We ask whether a two-country real business cycle (RBC) model can explain both what

happens inside countries (domestic business cycles) and what happens between countries
(international comovements) at the same time. To test this, we compare what the model
predicts with real-world data.

The biggest problem we see is this:

 In the real world, when we compare different countries, output is more correlated
than consumption.
 In the model, we get the opposite: consumption is more correlated than output.

In past research focused only on one country (a "closed economy"), RBC models could
explain a lot of patterns in U.S. economic data. We now want to see if this kind of model can
also explain global patterns — like how economic variables across countries (such as
output, consumption, trade) tend to move together.

In a typical RBC model, countries are hit by technology shocks (random changes in
productivity), and in the global version of the model, these shocks are not perfectly linked
across countries. The way countries interact through international borrowing and lending
could have a big effect on how these shocks influence the economy.

In an open economy (where countries trade and invest with each other):

 A country’s consumption and investment don’t have to match its own production
anymore.
 For consumption, you'd expect international trade to help people "share risks",
meaning consumption would be smoother and less connected to domestic output.
 For investment, capital would likely flow into the country that just had a positive
technology shock, making investment more volatile.

This global view also makes us look at how economic variables move together between
countries.
For example:

 Open economies can run trade deficits or surpluses by borrowing or lending

internationally.
 The trade balance (exports minus imports) changes depending on how much people
want to smooth their consumption and how much investment varies due to global
capital flows.
 Saving and investment are usually tightly linked in closed economies. In open ones,
they can diverge if countries are using international financial markets.

The theory also predicts that:

 With complete markets (where countries can perfectly insure each other against
shocks), consumption should be very closely linked across countries, even if
output isn’t.
So, the big question is:

Can this two-country RBC model explain both how economic variables move within
countries and how they move across countries?

We focus on things that directly involve international markets, like:

 How consumption and output are correlated across countries

 How net exports move with output
 How saving and investment are related

The Model

We build on Kydland and Prescott’s famous RBC model (1982) by extending it to two
countries. Like them, we assume:

 Only one good is produced

 Complete markets exist (so agents can insure against different possible future
events)

But we change two key things:

1. Countries get different technology shocks in each period.

2. People and firms can participate in international financial markets.

We also assume that:

 These shocks can be correlated across countries, meaning a shock in one country
may affect the other.
 There can be "spillovers", where a productivity increase in one country spreads to
the other.

The shock data (how often and how strong the shocks are) comes from Solow residuals for
the U.S. and a group of European countries.

Main Results (Benchmark Economy)

In our main version of the model:

 Consumption is smoother than in the data (standard deviation is 0.40 vs. 0.49 in the
U.S. data)
 Investment is way more volatile than in the real world (10.94 vs. 3.15)
 The link between investment and output is weaker than in the data (0.27 in the
model vs. 0.90 in U.S. data)
So, in these cases, the closed-economy model actually matches the real data better than
the open one.

Looking at international variables:

 Trade balances are much more volatile in the model than in actual data.
o Model: 2.90%
o Canada: 0.79%, Germany: 0.85%, Japan: 0.89%, U.S.: 0.42%
 Output is not positively correlated between countries in the model (it’s actually
negative at -0.18), even though in real life, countries' outputs usually move together.
 Consumption is very strongly correlated across countries in the model (0.88), but
not in the real data (which ranges from -0.23 to 0.65)

The Key Anomaly

This last point is the most persistent issue:

The model always predicts that consumption is highly correlated between countries, even
more than output.
But in reality, the opposite is true.

Why is this happening? We think it’s because the model lets people freely trade assets and
move capital between countries. This leads to:

 Very high consumption correlations

 Low or even negative output correlations
 Highly volatile investment and net exports

Trying to Fix the Problem

To fix this, we try two things:

1. Add a small trading cost (like a small "fee" for shipping goods across countries)
o It helps: makes investment and trade less volatile, and investment becomes
more procyclical (moves with output)
o But it doesn’t fully fix the problem: consumption correlation is still too high
2. Cut off all international trade completely (autarky)
o Even this doesn’t fix the problem
o Consumption is still more correlated than output, so the issue isn’t just due
to perfect international insurance

Final Thoughts
This paper adds to existing research on international RBC models by actually putting
numbers to the theory and comparing it to real-world data. The remaining big challenge is:

Why do models predict such strong international consumption links when the data
shows the opposite?

More work is needed to:

 Improve the theory

 Include other real-world features like different country sizes, types of shocks, or
financial market imperfections

I. Properties of International Business Cycles (Simplified Explanation)

This section reviews how business cycles behave in developed countries during the postwar
period. The authors analyze quarterly data that has been processed using the Hodrick-
Prescott (HP) filter. This filter is commonly used in macroeconomics to separate short-term
fluctuations (i.e., business cycles) from long-term trends. The authors focus specifically on
the medium- and high-frequency changes, which are typically associated with cyclical
economic activity.

The HP filter has been widely applied in previous studies to analyze macroeconomic data
(e.g., Kydland and Prescott, Hansen, Prescott, Christiano and Eichenbaum). Figure 1 in the
original paper illustrates how the HP filter separates the trend from the actual values using
U.S. real GDP data. The analysis focuses on the difference between the observed data and the
filtered trend.

Table 1 presents the cyclical characteristics of the U.S. economy from 1954 to 1989. Here
are the key points from the table:

 The standard deviation of output (a measure of how much it fluctuates) is 1.71%,

and this value will serve as a benchmark for comparing model predictions.
 Consumption (of nondurables and services) is about half as volatile as output.
 Investment in fixed capital is more than three times as volatile as output.
 Hours worked are slightly less volatile than output.
 All three variables — consumption, investment, and hours — are strongly
procyclical (they increase when output increases).
 The trade balance (net exports relative to output) tends to be countercyclical — it
usually declines when output increases. The correlation between net exports and
output is −0.28.

Many of these findings are consistent across other developed economies, as shown in studies
like Danthine and Donaldson.

Table 2 presents international data for 12 developed countries and a European aggregate. It
shows:
  The correlation of output between each country and the U.S. varies but is mostly
positive (except for South Africa). For Japan and major European countries, these
correlations range from 0.22 to 0.48.
 The correlation of consumption between countries is lower than that of output.
The highest is 0.65 for Canada. For the U.S. and the European aggregate,
consumption correlation is 0.46, while output correlation is 0.70.
 The difference between individual countries and the European aggregate is partly due
to the shorter time period used for the European data (the 1970s had higher
correlations than the 1960s or 1980s). However, in both the aggregate and individual
cases, output correlations exceed consumption correlations.

The focus on consumption correlations comes from a well-known result in economic

theory: in models with complete markets, where agents can fully insure against shocks,
consumption across countries should be highly correlated. If all agents have identical and
homothetic preferences, then their consumption patterns should be perfectly correlated,
regardless of differences in income.

This idea is considered a benchmark: if consumption correlations across countries are low, it
suggests that the model does not accurately capture international economic behavior.

The third column in Table 2 shows the correlation between saving and investment rates
within each country. A famous study by Feldstein and Horioka (1980) found that these two
variables are highly correlated in international data, which seems to contradict the idea of
perfectly integrated global capital markets. According to theory (based on the Fisher
Separation Theorem), if capital is mobile, saving and investment decisions should be
independent — yet, in the data, they are strongly correlated.

Subsequent studies (e.g., Obstfeld, Dooley et al., Tesar) confirmed this finding, especially at
low frequencies. However, at the higher frequencies examined in this paper (e.g., quarter-
to-quarter movements), the relationship is less consistent.

There are some theoretical and empirical complications when comparing saving-
investment correlations in models and data:

 Theoretically, the concept of saving depends on the structure of financial markets in

the model (i.e., which assets are available for trade), which affects how saving is
defined.
 Empirically, saving is difficult to measure precisely. The ideal measure would track
changes in the market value of national wealth, but this is hard to observe.
National accounts typically define saving as income minus expenditures, which
ignores capital gains or losses.

For this reason, the authors choose to avoid using the current account or complex saving
definitions. Instead, they use a simpler and more transparent approach:

Saving = output − consumption − government spending,

which is easy to compute both in the model and from real data. This approach captures the
essence of Feldstein and Horioka's argument — that saving and investment can diverge in
open economies.
According to this definition, the authors find, consistent with Obstfeld and Tesar, that the
correlation between saving and investment varies across countries, but is strong and
positive for Germany, Japan, and the United States.

The final two columns of Table 2 refer to net exports. The authors measure trade as the ratio
of net exports to output, and they look at its volatility (using the standard deviation of that
ratio). These measures vary over time and across countries. In every country examined, net
exports are countercyclical — their correlation with output is negative. This means that
when output rises, the trade balance tends to worsen (either through rising imports or falling
exports).

This pattern of countercyclical trade balances has also been observed in:

 Historical data from before World War I and the interwar period (Backus and Kehoe),
 Postwar data using spectral methods (Dellas),
 Keynesian-style empirical models (e.g., Krugman and Baldwin) where income affects
import demand.

Final Summary of Section I:

 Across countries, investment is more volatile than output, consumption is less

volatile, and hours worked are similar in volatility to output. All are procyclical.
 Net exports tend to be countercyclical.
 Output correlations between countries are generally higher than consumption
correlations, which contradicts standard theory with complete markets.
 The correlation between saving and investment varies and is a subject of ongoing
debate.

Here is a simplified and clearer version of Section II ("A World Economy") — made easier
to understand while preserving academic formality and technical accuracy:

II. A World Economy (Simplified Explanation)

In the theoretical model, the world economy consists of two countries. Each country
contains a large number of identical consumers and has its own production technology.
Both countries:

 Produce the same good,

 Share the same functional forms for preferences and technology,
 Use the same parameter values for these functions.

However, there are two key differences between them:

1. Labor used in each country’s production is entirely domestic.

 2. Each country is subject to its own country-specific technology shocks, meaning
productivity changes randomly and independently in each country.

The structure of preferences and technology is mostly taken from Kydland and Prescott's
(1982) closed-economy RBC model. In both the home country (h) and the foreign country (f),
a representative consumer maximizes expected lifetime utility:

E0∑t=0∞βtU(ct,It)for i=h,fE_0 \sum_{t=0}^{\infty} \beta^t U(c_t, I_t) \quad \text{for } i =

h, f

The utility function is:

U(c,I)=(c1−η⋅Iη)γU(c, I) = \left(c^{1 - \eta} \cdot I^\eta \right)^\gamma

Where:

 0<η<10 < \eta < 1, and γ<1\gamma < 1

 ctc_t is consumption,
 ItI_t is a distributed lag of leisure, which represents time spent outside the labor
market (e.g., rest, home activities).
 When γ=0\gamma = 0, the utility becomes a logarithmic function.

Time is normalized to 1 per period. The leisure index is defined as:

It=1−nt+∑j=1∞ajnt−jI_t = 1 - n_t + \sum_{j=1}^{\infty} a_j n_{t-j}

Where ntn_t is the amount of time spent working, and the parameters aja_j determine how
past leisure affects current utility. If a=1a = 1, utility depends only on current leisure; if a<1a
< 1, past leisure also matters.

Technology Shocks:

The model assumes each country’s productivity evolves based on a bivariate autoregressive
process:

Xt+1=AXt+εt+1X_{t+1} = A X_t + \varepsilon_{t+1}

Where:

 Xt=(Xth,Xtf)X_t = (X^h_t, X^f_t) is a vector of technology levels for both countries.

 AA is a matrix of coefficients.
 εt=(εth,εtf)\varepsilon_t = (\varepsilon^h_t, \varepsilon^f_t) is a vector of random
shocks.
 The shocks are jointly normally distributed and may be correlated across
countries at the same time.

The off-diagonal elements of matrix AA represent spillovers — that is, the degree to which
a shock in one country affects the other in future periods.
Consumers and producers in both countries observe the current state of technology (XtX_t)
when they make decisions.

The authors simplify the original Kydland-Prescott model by removing temporary shocks
and indicator variables, since those features are not central to understanding international
comovements.

Equilibrium:

To find the competitive equilibrium of this economy, the authors use the equivalence
between competitive equilibria and Pareto optimal allocations (a method developed in
earlier work by Negishi and Mantel). Because the utility function is concave, an optimal
allocation can be found by solving a central planner's problem.

The planner maximizes a weighted sum of utilities for both countries:

max⁡∑t=0∞βt[φU(cth,Ith)+(1−φ)U(ctf,Itf)]\max \sum_{t=0}^{\infty} \beta^t \left[ \varphi

U(c^h_t, I^h_t) + (1 - \varphi) U(c^f_t, I^f_t) \right]

Subject to all the relevant constraints in the model. The authors focus on the case where
φ=0.5\varphi = 0.5, giving equal weight to both countries' welfare.

They solve this problem by:

1. Substituting out the one nonlinear constraint (equation 3),

2. Approximating the objective function with a quadratic function around the steady
state,
3. Maximizing the result, subject to the model's other constraints.

Production:

In each country, the production function uses capital ktk_t, labor ntn_t, and inventories
ztz_t. Output depends on the technology shock Xt>0X_t > 0:

yt=F(Xt,kt,nt,zt)y_t = F(X_t, k_t, n_t, z_t)

Where the functional form is:

F(X,k,n,z)=[(Xkαn1−α)1−ν+μzν]1/(1−ν)F(X, k, n, z) = \left[ (X k^\alpha n^{1 - \alpha})^{1 -

\nu} + \mu z^\nu \right]^{1 / (1 - \nu)}

 0<α<10 < \alpha < 1, ν>−1\nu > -1, μ>0\mu > 0

 This formulation follows Kydland and Prescott (1988) and differs slightly from their
1982 version.
 The technology shock affects the productivity of the capital-labor composite.
World output is the sum of output from both countries. It is used for:

 Consumption,
 Fixed investment,
 Inventory accumulation.

The net exports of country ii are:

NXti=yti−(cti+xti+zt+1i−zti)NX^i_t = y^i_t - (c^i_t + x^i_t + z^i_{t+1} - z^i_t)

Capital Accumulation and Time-to-Build:

The model includes a time-to-build structure for capital formation:

 Adding to the capital stock requires investment projects that take multiple periods to
complete.
 Capital in period t+1t+1 is:

kt+1=(1−δ)kt+s1tk_{t+1} = (1 - \delta)k_t + s_{1t}

Where:

 δ\delta is the depreciation rate,

 sjts_{jt} is the number of projects started jj periods ago that are now jj periods from
completion.

Each investment project adds to capital over J periods, with equal shares of cost each
period:

λj=1Jfor j=1,…,J\lambda_j = \frac{1}{J} \quad \text{for } j = 1, \dots, J

So, fixed investment in period tt is the total cost of all investment projects that are underway:

xt=∑jλjsjtx_t = \sum_j \lambda_j s_{jt}

Here is a simplified and academically appropriate explanation of Section III: Steady State
and Parameter Values, keeping technical details intact but easier to follow:

III. Steady State and Parameter Values (Simplified Explanation)

The authors examine the behavior of their theoretical two-country economy under the
assumption that both countries are structurally identical to the closed-economy model
used by Kydland and Prescott (1982, 1988). This includes the same preferences,
technologies, and parameter values — except for the parameters that describe how
technology shocks are correlated between the two countries (these are defined by the
matrix A and the covariance matrix V).
The technology shock parameters are based on real international data, not chosen to help
the model match business cycle patterns across countries.

What Is the Steady State?

The steady state is the long-run condition of the economy when technology shocks are not
active (i.e., when their variances are zero). Most of the model’s parameters are set to match
the average behavior of the U.S. economy after World War II. Since the model is
symmetric (both countries are identical), the world economy’s steady state is just twice the
steady state of the single-country model.

In the steady state:

 Consumption, labor, capital stock, and inventories are all constant.

 The real interest rate is given by:

r=1−ββr = \frac{1 - \beta}{\beta}

 Investment equals depreciation, and inventory investment is zero.

 The resource constraint becomes:

c+δk=yc + \delta k = y

 The rental price of inventories equals the interest rate, rr.

 The cost of producing one unit of capital (in consumption terms) is:

q=1(1+r)Jq = \frac{1}{(1 + r)^J}

 The rental price of capital is then q(r+δ)q(r + \delta).

First-Order Conditions (Firms and Households)

From the firm's optimization problem, the first-order conditions determine how capital, labor,
and inventories are used. These yield a relationship (equation 9) that links the interest rate,
inventory share, and marginal products of capital and labor.

On the consumer side, the first-order condition relates the marginal utility of consumption to
the marginal disutility of labor (through the wage, ww), expressed in equation (10).

Calibrating the Model

The authors use real-world data on long-term averages and growth rates to guide the values
of the model’s parameters.
Key steady-state values:

 Consumption is 75% of output: c/y=0.75c/y = 0.75

 Inventory-to-output ratio is 1: z/y=1z/y = 1
 Quarterly real interest rate is 1%: r=0.01r = 0.01, which gives β=1/(1+r)≈0.99\beta
= 1 / (1 + r) \approx 0.99

Technology Parameters

 A Cobb-Douglas production function is used to capture how output depends on

capital and labor.
 Based on U.S. data, labor’s share of income is about 64%, so the capital share θ\
theta is set such that 1−θ=0.641 - \theta = 0.64.
 The depreciation rate is 2.5% per quarter: δ=0.025\delta = 0.025
 These values imply a capital-output ratio of about 10.
 The inventory share, based on equation (9), is set at 1%, so μ=0.01\mu = 0.01
 The parameter ν\nu, which controls how substitutable inventories are with capital-
labor, is harder to estimate directly. Based on firm-level data, Kydland and Prescott
set ν=3\nu = 3.

Time-to-Build and Preferences

 Time-to-build for capital investment is set to four periods (J = 4), following

Kydland and Prescott (1982).
 Preferences between consumption and leisure follow a Cobb-Douglas structure,
which aligns with the historical stability in average working hours, despite rising real
wages.
 The parameter λ\lambda, which determines the share of time devoted to market work,
is chosen to match 30% of available non-sleep time spent working. This leads to a
value of λ=0.34\lambda = 0.34.
 The curvature parameter γ\gamma determines risk aversion and intertemporal
substitution. Based on empirical studies, the authors choose γ=−1\gamma = -1,
which allows for non-separable preferences (i.e., consumption and leisure are not
independent).
 This non-separability is important because it helps explain the imperfect correlation
of consumption between countries — a major empirical feature not captured by
simpler models.

Unless otherwise stated, the model assumes no lagged leisure effects (i.e., α=1\alpha = 1).
However, one version of the model uses α=0.6\alpha = 0.6 and q=0.1q = 0.1, following Hotz,
Kydland, and Sedlacek (1988), to study the effect of leisure history on utility.

Technology Shock Parameters (Estimates from Data)

The interaction between foreign and domestic productivity shocks is new in the two-country
model. The authors estimate the bivariate shock process using Solow residuals (a measure
of total factor productivity) for:

 The United States

 A group of European countries (Austria, Finland, Germany, Italy, Switzerland, UK)

They compute:

log⁡A=log⁡y−(1−θ)log⁡n\log A = \log y - (1 - \theta)\log n

This avoids using capital stock data, which is less reliable and less relevant in the short term.

They estimate the process:

Xt+1=AXt+εt+1X_{t+1} = A X_t + \varepsilon_{t+1}

using quarterly data from 1970:Q1 to 1986:Q4. For the U.S. and Europe, they find:

A=[0.9040.0520.1490.908](standard errors in parentheses)A = \begin{bmatrix} 0.904 &

0.052 \\ 0.149 & 0.908 \end{bmatrix} \quad \text{(standard errors in parentheses)}

The standard deviations of the shocks:

 U.S.: 0.00906
 Europe: 0.00797

The correlation between shocks: 0.258

The eigenvalues of A: 0.994 and 0.818, indicating a fairly persistent process.

They repeat this estimation using U.S. and Canadian data and get similar results:

A=[0.7960.1311.0000.989]A = \begin{bmatrix} 0.796 & 0.131 \\ 1.000 & 0.989 \

end{bmatrix}

Standard deviations:

 U.S.: 0.00874
 Canada: 0.01023
Shock correlation: 0.434
Eigenvalues: 0.989 and 0.796

In both cases, the off-diagonal elements are positive, suggesting technology spillovers: a
shock in one country tends to affect the other over time.

Benchmark Calibration
For their main simulations, the authors use a symmetric version of the shock process, which
fits the model’s structure. The symmetric matrix A is:

A=[0.9060.0880.0880.906]A = \begin{bmatrix} 0.906 & 0.088 \\ 0.088 & 0.906 \

end{bmatrix}

 Standard deviation of shocks for both countries: 0.00852 (average of U.S. and
Europe)
 Correlation between shocks: 0.258

This choice simplifies analysis by allowing them to report results for just one country, since
both behave identically in the benchmark case.

Here is a clear and formal simplification of Section IV: Findings, preserving the technical
content while making the language more accessible for class discussion:

IV. Findings (Simplified Explanation)

The authors now examine how their theoretical two-country model behaves using the
benchmark parameter values described in Section III (and summarized in Table 3). They
simulate the model 50 times, each simulation covering 100 periods, which is roughly equal
to the number of observations used in real-world international comparisons (like in Table 2).
As in earlier sections, all the time series are detrended using the Hodrick-Prescott (HP)
filter, to isolate business cycle fluctuations.

Benchmark Results (Table 4)

Key findings from the benchmark model:

 Standard deviation of output: 1.55%, which is 91% of the actual U.S. value (1.71%
from Table 1). So, output variability in the model is reasonably close to the data.
 Relative volatility of components:
o Consumption volatility relative to output: 0.40 (vs. 0.49 in the U.S. data) →
fairly close.
o Investment volatility relative to output: 10.94, which is more than three
times what is observed in U.S. data (3.15) → highly exaggerated in the model.
 Trade balance:
o Standard deviation is about 7 times larger than in U.S. data.
o Contemporaneous correlation with output is almost zero (−0.02), unlike in
actual data where it is clearly countercyclical.
 Saving and investment rates:
o Positively correlated in the model, but not strongly.
o In real data, high-frequency movements show no clear pattern.
  International correlations:
o Output correlation between home and foreign countries is negative (−0.18),
while in the real world it is positive for nearly all country pairs.
o Consumption correlation between countries is very high in the model (0.88),
which contradicts the data, where consumption correlations are usually lower
than output correlations.

Understanding the Model’s Behavior (Figure 2)

To explain the model’s behavior, the authors analyze what happens when the home country
experiences a one-time positive technology shock.

Figure 2a (Home Country Response):

 Productivity increases, then slowly returns to normal.

 Output, investment, and consumption all rise.
 Investment increases the most, creating a trade deficit (since investment +
consumption > output).

Figure 2b (Foreign Country Response):

 Foreign productivity eventually rises due to technology spillovers, but:

o Output and investment initially fall, as resources (capital and labor) shift
toward the more productive home country.
o Foreign consumption rises slightly, even though output falls.
o This explains why foreign and domestic consumption move in the same
direction, but outputs move in opposite directions — leading to the model’s
negative output correlation.

Model vs. Reality

In summary:

 The model produces excessively volatile investment and net exports.

 It predicts too strong a consumption correlation across countries.
 It shows too weak or even negative output correlations, which contradicts actual
data.

Sensitivity to Parameter Changes (Table 5)

The authors test how sensitive these findings are to changes in model assumptions:
1. Asymmetric Spillovers

 Uses an estimated matrix A from U.S.–Europe data (rather than the symmetric
benchmark).
 Home country results:
o Investment/output correlation drops (from 0.27 to −0.08).
o Saving/investment correlation drops (from 0.28 to −0.04).
 In the foreign country, these correlations are positive (0.39 and 0.34), showing that
small changes in technology parameters can significantly affect saving and
investment relationships.

Still, investment and net exports remain highly volatile, and consumption is still more
correlated than output, which conflicts with real data.

2. Large Spillovers

 Raises the off-diagonal elements of A from 0.088 to 0.2.

 Increases the correlation between shocks from 0.258 to 0.5.
 These adjustments are stronger than what data might justify but are tested to explore
model behavior.

Results:

 Volatility of investment and net exports falls significantly (by more than 70%).
 Output correlation increases from −0.18 to 0.38 → much closer to real data.
 Consumption correlation increases further from 0.88 to 0.95 → now even farther
from actual data.

So, these changes improve some aspects of the model (e.g., output comovement) but worsen
others (e.g., consumption correlation).

3. Increased Risk Aversion

 Decreases utility parameter γ from −1 to −5 (increasing risk aversion).

 Effects:
o Small reductions in investment and net export volatility.
o Output correlation improves slightly (to −0.11).
o Consumption correlation falls from 0.88 to 0.74 → still too high compared
to output.

4. Distributed Lag on Leisure

 Introduces durability in leisure by setting α < 1.

 Results:
 o Makes output and investment more volatile.
o Volatility of hours worked increases (not shown in table).
o However, it has little impact on output vs. consumption correlations across
countries.

5. Reducing Time to Build

 Changes investment delay from 4 periods (J = 4) to 1 period (J = 1).

 Effects:
o Output volatility increases by 45% (from 1.55 to 2.24).
o Investment volatility relative to output increases dramatically (10 times
larger than in data).
 In a closed economy, time to build has little effect, but in this open-economy model,
it plays a much more significant role.

Summary of Section IV:

 The benchmark model captures some features of business cycles (e.g., output
volatility) but fails to match key international facts, especially:
o Excessive consumption correlation across countries.
o Insufficient or negative output correlation.
o Excessively volatile investment and trade flows.
 Several adjustments were tested (shock correlation, risk aversion, time-to-build), but
none fully resolved all discrepancies.

Here’s a clear and academically appropriate simplification of Section V: Trading

Frictions:

V. Trading Frictions (Simplified Explanation)

In this section, the authors explore how certain frictions (restrictions or costs) in the model's
trading system affect its ability to match real-world data.

Their intuition is that the model’s biggest discrepancies—such as too much investment
volatility, negative output correlations, and excessively high consumption correlations
across countries—stem from the fact that agents in the model can:

 Freely move capital and labor between countries,

 Fully insure themselves by trading complex financial assets (called state-contingent
claims).
This flexibility allows the model to shift too many resources to the country with the better
technology shock. That creates:

 Large investment swings

 Low or negative output correlations
 But relatively stable consumption patterns, because people smooth consumption
using international insurance.

1. Adding a Trading Friction (Transport Cost)

To test this, the authors first introduce a small transport cost—a cost for moving goods
between countries. This modifies the model’s resource constraint (which used to assume
costless trade).

 Instead of exact transport costs (which are hard to model due to their sharp “corners”),
they approximate them using a quadratic function:

G(nx)=T⋅nx2G(nx) = T \cdot nx^2

where nxnx is net exports and T>0T > 0 is a parameter that controls the size of the
cost.

 The marginal cost of trade is then:

2T⋅nx2T \cdot nx

They choose T=0.1/yT = 0.1 / y, where yy is steady-state output. This corresponds to a

0.58% cost when the economy is trading at its benchmark level of net exports.

Results with the Transport Cost (Table 5):

 Net export volatility drops dramatically:

from 2.90% to 0.16% (a 95% reduction).
 Investment volatility relative to output drops from 10.94 to 2.60.
 Output volatility falls slightly: from 1.55 to 1.38.
 Output correlation across countries improves from −0.18 to 0.02.
 Consumption correlation across countries rises slightly: from 0.88 to 0.91.

➡️Conclusion:
Adding a small transport cost greatly reduces trade and investment volatility but does not
correct the gap between consumption and output correlations across countries.

2. Autarky Experiment (No Trade at All)

Next, they go to the extreme case of autarky—a complete shutdown of trade in goods and
financial assets. The only link between countries is the correlation of technology shocks.

Results:

 Output volatility drops even further (from 1.38 to 1.33).

 All other statistics are very similar to those in the transport cost model.
 The saving-investment correlation becomes strong again.
o Without inventories, it would be exactly one (since countries cannot shift
savings abroad).
 Surprisingly, consumption correlation is still high—higher than output correlation
—even though no risk-sharing is possible through international markets.

Why is Consumption Still Correlated Without Trade?

To understand this, look at Figure 3, which shows what happens when the home country
experiences a technology shock under autarky:

 Home investment increases, but less than in the free-trade case.

 Foreign investment decreases.
 Consumption increases in both countries.

Even though there is no trade or asset market, the foreign country expects to benefit later
from the home country’s shock (due to technology spillovers). As a result, consumers
anticipate future income gains and increase consumption today. This is consistent with
the Permanent Income Hypothesis, which says people base consumption on expected
lifetime income, not just current income.

➡️So, the high consumption correlation does not result from international financial markets
alone. It also arises from expectations about future productivity.

Why Does a Small Trading Cost Have Such a Big Effect?

A key insight is that a very small friction (like a 0.58% cost) leads to dramatic changes in
the model’s behavior. Why?

The authors refer to Cole and Obstfeld (1991), who argue that if gains from trade are
small, then even small costs can discourage trade significantly.

To explore this, the authors measure the welfare gain from trade by comparing:

 Welfare under free trade (benchmark economy),

 Welfare under autarky.

They find:
  To be as well-off under autarky, consumption would have to increase by just 0.3%.
 In other words, the gains from international trade (as modeled here) are very small.

This helps explain why such a small trading cost has such a large effect—it discourages trade
because the benefits are not large enough to outweigh even minimal costs.

Final Takeaways from Section V:

 Small frictions in trade (even less than 1%) can sharply reduce volatility in
investment and net exports and moderate extreme model behavior.
 However, they do not fix the main puzzle: the model still predicts too much
correlation in consumption and not enough in output.
 The model's assumption of high spillovers and rational expectations contributes to
high consumption correlations—even when no trade is allowed.

Here is a clear and formal simplification of Section VI: Final Remarks, suitable for
academic discussion:

VI. Final Remarks (Simplified Explanation)

Real business cycle (RBC) theory has traditionally been used to study how technology
shocks affect overall economic fluctuations in closed economies (economies without
international trade or capital flows). In this paper, the authors extended the RBC framework
to an open-economy model, where:

 There is one homogeneous good,

 Labor is immobile across countries,
 Countries are allowed to trade goods and financial assets.

This extension significantly alters the behavior of macroeconomic variables compared to the
closed-economy case.

In the open-economy version of the model, the authors find that:

 Consumption correlations across countries are too high,

 Output correlations are too low,
 Investment and trade balances are far more volatile than observed in real-world
data.

When they introduced small frictions to trade (e.g., transport costs), the model showed:

 A sharp drop in investment and net export volatility,

 However, the main discrepancy—that consumption is more correlated than
output across countries—persisted.
This issue, which remains under all tested variations of the model (including with and
without trading frictions, and across many different parameter settings), is considered a
robust anomaly.

The Consumption/Output Anomaly

In real-world data, output is typically more correlated across countries than consumption.
In the model, it’s the opposite: consumption is more correlated.

Because this pattern persists despite changes in assumptions and parameters, the authors
argue that this is not just a limitation of their model—it is a deeper issue. It suggests that
current international versions of RBC models are incomplete, and further theoretical
development is necessary.

Where Should Theory Go Next?

To improve international business cycle models, researchers may need to:

 Consider additional sources of shocks beyond technology (e.g., policy shocks,

financial disturbances).
 Include relative prices and study how they behave alongside the trade balance.
 Explore how consumption and output co-move internationally in more realistic
ways.

The authors cite recent work that introduces:

 Asymmetric country sizes (Baxter & Crucini, 1991),

 Multiple types of shocks (Cardia, 1991),
 Alternative utility functions (Mendoza, 1991; Devereux, Gregory & Smith),
 More than one good being produced and traded (Ravn, 1990; Stockman & Tesar,
1990).

It is still an open question whether these extensions can solve the consumption/output
anomaly convincingly.

A Related Issue: Low-Frequency Comovements

While this paper focuses on business cycle (medium-frequency) behavior, the authors note
that low-frequency patterns also pose challenges for theory. For example:

 Many poor but fast-growing countries borrow less from rich countries than the
theory predicts.
These and other long-run inconsistencies between data and theory highlight additional
areas that need exploration in international macroeconomics.