Research Statement - Gregor Boehl
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Research Statement
Gregor Boehl∗
June 10, 2020
This research statement gives a concise outline on the research I am currently working on,
the work I have conducted during my PhD studies, and gives an idea about the projects I plan
to carry out in the near future. Accordingly, I split this statement in three subsections.
Current projects
Since the completion of my PhD thesis, the main focus of my work has been to understand the
empirical dynamics of the Global Financial Crisis and the Great Recession from the perspective
of structural economic models. As a consequence of the Crisis, economists were confronted with
phenomena that are new to many advanced economies. One is the zero lower bound (ZLB) on
nominal interest rates. In response to the sharp decline in economic activity, central banks in
major advanced countries reduced nominal interest rates to historically low levels. Unable to
lower interest rates any further, many central banks resorted to unconventional policy tools such
as quantitative easing (QE) in order to deliver additional monetary stimulus, and governments
in particular in Europe have employed large fiscal stimulus packages. Yet, economists and policy
makers are still debating about the macroeconomic effects of these unconventional measures –
did the measures have any effects, and if so, what are the effects? Answers to these questions are
also crucial against the backdrop of the Corona-crisis because both, the US and the Euro Area
have again seen large scale asset purchases in combination with massive fiscal stimulus.
The binding ZLB constraint on nominal interest rates poses a major problem for a quantitative-
structural analysis. Traditional solution, filtering and estimation methods typically do no work
in the presence of a nonlinearity such as the ZLB. Existing alternatives tend to be compu-
tationally demanding. In Boehl (2020) I provide new methods to solve, filter, and estimate
dynamic-stochastic general equilibrium (DSGE) models with a binding ZLB efficiently, robust,
and fast. The estimated model enables to structurally assess the macroeconomic effects of QE as
well as fiscal stimulus packages, and to conduct counterfactual analysis, run policy simulations
and investigate the risks of QE both in the short and long-run.
This methodology is applied in Boehl et al. (2020) to study the macroeconomic effects of
unconventional monetary policy conducted by the FED during the last decade. We extend a
medium-scale DSGE model with heterogeneous households and include a banking sector, as well
financial frictions inspired by Gertler and Karadi (2013). We incorporate several important
channels to which QE can affect the economy. We find that from 2009 to 2015 the overall QE
measures increased output moderately by about 1.2 percent. We find that the effects of liquidity
provision were negligible but both government bond and especially Mortgage Backed Securities
(MBS) purchases had an positive effect on investment of nearly 9%. However, the effects on
consumption were actually negative and led to a decrease by 0.7 percent. We report that both,
government bond and capital asset purchases were effective in improving financing conditions.
Especially capital asset purchases significantly facilitated new investment and increased the pro-
duction capacity. Against the backdrop of a fall in consumption, supply side effects dominated
which led to a mild disinflationary effect of about 0.25 percent annually.
∗
University of Bonn; email: gboehl@uni-bonn.de; web: https://gregorboehl.com
1Gregor Boehl Research Statement
Using the methods from Boehl (2020), in Boehl and Strobel (2020) we estimate a selection
of medium-scale models of the US economy before and during the ZLB period. We identify the
structural shocks (in particular during the Great Recession) and compare the performance of
several off-the-shelf models during this sample. We find that the standard model and variants
augmented with financial frictions or household heterogeneity á la TANK remain unable to
provide a simultaneous explanation for the core dynamics of the Great Recession: a drastic fall
in investment, a more modest decline in consumption, and a temporary dip of inflation. The
standard model does the best job in accounting for the differential of the drop in consumption
and investment, but absent additional shocks predicts persistently low inflation. TANK captures
the recovery of inflation but worsens the drop differential. A financial frictions extension fails in
assigning a common causal driver to any combination of the three. Associated financial shocks
mis-predict an increase in consumption on impact. TANK is also rejected in the pre-2009 sample
because it implies a counter movement of investment in response to impulses from wages. Our
results stress the overall importance of elevated risk premia for households following the crisis.
In contrast, using pre-2009 based estimates for the analysis of the post-crisis period overtaxes
the role of investment distortions. We also investigate on the cost of a binding ZLB for the
US economy, which is about 1.5% of GDP. To my best knowledge, this paper is the first one
estimating the complete model of Smets and Wouters (2007) – and variants thereof – while fully
including an endogenous ZLB. Thus, I believe it has the potential to provide a new reference
calibration for future research.
Previous work
The work that directly followed from my PhD thesis evolved around several, distinct topics.
In Boehl (2017) I study whether monetary policy can mitigate spillovers from speculative asset
prices to the real economy. My analysis is based on an estimated model with credit constraints in
which excess volatility of stock markets is endogenously amplified through behavioral speculation.
I find that speculative behavior, and its feedback to asset prices, are key features to replicate
central empirical moments. Standard monetary policy rules can be shown to amplify stock
price volatility. Numerical analysis suggests that asset price targeting can offset the impact of
speculation on either output or inflation (but not on both) and can dampen excess volatility.
The dampening effect of this policy is limited due to its undesirable response to non-financial
shocks. A particular strength of my modeling approach is, that it allows to study endogenous
financial crises that are triggered by speculative behavioral dynamics at the asset market, and
can generate server spillovers to the macroeconomy. While I find that monetary policy should
rather abstain from interfering with asset markets in normal times, my simulation studies suggest
that financial tumult can motivate central banks to lean against asset prices to prevent further
hazard.
In Boehl and Hommes (2020) we contribute to the large literature on bounded rationality.
We analyse the interaction of perfectly rational agents in the context of an asset market with
coexisting boundedly rational traders. Whether an individual agent is perfectly rational or
boundedly rational is determined endogenously, depending on the market performance of each
type. Perfect rationality implies full knowledge of the model including the non-linear switching
process itself. I use projection methods to find a recursive minimal state variable solution of a
system with complex nonlinear dynamics. This is novel to the literature, as previous work had -
due to technical limitations - to impose strong limitations on the degree of rationality of rational
agents. Depending on the parameterization, the fact that rational agents are able to predict the
behavior of less sophisticated agents can trigger complicated endogenous fluctuations that are
well captured by the solution algorithm. We find that, contrasting conventional wisdom, in a
financial market setup boundedly rational agents are not necessarily driven out of the market.
While up to a certain point the presence of fully rational agents tends to have stabilizing effects
2Gregor Boehl Research Statement
it may later even amplify endogenous fluctuations. The methodological contribution of this work
– the combination of complex and potentially chaotic nonlinear dynamics with iterative solution
methods – is completely novel to my best knowledge.
The last chapter of my thesis adds to the growing literature that studies the dynamics of
wealth inequality. In Boehl and Fischer (2017) we show that the degree of capital gains taxation
can retrace the data of the US from the 1920s up to the most recent observations. Precisely
matching up- and downturns and levels of top shares, it has high forecasting power. This result
is drawn from an estimated, micro-founded portfolio-choice model where idiosyncratic return
risk and disagreement in expectations on asset returns generate an analytically tractable fat-
tailed Pareto distribution for the top-wealthy. This allows us to decompose the sample into
periods of transient and stationary wealth concentration. The model generates good out-of-
sample forecasts. As an addition we predict the future evolution of inequality for different tax
regimes.
Future Work
For the near future I aim to continue working within the nexus of structural-empirical analysis
of fiscal and monetary policy, including the ELB and the methodology I am providing in Boehl
(2020). In particular, the method naturally allows to analyze the fiscal stimulus packages and
the measures of unconventional monetary policy during and after the financial crisis in Europe
similar to the work in Boehl et al. (2020). As such it can answer the question whether the
large-scale bond purchases in the Euro area (EA) were successful in preventing worse outcomes.
It can further quantify the role of the various fiscal stimulus packages for the economic recovery.
A third and much debated question that I can potentially answer is, whether the ELB was –
at all – binding in the EU or if the policy of negative interest rates was able to circumvent the
problem. So far, the literature was unable to provide a structural answers to these matters,
which is mainly due to the technical difficulties tackled by my methodological contribution. I
am also focussing efforts to see the papers discussed above published.
In the longer term I plan to concentrate my research on three pillars. The first is concerned
with improving the empirical performance of macroeconomic models, and thereby the quality of
their policy implications. Second, I want to investigate potential connection of secular stagnation
and inequality, and design appropriate fiscal and monetary responses. Third, I want to further
contribute to computational advances in my field.
Effects of economic inequality
The bulk of the new literature on heterogeneity in macroeconomics stresses the importance of
economic inequality when reevaluating macroeconomic phenomena. For example, Auclert and
Rognlie (2017, 2018) study the effects of income concentration on aggregate demand. Much of this
analysis centers around the idea that households with different levels of income have different
marginal propensities to consume, and hence respond differently to changes in the economic
environment. I build on similar mechanisms together with Alexander Clymo of the University of
Essex to contribute to the debate on secular stagnation: the natural rate, together with the labor
share seems to be decreasing over the last decades, while income, wealth and firm concentration
are skyrocketing, jointly with firm markups. The sharp increase in US wealth concentration in
the last decades seems to translate almost one-to-one to the decrease in estimates of natural
interest rates.1
We argue that these findings can largely be attributed to the process of digitalization, and its
direct effects on the distribution of income and wealth. Digitalization disproportionally favors
1
The effects of the wealth concentration on aggregate demand is also discussed in a recent working paper of
Mian et al. (2020).
3Gregor Boehl Research Statement
high-skilled labor, and has led to an increase in the dispersion of firm sizes. We formulate
a two-agent-two-firms real business-cycle (TARBC) model with monopolistic competition á la
Kimball, preferences for wealth and different skill levels across firms and households. A relative
productivity boost that affects only some firms can inflate average markups and run up the
profit share, which lowers the relative labor share. As the profits of owners, and wages of
workers associated with these firms increase, this drives up inequality. As a result, aggregate
overinvestment – hence a large supply – is a plausible explanation to the observation of the
rapid decline in natural interest rates, which are the price for investment. Given our utility
specification, consumption of the rich saturates, and the concentration of wealth leads to an
increase in savings, leading to an overall low natural rate. This model allows to study the effects
of targeted fiscal policy as well as redistribution policies on the distribution of income and wealth
and on aggregate measures alike.
In Boehl et al. (2020) we find that the measures of Quantitative Easing in the US were not
effective to foster inflation. Instead, the improvement in firms’ financing conditions manifested
in a net-decrease in prices, while the net effect on aggregate consumption remained negative. If
we can confirm this finding for the large-scale asset purchases in the Euro Area, this implies that
central banks are left without any tools to stimulate inflation when interest rates are constrained
by the ELB. One alternative is the use of helicopter money – a massive monetary transfer
directly to households – to stimulate consumption immediately. This alternative is not well
studied in structural frameworks as it requires the combination of methods for heterogeneous
agents with methods for occasionally binding constraints. My method (Boehl, 2020) allows to
be straightforwardly combined with linearized heterogeneous agent methods as e.g. Bayer and
Luetticke (2018). This would even allow to estimate a medium-scale HANK model including the
ZLB before conducting policy simulations such as drops of helicopter money and redistribution
policies.
The mystery of the Phillips Curve
It is acknowledged that, despite very serious and concentrated research efforts, contemporary
macro models still do a bad job in accounting for the empirical data. Macroeconomists like
to brand these inconsistencies between theory and the empirical evidence as “Puzzles”. In an
attempt to document and structure these shortcomings, I recently started a collection of Macro
Puzzles 2 . A prominent example is the well documented (Del Negro et al., 2007; Linde et al.,
2017; Gust et al., 2017; Boehl et al., 2020; Boehl and Strobel, 2020) artefact of the flattening New
Keynesian Phillips Curve. Many advanced economies experienced a shallow decline in inflation
despite large negative output gaps during the Great Recession. The absence of a persistent decline
in inflation, known as the missing deflation puzzle, is at odds with macroeconomic theory. It
calls into question one of the fundamental economic relationships: the Phillips curve, linking
inflation to real economic activity. Many economists argue that the Phillips curve has flattened
or that the relationship between inflation and output described by the Phillips curve entirely
broke down. Naturally, the ability to explain and predict inflation is of particular importance for
monetary policy. Another example is the widely accepted view that the monetary transmission
channel through direct effects on households is at odds with the empirical evidence.3
Such misalignment between theory and empirics pushes towards a reevaluation of alterna-
tives. Given that a good share of the macroeconomic literature since Bernanke et al. (1999) has
underlined the importance of financial frictions, we should likewise acknowledge the importance
of credit in the monetary transmission channel (see e.g. Sanches, 2016 and Gu et al., 2016).
Although the recent structural-empirical research suggests that modelling financial intermedia-
tion is essential in order to understand the last two decades of macroeconomic data, a granular
2
The collection can be found on GitHub.
3
See e.g. Kaplan et al. (2018) for a survey.
4Gregor Boehl Research Statement
integration of financial variables into models with financial frictions is a much needed, but com-
plex task. As such, I find it promising to reevaluate the role of endogenous money creation with
regard to the transmission channel as well as for financial shocks (see e.g. McLeay et al., 2014;
Jakab and Kumhof, 2015).
In ongoing research with several coauthors I propose a novel explanation for the missing
deflation puzzle in this proposal. Considering a DSGE model with financial friction, I argue
that a binding ZLB on nominal interest rates helps to explain low deflation during the Great
Recession. In this model, firms face financing cost which – together with marginal factor costs
– affect their marginal production cost. Financing costs are composed of the nominal interest
rate and the risk spread, which then both affect firms’ price decisions. Financial frictions allow
for risk spreads to be endogenous. Then, financing conditions are a key determinant for firms’
marginal cost affecting their price-setting and, hence, inflation. I show that the movement of
the interest rate as determined by a Taylor rule approximately offsets the effect of an increasing
risk premium in normal times, e.g. mild recessions. However, in a deep recession with a binding
ZLB on nominal interest rates, two counteracting effects on marginal costs emerge. On the one
hand, lower demand reduces real marginal costs of firms. On the other hand, marginal financing
cost are particularly high at the ZLB because the risk free rate cannot be lowered any further to
counteract the tight credit market. To test the quality of this explanation, the model has to be
brought to the data using my methodology.
Computational methods
Much of the contemporary research in macroeconomics is constrained by technical and compu-
tational boundaries. I feel that in particular at the computational frontier, there is much room
for improvement. Economists (1) seek accurate approximation of relevant features of nonlinear-
ities in very little computation time.4 We (2) would like to advance in the field of simulations
with heterogeneous agents (see e.g. Den Haan and Rendahl, 2010). And lastly (3), we want to
estimate nonlinear models and obtain good approximations of the distribution of hidden states
at low computation cost.
Macroeconomists are dealing with increasingly complex methods in combination with in-
creasingly complex models. Solving, simulating and estimating a nonlinear model easily involves
several ten thousand lines of code5 . I identify two core-problems here. First, with the complexity
of the methods used, the quality of their implementation increases in relevance (see e.g. Coleman
et al., 2018). Poor implementation in terms of code quality does not only increase complexity,
but also increases the probability of numerical inaccuracies and slow performance. Additionally,
badly written code complicates reusability, and hence slows down progress. Unfortunately, many
macroeconomists lack profound computational training to do this right.
Second, as the size of the code increases, interaction of different groups of researchers and
sharing of code becomes more important. Together with colleagues from the University of Bonn
we recently started the Open-Source-Economics initiative (open-econ.org) where we work with
well-established institutions like the Hausdorff Center for Mathematics to make it easier for gen-
erations to come to work with proper code. As advocates of free and open source software (as
opposed to proprietary programs) we currently also reach out for third party funding to con-
centrate efforts in this field. Additionally, together with an international team of computational
economists, we are organizing a series of symposia at the annual PASC to gather motivated
researchers and bring the topic of reusable and open code on the agenda.
4
E.g. Meyer-Gohde (2014) shows how to capture motives of risk aversion in a linear representation.
5
For example, when ignoring all documentation strings, my package pydsge counts 4925 lines of code.
5Gregor Boehl Research Statement
References
Auclert, A., Rognlie, M., 2017. Aggregate demand and the top 1 percent. American Economic
Review 107 (5), 588–92.
Auclert, A., Rognlie, M., 2018. Inequality and aggregate demand. Tech. rep., National Bureau
of Economic Research.
Bayer, C., Luetticke, R., 2018. Solving heterogeneous agent models in discrete time with many
idiosyncratic states by perturbation methods .
Bernanke, B. S., Gertler, M., Gilchrist, S., 1999. The financial accelerator in a quantitative
business cycle framework. Handbook of macroeconomics 1, 1341–1393.
Boehl, G., 2017. Monetary policy and speculative stock markets. Tech. rep., IMFS Working
Paper Series.
URL https://www.imfs-frankfurt.de/forschung/imfs-working-papers
Boehl, G., 2020. Efficient solution, filtering and estimation of models with OBCs. Tech. rep.
URL https://gregorboehl.com/live/obc_boehl.pdf
Boehl, G., Fischer, T., 2017. Can taxation predict us top-wealth share dynamics? Tech. rep.,
IMFS Working Paper Series.
URL https://www.imfs-frankfurt.de/forschung/imfs-working-papers
Boehl, G., Goy, G., Strobel, F., 2020. A Structural Investigation of Quantitative Easing. Tech.
rep.
URL https://gregorboehl.com/live/qe_bs.pdf
Boehl, G., Hommes, C., 2020. On the Evolutionary Fitness of Rational Expectations. In prepa-
ration.
Boehl, G., Strobel, F., 2020. US Business Cycles at the Zero Lower Bound. Tech. rep.
URL https://gregorboehl.com/live/recession_elb_bs.pdf
Coleman, C., Lyon, S., Maliar, L., Maliar, S., 2018. Matlab, python, julia: What to choose in
economics? .
Del Negro, M., Schorfheide, F., Smets, F., Wouters, R., 2007. On the fit of new keynesian models.
Journal of Business & Economic Statistics 25 (2), 123–143.
Den Haan, W. J., Rendahl, P., 2010. Solving the incomplete markets model with aggregate
uncertainty using explicit aggregation. Journal of Economic Dynamics and Control 34 (1),
69–78.
Gertler, M., Karadi, P., 2013. Qe 1 vs. 2 vs. 3...: A framework for analyzing large-scale asset
purchases as a monetary policy tool. international Journal of central Banking 9 (1), 5–53.
Gu, C., Mattesini, F., Wright, R., 2016. Money and credit redux. Econometrica 84 (1), 1–32.
Gust, C., Herbst, E., López-Salido, D., Smith, M. E., 2017. The empirical implications of the
interest-rate lower bound. American Economic Review 107 (7), 1971–2006.
Jakab, Z., Kumhof, M., 2015. Banks are not intermediaries of loanable funds–and why this
matters .
Kaplan, G., Moll, B., Violante, G. L., 2018. Monetary policy according to hank. Tech. Rep. 3.
6Gregor Boehl Research Statement
Linde, J., Maih, J., Wouters, R., 2017. Estimation of operational macromodels at the zero lower
bound. Tech. rep., manuscript.
McLeay, M., Radia, A., Thomas, R., 2014. Money creation in the modern economy. Bank of
England Quarterly Bulletin , Q1.
Meyer-Gohde, A., 2014. Risky linear approximations. Tech. rep., SFB 649 Discussion Paper.
Mian, A. R., Straub, L., Sufi, A., 2020. The saving glut of the rich and the rise in household
debt. Tech. rep., National Bureau of Economic Research.
Sanches, D., 2016. On the inherent instability of private money. Review of Economic Dynamics
20, 198–214.
Smets, F., Wouters, R., 2007. Shocks and frictions in us business cycles: A bayesian dsge ap-
proach. American Economic Review 97 (3), 586–606.
7You can also read
