Integration of Microsimulation Model Into Dynamic Scoring Model
Faculty Mentor: Richard Evans, Economics
We integrated individual tax rates produced by a microsimulation tax policy model with a
dynamic general equilibrium tax policy model. We can use this to conduct macroeconomic
analysis or score hypothetical tax policies. This approach captures the rich heterogeneity,
realistic demographics, and tax-code nuance of the microsimulation model and includes this
nuance to increase the accuracy of a general equilibrium model with an elevated level of
heterogeneity. Furthermore, we derive a functional form which suggests that tax rates depend
both on capital income and labor income. Applying this approach to a canonical example of tax
policy change—a cut of 10% in statutory marginal income tax rates and a two-fold increase of
the standard deduction, we find an increase of gross domestic product (GDP) by about 1.4%.
Accounting for these macroeconomic effects of the tax change results in an offset of about 53%
of the static cost of the tax cuts.
Heterogeneous agent models are standard and canonical in macroeconomics. This development
has made macroeconomic models more realistic and allowed exploration of topics related to
distributional issues. However, policy instruments usually lack detail, despite being incorporated
into dynamic general equilibrium models. This disparity between high levels of heterogeneity in
model agents and lack of policy detail can be especially prevalent models used to evaluate tax
policy. This gap is often due to the intractability of modeling the fine details of real-world policy.
Our contributions are primarily methodological. First, we derive a flexible functional form for
tax rates that has the properties necessary for solving a dynamic general equilibrium (DGE)
model while retaining much of the heterogeneity found in microsimulation model tax data.
Second, we describe a methodology where one can easily fit these tax functions using the output
of a microsimulation model. The use of a microsimulation model is important in that these
models are able to capture the rich detail of tax policy and how it affects households with
varying economic and demographic characteristics. The tax functions we propose then map the
results of the microsimulation model, the computed average and marginal tax rates, into
functions that can be used in a macroeconomic model. We tailor our functions here to a specific
microsimulation model and DGE model, but the methodology we propose can be scaled up or
down to account for models with more or less heterogeneity. Our approach has two distinct
advantages. It allows the DGE model to capture more detail of tax policy in than previously used
methods. It also greatly reduces the cost to incorporating rich policy detail and counterfactuals
into macroeconomic analysis. The bridge we build between the microsimulation model and the
macroeconomic model essentially automates this process.
Our goal was to introduce a methodology through which researchers and policy analysts could
integrate the strengths of a microsimulation model of tax policy into aggregate models that allow
one to understand the macroeconomic impacts of fiscal policy. We apply this methodology by
estimating the revenue effects of a canonical example policy change in a microsimulation model
and a DGE model. More broadly, we note that the methodology and underlying source code for
the tax function estimation can be applied to link other microsimulation or macroeconomic
models. Thus these methods can be quite general and utilized by a wide class of models. The
important consideration is the flexible specification of the tax functions that allow details of tax
policy to be mapped to parametric functions used in macroeconomic models. A compelling
direction of future work integrating microsimulation models with general equilibrium models
lies providing consistency between the macroeconomic assumptions underlying the
microsimulation model and the macroeconomic effects found in the general equilibrium model.
In our microsimulation model, it is assumed that macroeconomic variables either remain
unchanged by policy experiments or a new path for the macroeconomic variables are produced
by time series models. We could expand our methodology of integration of the microeconomic
and macroeconomic models by providing for the following iterative procedure: Obtain solutions
to the microsimulation model given an assumption about macroeconomic variables. Solve the
macroeconomic model given the microeconomic results. Use the new macroeconomic variable
time paths in the microsimulation model and re-compute the microsimulation results. Repeat
until the macroeconomic assumptions of the microsimulation model are consistent with the
macroeconomic results of the general equilibrium model.