BUSINESS CYCLES IN SMALL OPEN ECONOMIES: EVIDENCE FROM PANEL DATA BETWEEN 1900 AND 2013

• Author: Wataru Miyamoto,Thuy Lan Nguyen
• DOI: http://doi.org/10.1111/iere.12243
• Published date: 01 August 2017
• Date: 01 August 2017

INTERNATIONAL ECONOMIC REVIEW

Vol. 58, No. 3, August 2017

BUSINESS CYCLES IN SMALL OPEN ECONOMIES: EVIDENCE FROM PANEL

DATA BETWEEN 1900 AND 2013∗

BYWATARU MIYAMOTO AND THUY LAN NGUYEN1

Bank of Canada, Canada; Santa Clara University,U.S.A.

Using a novel data set for 17 countries between 1900 and 2013, we characterize business cycles in both small

developed and developing countries in a model with financial frictions and a common shock structure. We

estimate the model jointly for these 17 countries using Bayesian methods. We find that financial frictions are an

important feature for not only developing but also small developed countries. Furthermore, business cycles in

both groups of countries are marked with trend productivity shocks. Common disturbances explain one third of

the fluctuations in small open economies, especially during important worldwide phenomena.

1. INTRODUCTION

Recent studies have examined whether economic fluctuations in developing countries are de-

scribed well by a frictionless real business cycle (RBC) model with shocks to trend productivity.

Using data for Mexico, Aguiar and Gopinath (2007) argue that a frictionless RBC model with

trend productivity shocks goes a long way in explaining business cycles in developing coun-

tries, as trend productivity shocks can capture frequent regime switches in economic policies

and market failures, which are important for these countries. At the same time, Garcia-Cicco

et al. (2010) show that an RBC model with a reduced-form financial friction describes the data

for Argentina better than a frictionless model, and that it predicts the negligible role of trend

productivity shocks in aggregate fluctuations. Given these contrasting conclusions coming from

studies with relatively few countries, several questions on the nature of the business cycle in

small open economies remain. First, what are the important features of economic fluctuations

in developing countries in general? Do trend productivity shocks have a negligible role, while

financial frictions are at the front and center? Second, are business cycles in small developed

countries any different from those in developing countries over the long horizon?

This article answers these questions in a unified framework of a structural model for small

open economies and provides new evidence on the characteristics of business cycles using a

novel panel data set covering over 100 years of data for 17 small developed and developing

countries. Our structural model is a small open economy RBC model with financial frictions

and common shocks. The financial friction feature of the model provides us with a framework

to analyze whether a frictionless RBC model with trend productivity shocks or a model with

financial frictions can better describe key features of business cycles in small open economies

over the long horizon. Importantly, unlike previous papers in the literature, we introduce a

common shock structure into the model to capture the possibility that these countries are

∗Manuscript received March 2015; revised March 2016.

1We thank Emi Nakamura, Serena Ng, Stephanie Schmitt-Groh´

e, J´

on Steinsson, and Mart´

ın Uribe for their in-

valuable advice. We also thank two anonymous referees, Jonathan Dingel, Andres Fernandez, Alex Field, Pablo

Guerron-Quintana, Chris Otrok, and seminar participants at the 2012 Midwest Macroeconomic Meeting, Econometric

Society European Meeting, and Columbia Economic Fluctuation and Monetary Colloquia for their inputs. We also

thank Chris Otrok for his help with the dynamic factor approach, and Leandro Prados de la Escosura, Alpay Filitztekin,

Ola Gryten, Gudmundur Jonsson, Jari Kauppila, Pedro Lain, Eduardo Moron, Bruno Seminario, and Jose Ursua for

providing us the data. Please address correspondence to: Wataru Miyamoto, CEA, Bank of Canada, 234 Wellington

Street West, Ottawa, ON K1A 0G9, Canada. E-mail: wataru.miyamoto1@gmail.com.

1007

C

(2017) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

1008 MIYAMOTO AND NGUYEN

subject to some common outside shocks and to better utilize the information from the panel

data of many countries. More specifically, our model includes 17 small open economies, each of

which faces a reduced-form financial friction modeled as an endogenous interest rate premium

that responds to both the level of debt-to-output ratio and expected future productivity. The

model economy is buffeted by five types of shocks, including trend and stationary productivity

shocks and country premium shocks, each of which has two components: a world common shock

that affects all countries at the same time and a country-specific shock. These common shocks

are what connect these small open economies and can be interpreted as outside shocks.

To facilitate our analysis, we estimate the model using a new data set covering 17 developing

and developed small open economies between 1900 and 2013. Compared with previous studies,

our data set includes many more countries over a much longer horizon, providing us with new

evidence on the important features of business cycles in both small developed and developing

economies, which has been limited in both sample countries and sample periods. Furthermore,

given the panel structure of the data set, we pool all available information and estimate the

model jointly for these countries. Therefore, we obtain an efficiency gain in estimating key

parameters and can identify structural shocks more accurately. Even though long data series

may contain measurement errors, the fact that these series contain several business cycles makes

them suitable for our purpose, to characterize observed business cycles and identify structural

parameters in the model, especially those related to the trend productivity shock process.

Furthermore, as we pool data in our estimation, the problem with measurement errors is less

pronounced, to the extent that these measurement errors are independent across countries.

Our joint estimation for all 17 countries using Bayesian methods indicates that financial

frictions are an important feature of business cycles in both small developed and developing

countries. In other words, a frictionless RBC model is not supported by the data. In fact, all

17 small open economies in our sample face nonzero financial frictions, although the estimated

degree of financial frictions varies across countries. Our estimation results suggest that although

it is difficult for households to smooth their consumption by borrowing internationally, their

borrowing constraint is also relaxed when their expected future productivity is high.

An important finding of our analysis is that trend productivity shocks play a sizable role in

both small developed and developing countries. In particular, trend productivity shocks explain

about one third of output fluctuations in both small developed and developing countries, on

average. This result is substantially different from previous studies on the importance of trend

productivity shocks in emerging economies, such as Garcia-Cicco et al. (2010), who find that in

an estimated model with financial frictions for Argentina, trend shocks are a negligible source

of business cycles. Our estimation suggests that Argentina is a particular case, since in other

countries such as Taiwan and Portugal, trend productivity shocks are large and significant even

though these countries face substantial financial frictions. The contribution of trend productivity

shocks is, on average, much larger than that in Argentina. Nevertheless, on average, trend plays

a less significant role than stressed in Aguiar and Gopinath (2007), who examine Mexico after

1980. This result highlights the importance of using information from many countries over the

long horizon to understand the nature of business cycles in small open economies.

Although trend productivity shocks explain a significant fraction of business cycle fluctuations

in both developed and developing countries, the natures of trend productivity shocks differ

between these two groups of countries. When we decompose the importance of trend and

stationary shocks into common and country-specific components, our estimation finds that

although important trend shocks in small developed countries are common, country-specific

trend shocks are much more dominant in developing countries. We interpret this result as

follows: Developed countries are generally closer to the world productivity frontier, so they are

more prone to common trend productivity shocks. However, developing countries are subject

to various domestic policy and structural reforms, so the trend productivity shocks that are

important for them are not common but country specific.

Another finding in our article is that common disturbances across countries are an important

driving force of business cycle fluctuations in small open economies. These common shocks

BUSINESS CYCLES IN SMALL OPEN ECONOMIES 1009

capture worldwide phenomena in the last 100 years such as the Great Depression, the two

World Wars, the two oil price shocks, and the Great Recession. During these episodes, output

in all these countries dropped at the same time. Therefore, the estimation attributes a substan-

tial fraction of business cycle fluctuations in both developed and developing countries to these

common disturbances. In particular, all types of common shocks account for roughly 28% of

output fluctuations at an annual frequency over the last 100 years. Furthermore, the extracted

world common shocks are highly correlated with U.S. output over time. For example, in the

2008–09 recession, output in Canada and Mexico, which have strong ties with the United States,

declined significantly due to common shocks. These results suggest that the identified common

shocks include the general equilibrium effects of shocks from large countries, such as the United

States, to 17 small open economies through financial and trade linkages.2Finally, we document

that several sources of common shocks contribute to the fluctuations of macroeconomic vari-

ables, including common trend and stationary productivity shocks as well as common premium

shocks.

To examine whether business cycles have changed substantially over the last 100 years, we

estimate the model for the two subsample periods before and after 1950. We find that the

estimated parameters of the model, including those related to the financial friction, change over

time, consistent with the fact that some of the second moments, such as volatilities in the data,

are different across the two subsample periods. However, the main findings of our article are

robust: Trend productivity shocks as well as common shocks play a sizable role in business

cycles in these countries, and financial frictions are still an important feature for small open

economies.

Although the identification in Bayesian estimation relies on all the information and moments

in the data, our analysis suggests that the behaviors of trade balance, as well as output and

consumption growth rates, help to pin down the importance of trend and productivity shocks.

In particular, in a frictionless RBC model, although trend productivity shocks can lead to

countercyclical trade balance and the excess volatility of consumption, trend productivity shocks

lead to a near random-walk trade balance, as discussed in Garcia-Cicco et al. (2010). Therefore,

observing trade balance is important in identifying whether a frictionless RBC model is adequate

in explaining business cycles in small open economies or whether financial frictions are an

important feature for these economies. Furthermore, observing output and consumption growth

rates over the long horizon also helps to identify the persistence of productivity shocks, and

the estimation to distinguish trend and stationary productivity shocks. We identify the common

components of these shocks through both contemporaneous and dynamic correlations across all

country pairs. In the model, since each country is a small open economy, there is no correlation

across countries without common shocks. Therefore, the estimation attributes the comovements

across all countries to world common shocks and the fluctuations independent of other countries

to country-specific shocks. This identification implies that common shocks tend to be more

important for countries that are more correlated with the rest of the countries in the sample,

which is consistent with our findings.

1.1. Related Literature. Our article is related to several strands of the macroeconomics lit-

erature. First, we contribute to a large literature in the small open economy business cycle

studies, starting with Mendoza (1991), by providing new evidence on the role of trend shocks

and financial frictions in a large number of countries.3These papers often focus on only a few

countries, such as Argentina and Mexico, and use short time series. Although Garcia-Cicco

2It is possible that our common shocks include the shocks originating from one of the 17 countries transmitting to

the rest of the countries in the sample, which can overstate the importance of common shocks. However, this bias may

be small. The reason is that since our sample includes 17 small open economies, shocks originating from Argentina or

Canada are unlikely to affect other countries such as Taiwan or India. In other words, data from small open economies

can help to avoid some of the internal propagation among countries in the group.

3A number of papers including Neumeyer and Perri (2005), Uribe and Yue (2006), Aguiar and Gopinath (2007),

Chang and Fernandez (2013), ´

Alvarez-Parra et al. (2013), Fern´

andez-Villaverde et al. (2011), and Fernandez and