Baltic-Nordic Trade Flows, Exchange-Rate Volatility, And Structural Breaks: �An Application of Nonlinear Cointegration Analysis

Scott W. Hegerty, Ph.D.

Northeastern Illinois University (USA)

Evolving Challenges in European Economies

Tallinn, Estonia

14 June 2022

Introduction

• Estonia, Latvia, and Lithuania: well-known cultural, historic, and linguistic ties with their Nordic neighbors
• Going back centuries; continuing through EU membership to today
• Many studies on bilateral trade (exports, imports, trade balances); few on Baltic; fewer on Baltic-Nordic trade
• This study: Extends work on trade balances and structural breaks�🡪 Cointegration models (linear and nonlinear)
• Finds differences by country pair and by specification
• Structural breaks affect performance of model

Some previous studies:

• Bems and Jönsson Hartelius (2006) written before the Global Financial Crisis
• Ketenci (2016) focuses outside the Baltic
• Bošnjak et al. (2020) find hysteresis in CEE time series
• Kulbacki and Michalczuk (2021) show shifts in trade patterns
• Hegerty (2022a) focuses on Baltic trade outside of Europe
• Hegerty (2022b) begins with structural breaks*�

*Hegerty, S. W. (2022). Time-series dynamics of Baltic trade flows: structural breaks, regime shifts, and exchange-rate volatility. Journal of Economics & Management, 44, 96-118. https://doi.org/10.22367/jem.2022.44.05

​

This study:

• 1. Examines structural changes in bilateral trade �🡪 Events such as European Union accession in 2004, the 2008-2009 global financial crisis, and spillovers from Russia’s first incursion into Ukraine in 2014
• 2. Evaluates how such events (e.g. capital inflows in mid 2000s) are “priced into” the real exchange rate �🡪 May capture same information as breaks
• 3. Investigates how exchange-rate volatility and regime shifts affect these trade balances �🡪 Applies different models and specifications

Main hypotheses:

• 1) Historic partnerships/events “captured” by structural breaks
• 2) Events also “priced into” the real exchange rate�🡪 Will lead to redundant effects
• 3) Ambiguous volatility effect can be uncovered by nonlinear methods

Methodology:

• Monthly time series for bilateral exports from the International Monetary Fund’s Direction of Trade Statistics (1993-2021); deseasonalized (if necessary) using the Census-X13 procedure
• Export/import ratios for Estonia, Latvia, Lithuania and Finland, Norway, and Sweden; also comparison countries Poland, Russia, and the United States 🡪18 bilateral trade balances total (3 x 6)
• Additional variables: industrial production, consumer price indices, and nominal exchange rates (vs. USD and EUR), taken from the IMF’s International Financial Statistics
• Bilateral rates via euro cross rates (1999-)

Methodology:

• Structural breaks (changes in mean) in the trade balances are extracted using the method of Bai and Perron (1998)
• Used both for informative purposes (many coincide with the GFC, for example), as well as to create dummy variables for the cointegration model
• Real exchange-rate volatility is a GARCH(1,1) process, per Bollerslev (1986), with an ARMA model for log changes in real exchange rates as the underlying mean equation
• Structural breaks in volatility are also calculated and compared against those found in the underlying trade balances

Methodology:

Methodology:

Exchange Rates and Volatility

• Appreciations, depreciations vary�Appreciations vs. Nordic, fluctuation vs. U.S. and Russia
• Mix of stationarity orders
• Volatility “spikes” in 2008; others vary by country pair

Real Exchange Rates (EE, LV)

Real Exchange Rates (LT)

Stationarity �(Real Exchange Rate, q)

​

Often I(1); less so for Estonia

Exchange-Rate Volatility

Results: Structural Breaks

• Structural breaks differ between trade balances and real exchange rates
• Similarities: 2004 EU accession🡪 regime shifts in the trade balance (but no usable GARCH model) for Estonia-Finland
• Additional common break for Estonia and Russia in mid-2016
• Common breaks across countries in 2008

Trade Balances and Structural Breaks

Trade Balances and Structural Breaks

Trade Balances and Structural Breaks

Structural Break Dates

Results: Structural Break Dummies

• Structural-break dummies 🡪 larger F-statistic �Stronger support for cointegration in both linear and nonlinear ARDL models
• Real exchange rate incorporates much of the same information regarding economic shifts�🡪 Often has a more limited impact with dummies

Cointegration Results: Comparing Models

• Most evidence of cointegration (F-test): ARDL w/dummies
• Least evidence: NARDL w/o dummies
• Real exchange rate significant in 4/18 cases:�ARDL, w/o and NARDL w/ dummies

Cointegration Results: Macro Variables

• Role of macroeconomic determinants (including exchange-rate volatility) is generally weak
• Income and relative prices do have significant effects in some cases�e.g. Domestic income: �Estonia/Norway; Estonia/U.S.; Lithuania/Poland�q (real exchange rate): Estonia/Russia
• “Disconnect” between F-test and coefficient significance

F-Test Results

Cointegration: ARDL, no dummies

Cointegration: ARDL, w/ dummies

Cointegration: NARDL, no dummies

Cointegration: NARDL, w/ dummies

Conclusions

• This study:
• Looks specifically at Baltics (rather than general CEE)
• Helps examine the timing and potential causes of economic change in the Baltic Sea region
• Focuses on exchange-rate volatility: modeling and as a determinant of bilateral trade balances
• Finds evidence of structural breaks in Baltic bilateral trade balances
• Compares cointegration results between linear and nonlinear models

Conclusions

• Findings:
• Linear ARDL model performs better than NARDL
• Including structural break dummies better than not including them
• F test more significant than coefficients
• Some significant macro effects
• In general, real exchange rate and volatility not significant
• Important pair: Lithuania/Poland (income effects are strong!)
• Future steps: focus on q; incorporate additional variables

Questions?

• Thank you!
• S-Hegerty@neiu.edu