model for the baseline calibration. See text.
Table 4: Point estimates for hysteresis elasticity η
pegs (q = 1) 1890–2015 1948–2015

η 0.28 0.42
[95% confidence band] [0.10, 0.45] [0.33, 0.50]
Notes: The point estimates are estimated with a two-step classical minimum distance approach using the IRF of TFP and the IRF of
GDP to the monetary policy shock in the second step. Figure 3 report the IRFs for GDP and TFP.
are plotted in percent deviations from an exogenous trend. For real interest rate, we plot
the actual path of real interest rate. Inflation is in percent deviation from steady state level.
Time is in quarters.
The model replicates the estimated empirical patterns. There is a persistent decline
in capital stock, output and TFP. Furthermore, the endogenous growth model exhibits
considerable amplification to the transitory shock because of the large hysteresis elasticity.
We define the accumulated gaps in TFP growth rate as the hysteresis. We next show the
path of output, and capital when the central bank sets interest rates following an augmented
Taylor rule:


ρ R φπ !φy ! φ H 1− ρ R
1 + it 1 + i t −1  πt Yt Ht mp
= f ,t f ,t
 et ,
1 + iss 1 + iss πss Yt Ht
30

Figure 8: Response of Output, Capital Stock, Real interest rate and Inflation rate to a 10 bps increase in
nominal interest rate
Output Capital stock

0.02 0
0
-0.02
-0.02
-0.04
-0.04
-0.06
-0.08 -0.06
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Real interest rate Inflation rate

2.6 0.05
2.55
0
2.5
2.45
-0.05
2.4 CEE/SW Exogenous Growth
CEE/SW Endogenous Growth
Endo Growth with hysteresis target
2.35 -0.1
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Notes: The figure plots the model-implied IRFs for output, capital stock, real interest rate and net inflation rate to a transitory shock to
the assumed Taylor rule. Solid line reports the IRFs for endogenous growth model with η = 0.10, and dashed line reports IRFs for the
comparable exogenous growth benchmark i.e. η = 0. Time is in quarters. IRFs are traced following a one-time exogenous shock in the
federal funds rate of about 10 basis points. The IRFs for output and capital stock are plotted in percent deviations from an exogenous
trend. For real interest rate, we plot the actual path of real interest rate. Net inflation rate is in percent deviation from steady state.
where Ht is hysteresis and follows the law motion given by:
f
Ht = Ht−1 + gt − gt .
We set φH = 0.2 to plot the graphs. Solid red line with crosses in Figure 8 plots the
evolution of output, capital stock, and interest rates under this hysteresis targeting rule.
The real interest rate increases by less with hysteresis targeting rule relative to a standard
Taylor rule (Equation 14). This is because, under a hysteresis targeting rule, the central
bank accommodates above-target inflation at a later time in order to target zero output
hysteresis. Expectations of a high inflation rate induce less contraction in the economy as
well as a sharp recovery of GDP to the pre-shock trend.18
18 Alternately,a higher weight on inflation stabilization in the policy rule can also reduce the permanent
effects of a given monetary shock. An advantage of our simple framework is that a strict inflation targeting
prescription of textbook models (Galı́, 2015b) is an optimal policy to deal with temporary demand shocks.
See Garga and Singh (2016) for detailed analysis from a micro-founded model of TFP growth.
31

8. Conclusion
This paper challenges the widely accepted benchmark of long-run money neutrality.
To test for a causal relationship in long-run data over more than 140 years in 17 advanced
economies we use a trilemma instrument: standard international macroeconomic theory
and recent evidence shows that exogenous monetary policy shocks in open pegs can be
isolated using unanticipated monetary policy shifts in base (anchor currency) countries.
We show the instrument is strong, and argue that it is valid, but we also account
for potential biases. To do this, we develop further the approach of local projections
and, in particular, we explain why LPs are preferable to VARs in settings like this where
long-horizon responses need to be estimated.
In our annual panel data, using the trilemma-identified monetary policy shocks, we do
indeed find evidence of powerful hysteresis forces. Monetary policy shocks have long-run
effects on output, capital, and TFP, over a horizon of more than a decade. Notably, the
capital stock is found to be permanently lower after a temporary contractionary monetary
policy shock identified using instrumental variables.
We reconcile these findings to theory using a simple extension of the Smets and Wouters
(2007) medium-scale DSGE model, where TFP growth depends on deviations of output
from its flexible-price counterpart. Using the model, we show that a central bank that also
targets hysteresis in its policy rule can offset the long-run effects of monetary policy. The
analysis of optimal policy in this environment is an important goal for future research.
32

References
Anzoategui, Diego, Diego Comin, Mark Gertler, and Joseba Martinez. 2019. Endogenous Technol-
ogy Adoption and R&D as Sources of Business Cycle Persistence. American Economic Journal:
Macroeconomics 11(3): 67–110.
Bai, Jushan, and Pierre Perron. 1998. Estimating and Testing Linear Models with Multiple Structural
Changes. Econometrica 66(1): 47–78.
Barlevy, Gadi. 2004. The cost of business cycles under endogenous growth. American Economic
Review 94(4): 964–990.
Barro, Robert J. 2013. Inflation and Economic Growth. Annals of Economics and Finance 14(1): 85–109.
Basu, Susanto, and John G Fernald. 2002. Aggregate productivity and aggregate technology. European
Economic Review 46(6): 963–991.
Benigno, Gianluca, and Luca Fornaro. 2018. Stagnation Traps. The Review of Economic Studies 85(3):
1425–1470.
Bergeaud, Antonin, Gilbert Cette, and Rémy Lecat. 2016. Productivity Trends in Advanced Countries
between 1890 and 2012. Review of Income and Wealth 62(3): 420–444.
Bernanke, Ben S., and Ilian Mihov. 1998. The liquidity effect and long-run neutrality. Carnegie-
Rochester Conference Series on Public Policy 49(1): 149–194.
Bianchi, Francesco, Howard Kung, and Gonzalo Morales. 2019. Growth, slowdowns, and recoveries.
Journal of Monetary Economics 101(C): 47–63.
Blanchard, Olivier. 2016. Currency Wars, Coordination, and Capital Controls. Working Paper 22388,
National Bureau of Economic Research.
Blanchard, Olivier. 2018. Should We Reject the Natural Rate Hypothesis? Journal of Economic
Perspectives 32(1): 97–120.
Blanchard, Olivier J., and Lawrence H. Summers. 1986. Hysteresis and the European unemployment
problem. In NBER Macroeconomics Annual 1986, Volume 1, 15–90. MIT Press.
Bühlmann, Peter. 1995. Moving-average representation of autoregressive approximations. Stochastic

Processes and their Applications 60(2): 331–342.
Cerra, Valerie, and Sweta Chaman Saxena. 2008. Growth dynamics: the myth of economic recovery.
American Economic Review 98(1): 439–457.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 1999. Monetary policy shocks:
What have we learned and to what end? In Handbook of Macroeconomics, edited by Taylor, J. B., and
M. Woodford, volume 1 of Handbook of Macroeconomics, chapter 2, 65–148. Amsterdam: Elsevier.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. Nominal Rigidities and
the Dynamic Effects of a Shock to Monetary Policy. Journal of Political Economy 113(1): 1–45.
Coibion, Olivier. 2012. Are the effects of monetary policy shocks big or small? American Economic
Journal: Macroeconomics 4(2): 1–32.
33

Coibion, Olivier, Yuriy Gorodnichenko, and Mauricio Ulate. 2017. The Cyclical Sensitivity in
Estimates of Potential Output. Working Paper 23580, National Bureau of Economic Research.
Conley, Timothy G., Christian B. Hansen, and Peter E. Rossi. 2012. Plausibly Exogenous. The Review
of Economics and Statistics 94(1): 260–272.
DeLong, Bradford J., and Lawrence H. Summers. 2012. Fiscal Policy in a Depressed Economy.
Brookings Papers on Economic Activity 1: 233–297.
Durbin, J. 1959. Efficient Estimation of Parameters in Moving-Average Models. Biometrika 46(3/4):

306–316.
Evans, Charles L. 1992. Productivity shocks and real business cycles. Journal of Monetary Economics
29(2): 191–208.
Farmer, Roger E. A., and Giovanni Nicolò. 2018. Keynesian economics without the Phillips curve.
Journal of Economic Dynamics and Control 89: 137–150.
Farmer, Roger E. A., and Konstantin Platonov. 2019. Animal spirits in a monetary model. European
Economic Review 115: 60–77.
Fatas, Antonio. 2000. Do business cycles cast long shadows? Short-run persistence and economic
growth. Journal of Economic Growth 5(2): 147–162.
Fatás, Antonio, and Lawrence H. Summers. 2018. The permanent effects of fiscal consolidations.
Journal of International Economics 112: 238–250.
Fernald, John G. 2007. Trend breaks, long-run restrictions, and contractionary technology improve-
ments. Journal of Monetary Economics 54(8): 2467–2485.
Fernald, John G. 2014. A quarterly, utilization-adjusted series on total factor productivity. Working
Paper Series 2012-19, Federal Reserve Bank of San Francisco.
Galı́, Jordi. 2015a. Hysteresis and the European unemployment problem revisited. Economics
Working Papers 1488, Department of Economics and Business, Universitat Pompeu Fabra.
Galı́, Jordi. 2015b. Monetary policy, inflation, and the business cycle: an introduction to the new Keynesian
framework and its applications. Princeton University Press.
Gali, Jordi, and Tommaso Monacelli. 2005. Monetary policy and exchange rate volatility in a small
open economy. The Review of Economic Studies 72(3): 707–734.
Garga, Vaishali, and Sanjay R. Singh. 2016. Output Hysteresis and Optimal Monetary Policy. Brown
University. Unpublished.
Geman, Stuart, and Chii-Ruey Hwang. 1982. Nonparametric Maximum Likelihood Estimation by
the Method of Sieves. The Annals of Statistics 10(2): 401–414.
Gonçalves, Sı́lvia, and Lutz Kilian. 2007. Asymptotic and Bootstrap Inference for AR(∞) Processes
with Conditional Heteroskedasticity. Econometric Reviews 26(6): 609–641.
Gopinath, Gita, Şebnem Kalemli-Özcan, Loukas Karabarbounis, and Carolina Villegas-Sanchez.

2017. Capital Allocation and Productivity in South Europe. Quarterly Journal of Economics 132(4):
1915–1967.
34

Gordon, Robert J. 2016. The Rise and Fall of American Growth: The U.S. Standard of Living since the Civil
War. Princeton, N.J.: Princeton University Press.
Gourinchas, Pierre-Olivier. 2018. Monetary Policy Transmission in Emerging Markets: An Appli-

cation to Chile. In Monetary Policy and Global Spillovers: Mechanisms, Effects and Policy Measures,
edited by Mendoza, Enrique G., Ernesto Pastén, and Diego Saravia, volume 25 of Central Banking,
Analysis, and Economic Policies Book Series, chapter 8, 279–324. Central Bank of Chile.
Grenander, Ulf. 1981. Abstract Inference. Wiley Inter-Science. New York, N.Y.: Wiley.
Hume, David. 1752. ‘Of Money’, ‘Of Interest’ and ‘Of the Balance of Trade’. In Essays Moral, Political,
and Literary, edited by Miller, Eugene F. Indianapolis, Ind.: Liberty Fund.
Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. 2019. Exchange Arrangements Entering
the Twenty-First Century: Which Anchor will Hold? Quarterly Journal of Economics 134(2): 599–646.
Imbs, Jean M. 1999. Technology, growth and the business cycle. Journal of Monetary Economics 44(1):
65–80.
Itskhoki, Oleg, and Dmitry Mukhin. 2019. Mussa puzzle redux. Princeton University. Unpublished.
Jiang, Zhengyang, Arvind Krishnamurthy, and Hanno Lustig. 2019. Dollar Safety and the Global
Financial Cycle. Research Paper 19-16, Stanford University Graduate School of Business.
Jordà, Òscar. 2005. Estimation and Inference of Impulse Responses by Local Projections. American
Jordà, Òscar, Moritz Schularick, and Alan M. Taylor. 2016. The great mortgaging: housing finance,
crises and business cycles. Economic Policy 31(85): 107–152.
Jordà, Òscar, Moritz Schularick, and Alan M. Taylor. 2017. Macrofinancial History and the New
Business Cycle Facts. NBER Macroeconomics Annual 31(1): 213–263.
Jordà, Òscar, Moritz Schularick, and Alan M. Taylor. 2019. The effects of quasi-random monetary
experiments. Journal of Monetary Economics. Forthcoming.
Justiniano, Alejandro, Giorgio E. Primiceri, and Andrea Tambalotti. 2013. Is there a trade-off between
inflation and output stabilization? American Economic Journal: Macroeconomics 5(2): 1–31.
King, Robert G., and Mark W. Watson. 1997. Testing Long-Run Neutrality. Federal Reserve Bank of
Richmond Economic Quarterly 83(3): 69–101.
Kuersteiner, Guido M. 2005. Automatic Inference For Infinite Order Vector Autoregressions. Econo-
metric Theory 21(1): 85–115.
Lewis, Richard, and Gregory C. Reinsel. 1985. Prediction of multivariate time series by autoregressive
model fitting. Journal of Multivariate Analysis 16(3): 393–411.
Lusompa, Amaze B. 2019. Local projections, autocorrelation, and efficiency. Unpublished. UC Irvine.
Mankiw, N. Gregory. 2001. The inexorable and mysterious tradeoff between inflation and unem-
ployment. Economic Journal 111(471): 45–61.
35

Moran, Patrick, and Albert Queraltó. 2018. Innovation, productivity, and monetary policy. Journal of
Monetary Economics .
Nakamura, Emi, and Jón Steinsson. 2018. Identification in macroeconomics. Journal of Economic
Perspectives 32(3): 59–86.
Obstfeld, Maurice, and Kenneth Rogoff. 1995. Exchange rate dynamics redux. Journal of Political
Economy 103(3): 624–660.
Obstfeld, Maurice, Jay C. Shambaugh, and Alan M. Taylor. 2004. Monetary sovereignty, exchange
rates, and capital controls: The trilemma in the interwar period. IMF Staff Papers 51(3S1): 75–108.
Obstfeld, Maurice, Jay C. Shambaugh, and Alan M. Taylor. 2005. The Trilemma in History: Tradeoffs
Among Exchange Rates, Monetary Policies, and Capital Mobility. Review of Economics and Statistics
87(3): 423–438.
Plagborg-Møller, Mikkel, and Christian K. Wolf. 2018. Instrumental Variable Identification of

Dynamic Variance Decompositions. Princeton University. Unpublished.
Quinn, Dennis, Martin Schindler, and A Maria Toyoda. 2011. Assessing measures of financial
openness and integration. IMF Economic Review 59(3): 488–522.
Ramey, Valerie A. 2016. Macroeconomic Shocks and Their Propagation. In Handbook of Macroeco-
nomics, edited by Taylor, John B., and Harald Uhlig, volume 2 of Handbook of Macroeconomics,
chapter 0, 71–162. Amsterdam: Elsevier.
Ramey, Valerie A., and Sarah Zubairy. 2018. Government Spending Multipliers in Good Times and
in Bad: Evidence from US Historical Data. Journal of Political Economy 126(2): 850–901.
Reifschneider, Dave, William Wascher, and David Wilcox. 2015. Aggregate supply in the United
States: recent developments and implications for the conduct of monetary policy. IMF Economic
Review 63(1): 71–109.
Reinhart, Carmen M., and Kenneth S. Rogoff. 2009. The Aftermath of Financial Crises. American
Rey, Hélène. 2015. Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy
Independence. NBER Working Papers 21162, National Bureau of Economic Research, Inc.
Romer, Christina D., and David H. Romer. 2004. A New Measure of Monetary Shocks: Derivation
and Implications. American Economic Review 94(4): 1055–1084.
Schularick, Moritz, and Alan M. Taylor. 2012. Credit Booms Gone Bust: Monetary Policy, Leverage
Cycles, and Financial Crises, 1870–2008. American Economic Review 102(2): 1029–1061.
Shambaugh, Jay C. 2004. The Effect of Fixed Exchange Rates on Monetary Policy. The Quarterly
Journal of Economics 119(1): 301–352.
Smets, Frank, and Rafael Wouters. 2007. Shocks and frictions in US business cycles: A Bayesian
DSGE approach. American Economic Review 97(3): 586–606.
Stadler, George W. 1990. Business Cycle Models with Endogenous Technology. American Economic
Review 80(4): 763–78.
36

Stock, James H., and Mark W. Watson. 2018. Identification and Estimation of Dynamic Causal Effects
in Macroeconomics Using External Instruments. Economic Journal 128(610): 917–948.
Syverson, Chad. 2011. What determines productivity? Journal of Economic literature 49(2): 326–65.
van Kippersluis, Hans, and Cornelius A. Rietveld. 2018. Beyond plausibly exogenous. The Economet-
rics Journal 21(3): 316–331.
Wieland, Johannes F., and Mu-Jeung Yang. 2016. Financial Dampening. NBER Working Papers
22141, National Bureau of Economic Research, Inc.
Wold, Herman. 1938. A study in the analysis of stationary time series. Uppsala, Sweden: Almqvist &
Wiksell.
Yellen, Janet L. 2016. Macroeconomic Research After the Crisis. Speech at the conference “The
Elusive ‘Great’ Recovery: Causes and Implications for Future Business Cycle Dynamics,” Federal
Reserve Bank of Boston, October.
37

Appendix
A. Proof of consistency for impulse responses from MA(∞) DGP
This section provides the basic ideas behind the proofs of consistency for truncated VARs and LPs
when the true DGP is MA(∞). The reader is referred to the references cited for more details. Here
we sketch the main intuition.
A.1. Data generating process and main assumptions

Assume the data generating process for the m–dimensional vector process yt is as described in
section 4, namely
∞ ∞
yt = ∑ B j t − j ; B0 = I; ∑ || Bj || < ∞ ,
j =0 j =0
where || Bj ||2 = tr ( B0j Bj ) and B(z) = ∑∞

j=0 B j z j such that det{ B ( z )} 6 = 0 for | z | ≤ 1. Under these
assumptions:
∞ ∞
yt = ∑ A j yt− j + t ; ∑ || A j || < ∞; det{ A(z)} 6= 0 for |z| ≤ 1 .
j =1 j =1
Further, we make assumptions 1–4 following Lewis and Reinsel (1985), and Lusompa (2019) (Kuer-
steiner (2005) makes somewhat stronger assumptions because he later derives testing procedures to
determine the optimal lag length). These assumptions are:
Assumption 1 {yt } is generated by Equation 1.
Assumption 2 E|eit e jt ekt elt | ≤ γ4 < ∞ for 1 ≤ i, j, k, l ≤ m.
Assumption 3 The truncation lag p is chosen as a function of the sample size T such that
p2 /T → 0 as p, T → ∞.
Assumption 4 p is chosen as a function of T such that

∞
p1/2 ∑ || A j || → 0 as p, T → ∞.
j = p +1
Then, as discussed in the text, Lewis and Reinsel (1985) show:

p
|| Â j − A j || −
→0 as p, T → ∞.
A.2. Potential sources of bias in truncated VARs

In finite samples, inconsistent estimates of the impulse response function can arise from at least two
sources that can be quantified: (1) the truncation lag is too short given Assumptions 1–4; and (2)
A1

even if the truncation lag is appropriate, the impulse response is calculated for periods beyond the
truncation lag. To investigate the first source of bias, rewrite the VAR(∞) as
k
yt = ∑ A j yt − j + ut ,
j =1
p ∞
ut = ∑ A j yt − j + ∑ A j yt − j + t ,
j = k +1 j = p +1
where we assume k < p and p is the truncation lag that meets Assumptions 1–4 of the proof of
consistency. Hence rewrite the previous expression as
yt = A(k ) Xk,t−1 + ut ; A ( k ) = ( A1 , . . . , A k ); Xk,t−1 = (yt−1 , . . . , yt−k )0 .
The least-squares estimate of A(k ) is therefore

! ! −1
T T
1 1
Â(k ) =
T−k ∑ yt Xk,t
0
−1
T−k ∑ Xk,t−1 Xk,t
0
−1 .
p p
Hence
! ! −1
T T
1 1
Â(k ) = A(k ) +
T−k ∑ 0
ut Xk,t −1
T−k ∑ 0
Xk,t−1 Xk,t −1 .
p p
Given the three components of ut , it is easy to see that the source of inconsistency will come from
the term
! ! −1
T p T
1 1
T−k ∑ ∑ A j yt− j Xk,t ∑ Xk,t−1 Xk,t
0 0
−1 −1 ,
p j = k +1 T − k p
since the proof of consistency in Lewis and Reinsel (1985) shows that the other two terms vanish
asymptotically. The source of inconsistency can be quantified by noticing that
 −1
Γ (0) Γ (1) Γ(k)

! −1 ···
1 T  Γ (1) Γ (0) ··· Γ ( k − 1) 
∑ Xk,t−1 Xk,t
0
→  ..  → Γ− 1
k ,
 
−1 .. .. ..
T−k p  . . . . 
Γ ( k ) Γ ( k − 1) · · · Γ (0)
as shown in Lewis and Reinsel (1985), where E(yt yt0− j ) = Γ( j) and Γ(− j) = Γ( j)0 . Hence, asymptot-
ically, the source of inconsistency is
p
∑ A j (Γ( j − 1), . . . , Γ( j − k)) Γ−k 1 .
k +1
However, even when the lag-length p is chosen to be sufficiently large, another source of bias can
crop up into the estimation of the impulse response. In particular, following Durbin (1959), we know
A2

that
Bh = A1 Bh−1 + A2 Bh−2 + · · · + Ah−1 B1 + Ah .
If the VAR is truncated at lag k, for k ≤ p, it is easy to see that the previous expression becomes
Bh = A1 Bh−1 + A2 Bh−2 + · · · + Ak Bk−h + Ak+1 Bk−h−1 + · · · + Ah−1 B1 + Ah .

| {z }
BI AS
That is, even if k = p, and hence || Â j || → A j for j = 1, . . . , k, the fact remains that the omitted term
Ak+1 Bk−h−1 + · · · + Ah−1 B1 + Ah
is still unaccounted for.

(Of course, when k < p then one has to add to this problem the bias generated from having too
short a VAR relative to the sample, thus potentially compounding the bias problem.)
In practice, it seems advisable to choose relatively long-lags to accurately describe the impulse
response. Note that there is a variance-bias trade-off. We are not claiming a lower RMSE, but
rather, since interest is in uncovering the behavior of the impulse response at long horizons, we are
interested in trading precision for lower bias.
A.3. The consistency of the local projections estimator

In this section we use the same assumptions as in the previous section to establish the consistency
of the local projections estimator at any horizon.
Using the VAR(∞) representation of the DGP and recursive substitution, it is easy to see that
yt+h = Bh+1 yt−1 + {Ch+2 yt−2 + Ch+3 yt−3 + · · · } + et+h + B1 t+h−1 + t Bh ,
where
Ch+2 = Bh A1 + · · · + B1 Ah + Ah+1 ,
Ch+3 = Bh A2 + · · · + B1 Ah+1 + Ah+2 ,
..
.
Ch+k = Bh Ak−1 + · · · + B1 Ah+k−2 + Ah+k−1 .
Now, consider truncating the lag of the local projection at k = p, where p meets Assumptions 1–4 of
the Lewis and Reinsel (1985) consistency theorem discussed in the previous section.
Then the truncated local projection can be written as
y t + h = B h + 1 y t − 1 + C h + 2 y t − 2 + Ch + 3 y t − 3 · · · + C h + k y t − k + u t + h ,
ut+h = t+h + { B1 t+h−1 + B2 t+h−2 + · · · + Bh t } + {Ch+k+1 yt−k−1 + Ch+k+2 yt−k−2 + · · · } .
Define D = ( Bh , Ch+2 , . . . , Ch+k ) and Xt−1 = (yt−1 , . . . , yt−k )0 as defined earlier but where the
subscript k is omitted here for simplicity. Then the local projection can be compactly written as
yt+h = DXt−1 + ut+h .
A3

The least-squares estimate of D is simply
! ! −1
T −h T −h
1 1
D̂ =
T−h−k ∑ yt+h Xt0−1
T−h−k ∑ Xt−1 Xt0−1 ,
k k
from where consistency can be determined from the following expression

! ! −1
T −h T −h
1 1
D̂ = D +
T−h−k ∑ ut+h Xt0−1 T−h−k ∑ Xt−1 Xt0−1 .
k k
Lewis and Reinsel (1985) show that ||Γ− 1

k ||1 is uniformly bounded where we use the fact that
|| AB|| ≤ || A||1 || B|| ; as well as || AB|| ≤ || A||2 || B||21 where ||C ||21 = supl 6=0 l 0 C 0 Cl/l 0 l, the largest
2 2 2 2
eigenvalue of C 0 C (see Wiener and Masani, 1958).

Now we turn our focus to the terms
T −h T −h
1 1
T−h−k ∑ ut+h Xt0−1 =
T−h−k ∑ (t+h + B1 t+h−1 + · · · + Bh t )Xt0−1
k k
T −h
1
=
T−h−k ∑ (Ch+k yt−k−1 + Ch+k+1 yt−k−2 + · · · )Xt0−1 .
k
It is easy to see that
T −h
1
T−h−k ∑ t+h Xt0−1 → 0 ,
k
Bj T −h
T−h−k ∑ t+h− j Xt0−1 → 0 ,

k
since || Bj || < ∞ for j = 1, · · · , h. Hence, the only tricky part is to examine the terms
Ch + k + j T −h
T−h−k ∑ yt−k−( j+1) Xt0−1 for j = 0, 1, . . . .

k
Note that
Ch+k+ j = Bh Ak+k + · · · + B1 Ah+k+ j−1 + Ah+k+ j j = 0, 1, . . . ,
hence
∞ ∞
∑ ||Ch+k+ j || = ∑ || Bh Ak+k + · · · + B1 Ah+k+ j−1 + Ah+k+ j ||
j =0 j =0
∞ ∞ ∞
≤ ∑ || Bh Ak+ j || + · · · + ∑ || B1 Ah+k+ j−1 || + ∑ || Ah+ j+ j ||
j =0 j =1 j =0
∞ ∞ ∞
= || Bh ||1 ∑ || Ak+ j || + · · · + || B1 ||1 ∑ || Ah+k+ j−1 || + ∑ || Ah+k+ j ||.
j =0 j =1 j =0
A4

From the assumptions we know that the || Bj ||1 are uniformly bounded, and also that
∞ ∞
k1/2 ∑ || Ak+ j || → 0 =⇒ k1/2 ∑ ||Ch+k+ j || → 0 ,
j =0 j =0
and this condition can now be used to show that
∑∞
j = 0 Ch + k + j
T −h
T−h−k ∑ yt−k−( j+1) Xt0−1 → 0, as k, T → ∞ .

k
Summarizing, these derivations show that the same conditions that ensure consistency of the
coefficients estimates in a truncated VAR also ensure consistency of the local projections with
truncated lag length. However, because the coefficient for yt−1 in the local projection is a direct
estimate of the impulse response coefficient, then we directly get a proof of consistency for the
coefficients of the impulse response at any horizon regardless of truncation.
B. Imbs correction
The Imbs (1999) correction follows Burnside and Eichenbaum (1996) in endogenizing the capital
utilization rate in a partial equilibrium model. We assume perfectly competitive factor markets and
a technology which is constant returns to scale in effective capital and labor.19
In aggregate, and for the representative firm, the production function is
Yt = At (Kt ut )α ( Lt et )1−α ,
where Yt is output, Kt is capital stock, Lt is total hours worked, and ut and et denote the respective
factor utilizations. At is the utilization adjusted TFP. We assume perfect competition in the input
φ
and the output markets. Higher capital utilization increases the depreciation of capital δt = δut
where φ > 1. As a result, firms choose capital utilization rate optimally. Labor hoarding is calculated
assuming instantaneous adjustment of effort et against a payment of a higher wage w(et ), while
keeping fixed employment (determined one period in advance). The firm’s optimization problem is
given by:
max At (Kt ut )α ( Lt et )1−α − w(et ) Lt − (rt + δut ))Kt .
φ
et ,ut ,Kt
Households choose consumption, labor supply and effort to maximize their lifetime utility subject
to their budget constraint (with complete asset markets)
∞
" #
( e t L t )1+ ν
max ∑ β ln Ct − χ
t
.
ct ,Lt ,et t=0 1+ν
Normalizing the long-run capital-utilization and labor-utilization rates to one, the utilization rates
can be derived from
Yt
! r+δ δ Yt
! 1+1 ν
Kt Ct L
ut = Y ; et = Y ;
K C
L t
where Y, C, L and K are the steady-state values of output, consumption, labor, and capital.
19 We assume that these variables are stationary. In section 7, we construct capital-utilization in a general
equilibrium model along with endogenous growth.
A5

The Solow residual then can be decomposed into utilization-adjusted TFP and utilization
corrections, with
Yt
TFPt ≡ α 1−α = At × uαt e1t −α .
Kt L t
To construct steady state values of Y, C, L, and K, we extract a HP-filter trend from the data series.
We will show later that our empirical results are robust to computing moving averages over a 10
year window, using time-varying values of α constructed from labor-income data, and reasonable
parameters of the aggregate capital depreciation rate. Bergeaud, Cette, and Lecat (2016) constructed
capital stock for machines and buildings separately using the perpetual inventory method with data
on investment in machines and buildings and different depreciation rates. We will show robustness
of our results to choosing different depreciation parameters.
We present the responses of the various components of the Solow residual in Figure A9 for both
the full and post-WW2 samples. Utilization rates of labor and capital exhibit cyclical dynamics:
falling in the short-run and then recovering back to zero. In a sense, this pattern is mechanical. It
reflects the restriction imposed on the calculation that utilization rates must return to zero eventually
(Imbs, 1999).
Figure A10 presents the evolution of capital to utilization adjusted TFP ratio in response to the
trilemma shock. The hump-shaped response replicates the conventional justification for including
investment adjustment costs in typical DSGE models (Christiano, Eichenbaum, and Evans, 2005).20
C. A dynamic open economy model: Implications for the trilemma

We now write down a currency union model of a small open economy based on Gali and Monacelli
(2005). There is a continuum of measure zero small open economies.21 A representative household
in an economy has preferences over a consumption aggregator and derives disutility from supplying
labor. Final consumption is a CES aggregator over domestic and foreign goods with home bias
measured by 1 − α ∈ [0, 1]. The foreign good is a Dixit-Stiglitz aggregator of goods supplied by
a continuum of foreign economies with elasticity of substitution among varieties of countries is
given by γ > 0. η > 0 measures substitutability between domestic and foreign goods. And e > 0
measures elasticity of substitution among varieties produced by one country. Each differentiated
product within a country is produced by a monopolistically competitive producer that takes demand
as given and sets prices in a staggered fashion ala Calvo price setting. Since we will consider fixed
nominal exchange rate, producer or local currency pricing assumptions are equivalent. We assume
complete financial markets. We denote the world variables with asterisk. The home economy is too
small to influence any world economy variables.
Let Ŷt be home output, Ĉt be home consumption, π̂tH be domestic producer price inflation, ît be
home nominal interest rate, and Ŝt effective terms of trade (price of foreign goods in terms of home
goods) and Êt is the nominal exchange rate. Following the seminal work of Gali and Monacelli
(2005), we can derive the equilibrium as system of these six variables {Ŷt , Ĉt , π̂tH , ît , Ŝt , Êt } that satisfy
the following five equations, plus a sixth equation corresponding to the domestic policy rule, and
given foreign output and inflation {Yt∗ , π̂t∗ }, domestic productivity shocks êt , markup shocks µ̂t ,
20 We later show the robustness of using a more commonly used utilization adjusted TFP series constructed
by Fernald (2014) for the post-WW2 US, which requires more detailed data than is available for the historical
panel of 17 countries. Reassuringly, the results for utilization adjusted TFP using the two methods are
qualitatively similar.
21 De Paoli (2009)’s model treats foreign as one large economy. Similar results are obtained in that treatment.
A6

government spending shocks ĝt , and natural interest rates r̂tn :
1
Ŷt = Et Ŷt+1 − (ît − Et π̂tH+1 − r̂tn ) , (AD)
σα
π̂tH = βEt π̂tH+1 + κ ((σα + ϕ)Ŷt + (σ − σα )Ŷt∗ − (1 + ϕ)êt + µ̂t ) , (AS)
αω
Ŷt = Ĉt + Ŝt + ĝt , (RC)
σ
1−α
Ĉt = Ŷt∗ + Ŝt , (RS)
σ
∆Ŝt = ∆ Êt + π̂t∗ − π̂tH , (TT)
(1− βθ )(1−θ )
where κ = θ , where κ (1 + σϕ) is slope of the Phillips curve. It is microfounded from a
Calvo assumption where 1 − θ is the probability of firm price reset (and θ → 0 implies κ → ∞).
Here, σ−1 is the intertemporal elasticity of substitution, e is intratemporal elasticity of substituion,
ϕ−1 is Frisch elasticity of labor supply, and, σα = (1−ασ)+αω , ω = σγ + (1 − α)(ση − 1), and
Θ = (σγ − 1) + (1 − α)(ση − 1) = ω − 1. When σ = η = γ = 1, we have ω = 1. When
α → 0, we have σα → σ. In the closed economy limit α → 0, the eqn (AD) collapses to the
AD equation in a closed economy NK model (Galı́, 2015). The natural rate of interest is given
by r̂tn = σα (1 + αΘ)Et {∆ ĝt+1 } + σα αΘEt {∆Ŷt∗+1 }. The domestic component of the natural rate of
interest is given by σα (1 + αΘ)Et {∆ ĝt+1 }, and the foreign component to the natural rate of interest
σα αΘEt {∆Ŷt∗+1 } is driven by changes in foreign output growth.
A hard peg policy ∆ Êt = 0 in the presence of partially flexible prices introduces dynamics in real
exchange rate. As shown by Benigno (2004); Benigno, Benigno, and Ghironi (2007); Corsetti, Kuester,
and Müller (2013), the past level of the terms of trade (and hence inflation rate) is a state-variable
under fixed nominal exchange rate. One can analytically solve for law of motion of terms of trade
and obtain the following proposition.22
Proposition. In the Cole-Obstfeld case (log utility and unit trade elasticity, i.e., σ = η = γ = 1), a monetary
policy shock in world economy identifies an exogenous monetary policy shock in a small open economy that
pegs its currency to the union.
Under the Cole-Obstfeld assumption, there is no movement in net exports of a country. Further-
more, there are no terms of trade movements in response to a monetary policy shock in foreign
economy in this case. As a result, a monetary policy shock in the foreign economy acts like a
monetary policy shock in the home economy. The spillover effects of trade cancel out, and we can
identify a monetary multiplier as in the Mundell-Fleming model presented in the main text.
Using calibrated parameters of Gali and Monacelli (2005), we simulate the response of home
and foreign economy to a monetary policy shock in the foreign economy in Figure A11. The foreign
large economy is modeled as a closed economy in a textbook manner (Galı́, 2015). A contractionary
shock to the Taylor rule in foreign economy increases foreign real interest rate. As a result, foreign
output and inflation fall. Since the home economy follows a hard peg with foreign, the nominal
interest rate at home follows identical path as in the foreign economy. There are no changes in
terms of trade or net exports. The IRFs of home output and inflation are exactly like that of closed
economy’s response to a domestic monetary policy shock. The only difference is that there is no
systematic response of home’s central bank and it maintains a fixed nominal exchange rate.
22 See also Farhi and Werning (2012); Nakamura and Steinsson (2014); Itskhoki and Mukhin (2019).
A7

D. Additional figures
D.1. Trilemma graphs for GDP per capita
Figure A1: Baseline response to 100 bps trilemma shock: Real GDP per capita
Full Sample: 1890-2015 Post-WW2 Sample: 1948-2015

1
1
-1
-1
Percent
Percent
-3
-3
-5
-5
-7
-7
0 4 8 12 0 4 8 12
Year Year
sample: 1890–2015 (World Wars excluded). Post-WW2 sample: 1948–2015.LP-IV estimates displayed as a solid blue line and and 1 S.D.
and 2 S.D. confidence bands constructed using cluster-robust standard errors. See text.
D.2. Trilemma graphs for Post WW2 sample
Figure A2: Baseline response to 100 bps trilemma shock: Real GDP and components (Post ww2 sample)
(a) (b)
IRFs to a 100 bps trilemma shocks: 1948- 2015
real GDP total hours

0
0
0
Percent
Percent
-2
-2
-4
-4
-2
-6
-6
Percent
0 4 8 12 0 4 8 12
Year Year
capital stock TFP
-4
0
Percent
Percent
-2
-2
-6
-4
-4
0 4 8 12
Year
-6
-6
GDP TFP K L 0 4 8 12 0 4 8 12
Year Year
Notes: Response to a 100 bps shock in domestic interest rate instrumented with the trilemma. Responses for pegging economies.
Post-WW2 sample: 1948–2015. See text.
A8

Figure A3: Baseline response to 100 bps trilemma shock: a) Short term nominal interest rate, b) real interest
rate, and c) real interest rate multiplier
(a) (b) (c)
Short-term nominal interest rate Short term real interest rate Real interest rate multiplier
3
1.5
0
1
-.2
.5
Percent
1
0
-.4
0
-.5
-.6
-1
-1
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Notes: Response to a 100 bps shock in domestic interest rate instrumented with the trilemma. a) Responses of short term nominal interest
rate for pegging economies. b) Responses of short term real interest rate for pegging economies. Inflation expectations constructed
from the impulse response of consumer price level index to the same trilemma shock. c) Response of cumulative change in real GDP
divided by cumulative change in short term nominal interest rate for pegging economies. Sample: 1948–2015. See text.
D.3. Trilemma graphs: controls in levels, differences, and number of lags

We report the robustness of IRFs estimated in the baseline to adding the control variables x j,t in
levels instead of first differences in the left panel of Figure A4, as well as to including up to 5 lags of
the control variables in the right panel.
Figure A4: Response to 100 bps trilemma shock with controls in levels: Real GDP
Real GDP: controls in levels Real GDP: controls in differences

2
2
0
0
-2
-2
Percent
Percent
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
IV IV
sample: 1890–2015 (World Wars excluded). LP-OLS estimates displayed as a dashed red line, LP-IV estimates displayed as a solid blue
line and 1 S.D. and 2 S.D. confidence bands. See text and Jordà, Schularick, and Taylor (2019).
A9

D.4. Trilemma graphs for various macro-financial variables
Figure A5: Baseline response to 100 bps trilemma shock: Full Sample (1890-2015)
Real GDP per capita Real consumption per capita Real investment per capita
2
0
0
-2
-2
Percent
Percent
Percent
-4 -2
-4
-4
-6
-6
-8
-6
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Price level Short-term interest rate Long-term interest rate
1 1.5
1
0
Percentage pts
Percentage pts
.5
-5
Percent
.5
0
-10
-.5
-.5
-15
-1
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Real house prices Real stock prices Private credit/GDP
20
5
42
10
0
Percent
Percent
Percent
0
-5
-2
-10
-10
-4
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
sample: 1890–2015 (World Wars excluded). Confidence bands are one and two standard deviations using cluster-robust standard errors.
See text.
A10

Figure A6: Baseline response to 100 bps trilemma shock: Post-WW2 Sample (1948-2015)
Real GDP per capita Real consumption per capita Real investment per capita
2
0
0
-2
Percent
Percent
Percent
-2
-4 -2
-4
-4
-6
-6
-6
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Price level Short-term interest rate Long-term interest rate
1 1.5
1
0
Percentage pts
Percentage pts
-10 -8 -6 -4 -2
.5
Percent
.5
0
0
-.5
-.5
-1
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Real house prices Real stock prices Private credit/GDP
20
4
2
-8 -6 -4 -2 0
10
2
Percent
Percent
Percent
0
0
-10
-2
0 4 8 12 0 4 8 12 0 4 8 12
Year Year Year
Notes: Response to a 100 bps shock in domestic interest rate instrumented with the trilemma. Responses for pegging economies.
PostWW2 sample: 1948–2015 (World Wars excluded). Confidence bands are one and two standard deviations using cluster-robust
standard errors. See text.
A11

D.5. Trilemma graphs with structural breaks
Figure A7: Response to 100 bps trilemma shock with structural breaks: Real GDP
(a) (b)
5 breaks in TFP 5 breaks in GDP
Real GDP Real GDP

2
2
0
0
-2
-2
Percent
Percent
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
We next show an alternate selection of structural breaks where we allow agents to anticipate
structural breaks at upto 12 years ahead starting at time 0.
Figure A8: Response to 100 bps trilemma shock with anticipated structural breaks: Real GDP
5 breaks in TFP and expected TFP 5 breaks in current and expected GDP
Real GDP Real GDP

2
2
0
0
-2
-2
Percent
Percent
-4
-4
-6
-6
0 4 8 12 0 4 8 12
Year Year
A12

D.6. Trilemma graphs for TFP and utilization adjusted TFP
Figure A9: Baseline response to 100 bps trilemma shock: TFP and utilization
(a) Full sample (b) Post-WW2

1
1
0
0
Percent
Percent
-1
-1
-2
-2
-3
-3
0 4 8 12 0 4 8 12
Year Year
Raw TFP Util adj TFP K utilization L utilization
sample: 1890–2015 (World Wars excluded). Post-WW2 sample: 1948–2015. See text.
Figure A10: Baseline response to 100 bps trilemma shock: Capital to utilization adjusted TFP
(a) Full sample (b) Post-WW2
K/Util adj TFP K/Util adj TFP

2
1
0
0
Percent
Percent
-2
-1
-4
-2
-6
-3
0 4 8 12 0 4 8 12
Year Year
sample: 1890–2015 (World Wars excluded). Post-WW2 sample: 1948–2015. See text.
A13

D.7. Structural break dates in TFP growth and GDP growth
10 TFP growth for AUS TFP growth for BEL
40
20
0
0
-10
-20
-20
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for CAN TFP growth for CHE

20
30
20
10
10
0
0
-10
-10
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for DEU TFP growth for DNK

50
20
10
0
0
-50
-10
-100
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for ESP TFP growth for FIN

20
10
10
0
-10
0
-10
-20
-30
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
A14

TFP growth for FRA TFP growth for GBR
40
5
20
0
0
-5
-20
-40
-10
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for ITA TFP growth for JPN
20
30
0
20
-20
10
-40
0
-10
-60
-20
-80
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for NLD TFP growth for NOR

60
15
10
40
20
5
0
0
-5
-20
-10
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
TFP growth for PRT TFP growth for SWE

15
15
10
10
5
5
0
0
-5
-5
-10
-10
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
A15

TFP growth for USA
15
10
5
0
-5
-10
1890 1915 1940 1965 1990 2015

Year
GDP growth for USA

20
10
0
-10
-20
1890 1915 1940 1965 1990 2015

Year
A16

GDP growth for AUS GDP growth for BEL
20
40
10
20
0
0
-10
-20
-20
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for CAN GDP growth for CHE

20
15
10
10
5
0
0
-10
-5
-10
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for DEU GDP growth for DNK

50
20
10
0
0
-50
-10
-100
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for ESP GDP growth for FIN

20
10
10
0
0
-10
-10
-20
-30
-20
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
A17

GDP growth for FRA GDP growth for GBR
40
10
20
5
0
0
-5
-20
-10
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for ITA GDP growth for JPN
20
20
10
0
0
-10
-20
-20
-30
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for NLD GDP growth for NOR

60
15
40
10
20
5
0
0
-20
-5
-10
-40
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
GDP growth for PRT GDP growth for SWE

10
15
10
5
0
5
-5
0
-10
-5
-10
-15
1890 1915 1940 1965 1990 2015 1890 1915 1940 1965 1990 2015
Year Year
A18

D.8. Graph for Gali and Monacelli (2005) open economy model
Figure A11: Response to monetary policy shock in the base country
Home Inflation PPI Home Output Home Inflation CPI

0 0 0
-0.1
-0.5 -0.5
% dev from ss
-0.2
-1 -0.3 -1
-0.4
-1.5 -1.5
-0.5
-2 -0.6 -2
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Home Nom int Rate 10 -15Home terms of trade Nominal exchange rate
0 3 1
2 0.5
% dev from ss
-0.5
1 0
-1
0 -0.5
-1.5 -1 -1
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Foreign Output Foreign Inflation Foreign nom int rate

0 0 0
-0.1
-0.5
% dev from ss
-0.2 -0.5
-0.3 -1
-0.4 -1
-1.5
-0.5
-0.6 -2 -1.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Notes: Impulse response to a monetary policy shock in foreign economy using Gali and Monacelli (2005) parameters.
A19

D.9. Graphs for Romer and Romer (2004): Disaggregation of GDP
Figure A12: Baseline response to 100 bps Romer and Romer (2004) shock: U.S. postwar data
Instrument: Romer & Romer (2004) residuals
.5
0
Percent
-.5
-1
0 4 8 12 16 20 24 28 32
Quarter
GDP Util adj TFP K L LQ
Notes: Response to a 100 bps shock in overnight federal funds rate rate instrumented with policy forecast residuals (Romer and Romer,
2004; Wieland and Yang, 2016). Responses of real GDP, capital stock, total hours worked, labor composition/quality, and raw TFP for
U.S. Quarterly sample: 1969-Q1: 2007-Q4. Quarterly data series are taken from Fernald (2014). See text.
A20

D.10. Graphs for Romer and Romer (2004) Surprises: Different sample
selection for US
Figure A13: Response to 100 bps Romer and Romer (2004) shock
(a) Full sample: 1969Q2: 2008Q3 (b) Sample: 1973Q2: 2008Q3
Real GDP (Util. Adj.) TFP index (Fernald 2014) Real GDP (Util. Adj.) TFP index (Fernald 2014)
.5
1
1
.5
1
0
0
Percent
Percent
Percent
Percent
0
0
-1
-.5
-.5
-1
-1
-2
-2
-1
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
Quarter Quarter Quarter Quarter
IV OLS IV OLS
Hours, bus sector Capital input Hours, bus sector Capital input
.5
1
0
-.5
0
0
Percent
Percent
Percent
Percent
-.5
-1
-1
-1.5
-1
-1
-2
-1.5
-2
-2
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
IV OLS IV OLS
economy. Quarterly samples: a) 1969-Q2: 2008-Q3, b) 1973-Q2: 2008-Q3. Quarterly data series are taken from Fernald (2014).
A21

(a) Sample: 1979Q3: 2008Q3 (b) Sample: 1984Q1: 2008Q3
2
1
1
0
Percent
Percent
Percent
Percent
0
-1
-1
-1
-2
-2
-2
-2
-4
-3
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
IV OLS IV OLS
3
4
1
2
2
.5
2
1
1
0
Percent
Percent
Percent
Percent
0
0
-.5
0
-2
-1
-1
-1
-1.5
-2
-4
-2
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
IV OLS IV OLS
A22

(a) Sample: 1969Q2: 2002Q4 (b) Sample: 1987Q1: 2008Q3
4
2
2
2
1
.5
0
0
0
Percent
Percent
Percent
Percent
0
-2
-1
-2
-.5
-4
-2
-1
-6
-3
-4
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
IV OLS IV OLS
2
2
.5
1
0
Percent
Percent
Percent
Percent
-.5
0
0
-2
-1
-1
-1.5
-4
-2
-5
0 8 16 24 32 0 8 16 24 32 0 8 16 24 32 0 8 16 24 32
IV OLS IV OLS
A23

E. Additional tables
Table A1: LP-OLS vs. LP-IV. Attenuation bias of real GDP per capita responses to interest rates.
Trilemma instrument. Matched samples
Responses of real GDP per capita at years 0 to 10 (100 × log change from year 0 baseline).
(a) Full Sample OLS-IV (b) Post-WW2 OLS-IV
Year LP-OLS LP-IV p-value LP-OLS LP-IV p-value
(1) (2) (3) (4) (5) (6)
h=0 0.07∗∗ -0.01 0.35 0.03 0.11∗ 0.22
(0.03) (0.08) (0.02) (0.07)
h=2 -0.33∗∗ -1.55∗∗∗ 0.00 -0.32∗∗ -1.40∗∗∗ 0.00

(0.15) (0.29) (0.13) (0.30)
h=4 -0.22 -2.07∗∗∗ 0.00 -0.29 -1.64∗∗∗ 0.00

(0.24) (0.40) (0.22) (0.41)
h=6 -0.10 -2.76∗∗∗ 0.00 -0.19 -2.93∗∗∗ 0.00

(0.36) (0.57) (0.30) (0.59)
h=8 -0.30 -4.71∗∗∗ 0.00 -0.01 -3.13∗∗∗ 0.00

(0.34) (0.98) (0.33) (0.67)
h = 10 0.10 -4.27∗∗∗ 0.00 0.21 -3.03∗∗∗ 0.00

(0.34) (0.96) (0.37) (0.75)
h = 12 0.21 -4.02∗∗∗ 0.00 0.37 -3.72∗∗∗ 0.00

(0.35) (0.96) (0.39) (0.94)
KP weak IV 44.40 54.07
H0 : LATE = 0 0.00 0.00 0.01 0.00
Observations 838 765 658 585
Notes: ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Cluster robust standard errors in parentheses.Full sample: 1890 − 2015 excluding
WW1: 1914 − 1919 and WW2: 1939 − 1947. Post WW2 sample: 1948 − 2015. Matched sample indicates LP-OLS sample
matches the sample used to obtain LP-IV estimates. KP weak IV refers to the Kleibergen-Paap test for weak instruments.
H0 : LATE = 0 refers to the p-value of the test of the null hypothesis that the coefficients for h = 0, ..., 10 are jointly zero for a
given subpopulation. OLS = IV shows the p-value for the Hausmann test of the null that OLS estimates equal IV estimates.
See text.
A24

F. DSGE model
We present the different components of the model in the next few sections, starting with the
production side of the economy.
F.1. Monopolistically competitive producers

Assume there is a continuum of differentiated intermediated good producers that sell the inter-
mediate good Ykt . These goods can be aggregated into a Dixit-Stiglitz final composite good Yt as
follows: Z 1 1 1+ λ p
1+ λ p
Yt = Ykt dk ,
0
where λ p > 0 is the price markup. The iso-elastic demand for intermediate good k is given by:
− 1+λ λ p
Pkt p
Ykt = Yt .
Pt
The zero profit condition for competitive final good producers implies that the aggregate price index
is Z 1 −1
−λ p
λp
Pt = Pkt dk .
0
Each intermediate good k is produced by a price-setting monopolist using labor Lkt and physical
capital Kkt 23 :
Ykt = ( Zt Lkt )1−α Kkt
α
,
where Zt is the aggregate TFP. The variable Zt denotes a non-stationary TFP series that we describe
in the next subsection.24 Firms may not be able to adjust their price in a given period, but they will
always choose inputs to minimize total cost each period. The cost minimization yields the input
demand functions.
Kkt α

1− α
Wt = (1 − α) mckt Zt ,
Lkt
Kkt α−1

k 1− α
Rt = α mckt Zt .
Lkt
The first order condition implies that the capital labor ratio at the firm level is independent on
firm-specific variables:
Kkt Kt α Wt
= = ,
Lkt Lt 1 − α Rkt
Thus, (nominal) marginal cost is independent of firm specific variables:
α 1− α
Rkt

1 Wt
Pt mckt = Pt mct = .
Zt1−α α 1−α
23 We can append fixed cost in the production function to eliminate steady state profits. This is usually
done to justify no entry and exit in the steady state in a DSGE model (Christiano, Eichenbaum, and Evans,
2005; Justiniano, Primiceri, and Tambalotti, 2013; Smets and Wouters, 2007). In a micro-founded model of
growth, positive rents are needed to incentivize investment in growth (Romer, 1990).
24 Relative to the conventional literature, we will allow for an endogenous relationship between TFP growth
and output gap.
A25

Each firm k is assumed to set prices on a staggered basis following Calvo (1983). With probability
(1 − θ p ), a firm adjusts its price independent of previous history. A resetting firm chooses P̃kt to
maximize:
∞
P̃
Et ∑ θ sp−t Qt,s kt − mcs Ys (k ) ,
s=t Ps
subject to demand for its product
− 1+λ λ p
Pkt p
Ykt = Yt ,
Pt
where the stochastic discount factor in period s relative to period t is given by:
Λs Pt
Qt,s = βs−t ,
Λt Ps
and Λt is the marginal utility of consumption defined later. The first order condition is :
∞ − 1+λ λp p
P̃t P̃t
Et ∑ θ sp−t Qt,s − (1 + λ p ) mcs Ys = 0 .
s=t Πt,s Πt,s
By the law of large numbers, the law of motion of the aggregate price index Pt is given by:
1 1 1
λp
= (1 − θ p )( P̃t ) λ p + θ p Pt−p1 .
λ
Pt
F.2. Households
Rest of the model components are standard. We briefly summarize them here and leave the detailed
derivations to the appendix.
Households
Each household supplies differentiated labor indexed by j. Household j chooses consumption Ct ,
risk-free nominal bonds Bt , investment It and capital utilization ut to maximize the utility function,
with external habits over consumption:

∞ s−t ω 1+ ν
Et Σ s = t β log(Cj,s − hCj,s−1 ) − L ,
1 + ν j,s
where h is the degree of habit formation, ν > 0 is the inverse Frisch elasticity of labor supply,
ω > 0 is a parameter that pins down the steady-state level of hours, and the discount factor β
satisfies 0 < β < 1. We assume perfect consumption risk sharing across the households. As a result,
household’s budget constraint in period t is given by
Pt Ct + Pt It + Bt+1 = Bt (1 + it ) + BSj,t + (1 + τ w )Wt L j,t + Γt + Tt + RtK ut Ktu − Pt a(ut )Ktu , (11)
where It is investment, BSj,t is the net cash-flow from household j’s portfolio of state-contingent
securities. Labor income Wt L j,t is subsidized at a fixed rate τw . Households own an equal share of
all firms, and thus receive Γt dividends from profits. Finally, each household receives a lump-sum
government transfer Tt . Since households own the capital and choose the utilization rate, the amount
A26

of effective capital that the households rent to the firms at nominal rate RtK is :
Kt = ut Ktu ,
The (nominal) cost of capital utilization is Pt a(ut ) per unit of physical capital. As in the literature
(Smets and Wouters 2007) we assume a(1) = 0 in the steady state and a00 > 0. Following Christiano,
Eichenbaum, and Evans (2005), we assume investment adjustment costs in the production of capital.
Law of motion for capital is as follows:

It
Ktu+1 = 1 − S It + (1 − δk )Ktu ,
(1 + gss ) It−1
where gss ≡ µ̄ is the steady state growth rate of Zt . Utility maximization delivers the first order
condition linking the inter-temporal consumption smoothing to the marginal utility of holding the
risk-free bond
Λ t +1

Pt
1 = βEt (1 + i t ) . (12)
Λt Pt+1
The stochastic discount factor in period t + 1 is given by:
Λt+1 Pt
Qt,t+1 = β ,
Λt Pt+1
where Λt is the marginal utility of consumption given by:
1 hβ
Λt = − .
Ct − hCt−1 Ct+1 − hCt
The household does not choose hours directly. Rather each type of worker is represented by a
wage union who sets wages on a staggered basis. Consequently the household supplies labor at the
posted wages as demanded by firms.
We introduce capital accumulation through households. Solving household problem for invest-
ment and capital yields the Euler condition for capital:
" !#
Λt+1 RtK+1
qt = βEt ut+1 − a(ut+1 ) + qt+1 (1 − δk ) ,
Λt Pt+1
where the (relative) price of installed capital qt is given by

It 0 It It
qt 1 − S −S
(1 + gss ) It−1 (1 + gss ) It−1 (1 + gss ) It−1
Λ t +1 It+1 2 0

1 It+1
+β , q t +1 S =1
Λt (1 + gss ) It (1 + gss ) It
Choice of capital utilization rate yields:
RtK
= a0 (ut ) .
Pt
A27

Wage setting
Wage Setting follows Erceg, Henderson, and Levin (2000) and is relatively standard. Perfectly
competitive labor agencies combine j type labor services into a homogeneous labor composite Lt
according to a Dixit-Stiglitz aggregation:
Z 1 1
1+ λ w
1+ λ w
Lt = L j,t dj ,
0
where λw > 0 is the nominal wage markup. Labor unions representing workers of type j set wages
on a staggered basis following Calvo (1983), taking given the demand for their specific labor input:
− 1+λ λw −λw
Wj,t 1
Z −1
w
λw
L j,t = Lt , where Wt = Wj,t dj .
Wt 0
In particular, with probability 1 − θ, the type-j union is allowed to re-optimize its wage contract
and it chooses W̃ to minimize the disutility of working for laborer of type j, taking into account
the probability that it will not get to reset wage in the future.25 The first order condition for this
problem is given by:
∞
" #
L ν
Et ∑ ( βθw )s Λt+s (1 + τtW )W̃t − (1 + λw )ω
j,t + s
L j,t+s = 0 . (13)
s =0 Λt+s
By the law of large numbers, the probability of changing the wage corresponds to the fraction of
types who actually change their wage. Consequently, the nominal wage evolves according to:
1 1 1
Wtλw = (1 − θw )W̃tλw + θw Wtλ−w1 ,
where the nominal wage inflation and price inflation are related to each other :
Wt wt 1 1
πtw = = ,
Wt−1 w t −1 π t 1 + g t
where πt ≡ PPt−t 1 is the inflation rate, wt ≡ Wt

Pt Zt is the productivity adjusted real wage and gt is the
growth rate of Zt .
F.3. Government
The central bank follows a Taylor rule in setting the nominal interest rate it . It responds to deviations
in inflation, output and output growth rate from time-t natural allocations.


ρR φπ !φy 1−ρR
1 + it 1 + i t −1  πt Yt mp
= f ,t
 et , (14)
1 + iss 1 + iss πss Yt
where iss is the steady state nominal interest rate, πt is gross inflation rate, πss is the steady
f ,t
state inflation target, Yt is the time-t natural output, ρ R determines interest-rate smoothing and
25 We assume imperfect wage indexation in our nominal wage rigidity assumption. We ignore specifying it
here for ease of exposition. See appendix for details.
A28

mp
et ∼ N (0, σr ) is the monetary policy shock.
We assume government balances budget every period:
Z 1
Pt Tt = τ p pit xit di + τ w Wt Lt + Pt Gt ,
0
where Gt is the government spending, which is determined exogenously as as a fraction of GDP

!
1
Gt = 1 − g Yt ,
λt
and where the government spending shock follows the process:

g g g g
log λt = (1 − ρ g )λ g + ρ g log λt−1 + et ; et ∼ N (0, σg ) ,
F.4. Market clearing

Market clearing requires
Yt = Ct + It + a(ut )Ktu + Gt .
F.5. Stationary allocation

We normalize the following variables :
yt = Yt /Zt ,
ct = Ct /Zt ,
k t = Kt /Zt ,
kut = Ktu /Zt−1 ,
It = It /Zt , capital investment ,
Gt = Gt /Zt , government spending ,
wt = Wt /( Zt Pt ) ,
rtk = Rkt /Pt ,
λt = Λt Zt ,
Γt
Γ̃t ≡ .
Pt Zt
p p
Definition 1 (Normalized equilibrium). 19 endogenous variables {λt , gt+1 , it , πt , πtw , X1t , X2t , Xitw ,
w , c , y , z , k u , r K , I , q , u , k , w , L }, 3 shocks { λ g , emp , µ } given the natural rate allocation
X2t t t t t +1 t t t t t t t t t t
f f
variables {yt , gt+1 }.
Consumption Euler equation

λ t +1 1 + it
1 = βEt , (15)
λ t (1 + g t +1 ) π t +1
A29

1 hβ
λt = hct−1
− , (16)
ct − ct+1 (1 + gt+1 ) − hct
1+ gt
Price-setting
 1 −λ p
p λp
X1t 1 − θ p πt
p =  , (17)
X2t 1 − θp
1+ λ p
p λp p
X1t = λt yt mct + θ p βπt+1 X1t+1 , (18)
1
p 1 p
λt yt + θ p βπt+p 1 X2t+1 ,
λ
X2t = (19)
1 + λp
Wage-setting
1
!−λw +(1+λw )ν
w
X1t 1 − θw (πtw ) λw
w =
X2t 1 − θw
, (20)
− (1+λw,t +1 )(1+ν) (1+λw )(1+ν)

w w 1− ι w
= ωLt 1+ν + θw β (πss w
λw,t+1
X1t ) (πW,t )ιw πW,t+λ1w X1t +1 , (21)
1 + τw λ −1 1
w w 1− ι w w
X2t = λt wt Lt + θw β (πss ) (πW,t )ιw w,t+1 πW,t
λw
+1 X2t+1 , (22)
1 + λw
wt
πW,t = π t (1 + gt ) , (23)
w t −1
Capital investment
It 1 + gt kut

kut+1 = 1−S It + (1 − δk ) , (24)
It−1 1 + gss 1 + gt

λ t +1
qt = βEt rtK+1 ut+1 − a(ut+1 ) + qt+1 (1 − δk ) , (25)
λ t (1 + g t +1 )
It 1 + gt It 1 + gt It 1 + gt

0
qt 1−S −S
It−1 1 + gss It−1 1 + gss It−1 1 + gss
1 + g t +1 It +1 2 0 It +1 1 + g t +1

λ t +1
+β q t +1 S =1 (26)
λt 1 + gss It It 1 + gss
Capital utilization rate

kut
k t = ut , (27)
1 + gt
rtK = a0 (ut ) , (28)
A30

Production technologies
yt = kαt L1t −α , (29)
yt
rtk = α , (30)
kt
yt
w t = (1 − α ) , (31)
Lt
Endogenous growth
f
gt = log Zt − log Zt−1 = µt + η log(yt−1 /yt−1 ) , (32)
Government


ρR φπ !φy 1−ρR
1 + it 1 + i t −1  πt yt mp
= f
 et , (33)
1 + iss 1 + iss πss yt
Market clearing !
ku 1
y t = c t + It + a ( u t ) t + 1− g yt , (34)
1 + gt λt
Shocks
g g g g
log λt = (1 − ρ g )λ g + ρ g log λt−1 + et ; et ∼ N (0, σg ) , (35)
mp
et ∼ N (0, σr ) , (36)
F.6. Steady state

1 1+i
1=β ,
(1 + g ) π
(1 + g) − hβ
λ= ,
c (1 + g − h )
1
mc = ,
1 + λp
1 + τw
ωL1+ν = λwL ,
1 + λw
π w = π (1 + g ) ,
q = 1,
u = 1,
1 − δk u
(1 − )k = I ,
1+g
A31

1 K
1=β r + (1 − δk ) ,
1+g
ku
k= ,
1+g
r K = a 0 (1) ,
y = k α L 1− α ,
y
rk = α ,
k
y
w = (1 − α ) ,
L
g = µ,

1
y = c+I+ 1− g y,
λ
F.7. Approximate equilibrium

We log-linearize the variables around the steady state as follows: for any variable x,
x
t
x̂t = log ,
x
where x is the steady state, except for the following variables

1 + g t +1
ĝt+1 ≡ log ,
1+g

1 + it
ît = log ,
1 + iss
Definition 2 (Approximate equilibrium). An approximate competitive equilibrium in this economy
with endogenous growth is defined as a sequence of 16 endogenous variables {λ̂t , ĝt+1 , ît , π̂t , ĉt ,
π̂tw , m̂ct , ŷt , L̂t , ŵt , k̂ut+1 , Ît , q̂t , r̂tK , k̂ t , ût }, which satisfy the following equations, for a given sequence
mp
of one exogenous shock {êt } and given sequence of flexible price allocations.
Consumption Euler equation
(Et λ̂t+1 − λ̂t − ĝt+1 ) + ît − Et π̂t+1 = 0 , (37)

1+g 1+g h
λ̂t = ĝt − (ĉt + ĝt ) − ĉ (38)
1 + g − hβ 1+g−h (1 + g ) − h t −1

hβ 1+g h
+ Et (ĉt+1 + ĝt+1 ) − ĉt
(1 + g) − hβ 1+g−h (1 + g ) − h
Price-setting
β ιp
π̂t = Et π̂t+1 + π̂t−1 + κ p m̂ct , (39)
1 + ιpβ 1 + ιpβ
A32

(1−θ p β)(1−θ p )
where κ p = θ p (1+ ι p β )
.
Wage-setting
β ιw
π̂tw = Et π̂tw+1 + π̂ w + κw [−λ̂t + ν L̂t − ŵt ] , (40)
1 + ιw β 1 + ι w β t −1
(1−θw )(1− βθw )
where κw ≡ >0
θw (1+ι w β)(1+ν(1+ λ1w ))
π̂tw = ŵt − ŵt−1 + π̂t + ĝt , (41)
Capital investment
I 1 − δk h u i
k̂ut+1 = Î t + k̂ t − ĝ t + 1 , (42)
ku 1+g
rK 1 − δk
q̂t = Et λ̂t+1 − λ̂t − ĝt+1 + r̂tK+1 + K Et q̂t+1 ,

K
(43)
r + (1 − δk ) r + (1 − δk )
q̂t − S00 Ît − Ît−1 + ĝt + βS00 Ît+1 − Ît + ĝt+1 = 0 ,

(44)
where we parametrize S00 from Smets and Wouters (2007), or Justiniano, Primiceri, and Tambalotti
(2013) to be between 3 and 5.
Capital utilization rate

k̂ t = ût + k̂ut − ĝt , (45)
a00 (1)
r̂tK = ût . (46)
a 0 (1)
Production technologies
ŷt = αk̂ t + (1 − α) L̂t , (47)
r̂tk = ŷt − k̂ t , (48)
ŵt = ŷt − L̂t . (49)
Endogenous growth
f
ĝt = µ̂t + η (ŷt−1 − ŷt−1 ) , (50)
where η is the hysteresis elasticity.
Government h i
f mp
ît = ρ R ît−1 + (1 − ρ R ) φπ π̂ + φy (ŷt − ŷt ) + êt . (51)
Market clearing
1 c I a 0 (1) k
g
ŷt = ĉt + Ît + ût . (52)
λ y y y
A33

Appendix references
Benigno, Gianluca, Pierpaolo Benigno, and Fabio Ghironi. 2007. Interest rate rules for fixed exchange
rate regimes. Journal of Economic Dynamics and Control 31(7): 2196–2211.
Benigno, Pierpaolo. 2004. Optimal monetary policy in a currency area. Journal of international
economics 63(2): 293–320.
Bergeaud, Antonin, Gilbert Cette, and Rémy Lecat. 2016. Productivity Trends in Advanced Countries
between 1890 and 2012. Review of Income and Wealth 62(3): 420–444.
Burnside, Craig, and Martin Eichenbaum. 1996. Factor-Hoarding and the Propagation of Business-
Cycle Shocks. American Economic Review 86(5): 1154–1174.
Calvo, Guillermo A. 1983. Staggered prices in a utility-maximizing framework. Journal of Monetary

Economics 12(3): 383–398.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. Nominal Rigidities and
the Dynamic Effects of a Shock to Monetary Policy. Journal of Political Economy 113(1): 1–45.
Corsetti, Giancarlo, Keith Kuester, and Gernot J. Müller. 2013. Floats, pegs and the transmission of
fiscal policy. Central Banking, Analysis, and Economic Policies Book Series 17: 235–281.
De Paoli, Bianca. 2009. Monetary policy and welfare in a small open economy. Journal of international
Economics 77(1): 11–22.
Durbin, J. 1959. Efficient Estimation of Parameters in Moving-Average Models. Biometrika 46(3/4):

306–316.
Erceg, Christopher J., Dale W. Henderson, and Andrew T. Levin. 2000. Optimal monetary policy
with staggered wage and price contracts. Journal of Monetary Economics 46(2): 281–313.
Farhi, Emmanuel, and Ivan Werning. 2012. Dealing with the trilemma: Optimal capital controls
with fixed exchange rates. Working Paper 18199, National Bureau of Economic Research.
Fernald, John G. 2014. A quarterly, utilization-adjusted series on total factor productivity. Working
Paper Series 2012-19, Federal Reserve Bank of San Francisco.
Galı́, Jordi. 2015. Monetary policy, inflation, and the business cycle: an introduction to the new Keynesian
framework and its applications. Princeton University Press.
Gali, Jordi, and Tommaso Monacelli. 2005. Monetary policy and exchange rate volatility in a small
open economy. The Review of Economic Studies 72(3): 707–734.
Imbs, Jean M. 1999. Technology, growth and the business cycle. Journal of Monetary Economics 44(1):
65–80.
Itskhoki, Oleg, and Dmitry Mukhin. 2019. Mussa puzzle redux. Princeton University. Unpublished.
Jordà, Òscar, Moritz Schularick, and Alan M. Taylor. 2019. The effects of quasi-random monetary
experiments. Journal of Monetary Economics. Forthcoming.
Justiniano, Alejandro, Giorgio E. Primiceri, and Andrea Tambalotti. 2013. Is there a trade-off between
inflation and output stabilization? American Economic Journal: Macroeconomics 5(2): 1–31.
A34

Kuersteiner, Guido M. 2005. Automatic Inference For Infinite Order Vector Autoregressions. Econo-
metric Theory 21(1): 85–115.
Lewis, Richard, and Gregory C. Reinsel. 1985. Prediction of multivariate time series by autoregressive
model fitting. Journal of Multivariate Analysis 16(3): 393–411.
Lusompa, Amaze B. 2019. Local projections, autocorrelation, and efficiency. Unpublished. UC Irvine.
Nakamura, Emi, and Jón Steinsson. 2014. Fiscal stimulus in a monetary union: Evidence from US
regions. American Economic Review 104(3): 753–92.
Romer, Christina D., and David H. Romer. 2004. A New Measure of Monetary Shocks: Derivation
and Implications. American Economic Review 94(4): 1055–1084.
Romer, Paul M. 1990. Endogenous Technological Change. Journal of Political Economy 98(5): 71–102.
Smets, Frank, and Rafael Wouters. 2007. Shocks and frictions in US business cycles: A Bayesian
DSGE approach. American Economic Review 97(3): 586–606.
Wieland, Johannes F., and Mu-Jeung Yang. 2016. Financial Dampening. NBER Working Papers
22141, National Bureau of Economic Research, Inc.
Wiener, Norbert, and Pesi Masani. 1958. The prediction theory of multivariate stochastic processes,
II: The linear predictor. Acta Mathematica 99(1): 93–137.
A35

The Long-Run Effects of Monetary Policy

Uploaded by

Copyright:

Available Formats

The Long-Run Effects of Monetary Policy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Long-Run Effects of Monetary Policy

Uploaded by

Copyright:

Available Formats

The long-run effects of monetary policy ?

Òscar Jordà † Sanjay R. Singh ‡ Alan M. Taylor §

Keywords: monetary policy, money neutrality, hysteresis, trilemma, instrumental vari-

and CEPR ([email protected]).

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

The trilemma of international finance gives a theoretically justified source of exogenous

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

where a, b, c, c∗ , d, f , g, χ ≥ 0. Domestic output Y is equal to the sum of domestic absorption

rate pegs, i.e., g is smaller.

Electronic copy available at: https://ssrn.com/abstract=3529773

The textbook specification with hard pegs

Financial spillovers with soft pegs

Electronic copy available at: https://ssrn.com/abstract=3529773

4. Measuring long-run treatment effects in finite samples

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

4.2. Impulse responses from a VAR(p)

to ensure consistency of Â1 , . . . , Â p .

Bh = Ah + Ah−1 B1 + · · · + A1 Bh−1 . (4)

Electronic copy available at: https://ssrn.com/abstract=3529773

A p+1 Bh−( p+1) + · · · + Ah−1 B1 + Ah ,

so that the sign of the inconsistency is, a priori, unknown.

Electronic copy available at: https://ssrn.com/abstract=3529773

yt+h = Ah,1 yt−1 + · · · + Ah,p yt− p + ut+h . (7)

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

Impulse response Impulse response

Cumulated response Cumulated response

True VAR(4) LP(4)

Electronic copy available at: https://ssrn.com/abstract=3529773

y j,t+h − yi,t−1 = αi,h + ∆i

for h = 0, 1, . . . , H; i = 1, . . . , N; t = t0 , . . . , T, where y j,t+h is the outcome variable for

Electronic copy available at: https://ssrn.com/abstract=3529773

pegs (q = 1) All years PreWW2 PostWW2

5.1. Main results

Electronic copy available at: https://ssrn.com/abstract=3529773

h=2 -0.31∗∗ -1.37∗∗∗ 0.00 -0.29∗∗∗ -1.20∗∗∗ 0.00

h=4 -0.22 -1.89∗∗∗ 0.00 -0.15 -1.24∗∗∗ 0.01

h=6 -0.12 -2.67∗∗∗ 0.00 -0.07 -2.55∗∗∗ 0.00

h=8 -0.34 -4.83∗∗∗ 0.00 0.08 -2.93∗∗∗ 0.00

h = 10 0.13 -4.22∗∗∗ 0.00 0.34 -2.80∗∗∗ 0.00

h = 12 0.35 -3.82∗∗∗ 0.00 0.56 -3.31∗∗∗ 0.00

Electronic copy available at: https://ssrn.com/abstract=3529773

(a) Full sample: 1890–2015 (b) Post-WW2 sample: 1948–2015

Electronic copy available at: https://ssrn.com/abstract=3529773

Electronic copy available at: https://ssrn.com/abstract=3529773

6. Robustness and discussion

6.1. Spillover correction for the trilemma instrument

Electronic copy available at: https://ssrn.com/abstract=3529773

where ∆y is the outcome variable, ∆i is an endogenous variable, and z is an instrumental

Without loss of generality, we can assume that β = λφ where λ is an unknown parameter

(∆y − zφ̂(λ)) = ∆i β + v + z(φ̂(λ) − φ(λ)) ,

Electronic copy available at: https://ssrn.com/abstract=3529773

(a) Full sample: 1890–2015 (b) Post-WW2 sample: 1948–2015

IV OLS IV spillover corrected IV OLS IV spillover corrected

yt+h = Bh+1 yt−1 + {Ch+2 yt−2 + Ch+3 yt−3 + · · · } + et+h + B1 t+h−1 + t Bh ,

ut+h = t+h + { B1 t+h−1 + B2 t+h−2 + · · · + Bh t } + {Ch+k+1 yt−k−1 + Ch+k+2 yt−k−2 + · · · } .

T−h−k ∑ t+h− j Xt0−1 → 0 ,