Chapter

Chapter 2. Trade Finance and Trade Flows: Panel Data Evidence from 10 Crises

Author(s):
Jian-Ye Wang, and Márcio Ronci
Published Date:
February 2006
Share
  • ShareShare
Show Summary Details
Author(s)
Marcio Ronci10

This chapter assesses the effect of constrained trade finance on trade flows in countries undergoing financial and balance of payments crises. Most of the countries that had a major external crisis had a significant trade contraction, while trade-related finance declined sharply (Figure 2.1).

Figure 2.1.Outstanding External Short-Term Credit

((Observations centered around the crisis year, in percent of trade)

Sources: World Bank, Global Development Finance, and IMF, World Economic Outlook.

1Average of the 10 countries.

Despite anecdotal evidence that the contraction of trade financing may have affected trade,11 there has been to date only a few empirical studies assessing the effect of constrained trade finance on trade flows. In addition, trade may have also been affected by other variables such as world demand, domestic demand, banking crises, changes in export and import prices, and real exchange rate depreciation.12

A closer look at the data does not provides a clear-cut relationship between trade and trade financing. Table 2.1 summarizes trade indicators, external short-term credit (as a proxy for trade financing—see the section in this chapter on data), and real exchange rates for 10 crisis countries. Although overall export and import values in U.S. dollars fell, only import volumes contracted sharply, by 20 percent on average, while export volumes increased by 10 percent on average (albeit slightly below the three-year trend growth of 11.7 percent preceeding the countries’ crises).

Table 2.1.Trade, External Short-Term Credit, and Real Exchange Rate in 10 Crises(Annual percentage change)
     Outstanding External 
 Values in U.S. DollarsVolume IndexesShort-Term CreditReal Exchange
 ExportsImportsExportsImports(in U.S. dollars) 1Rate
Weighed average21997–98−4.3−29.110.6−21.3−19.9−25.6
Malaysia−7.3−26.64.5−17.8−41.9−20.6
Philippines16.9−18.619.4−19.0−37.5−18.5
Thailand−6.8−33.88.5−27.5−19.6−15.6
Indonesia−8.5−27.43.1−11.2−37.2−51.7
Korea−4.7−36.219.6−23.1−46.4−25.7
Russia−14.3−19.43.9−18.072.53−11.5
1998–99

Brazil
−6.1−14.77.7−4.9−3.0−33.6
2000–01

Argentina
0.8−19.84.6−17.3−30.11.9
Turkey1994–9511.9−26.815.6−23.0−44.1−17.8
Mexico14.0−21.48.3−25.5−8.4−33.1
Sources: IMF, World Economic Outlook and International Financial Statistics, and World Bank, Global Development Finance.

Deflated by U.S. whole industrial price.

Volume changes were weighed using exports (imports) share in the total exports (imports).

The increase in short-term credit during the crisis was largely due to a gas pipeline project under the Black Sea.

Sources: IMF, World Economic Outlook and International Financial Statistics, and World Bank, Global Development Finance.

Deflated by U.S. whole industrial price.

Volume changes were weighed using exports (imports) share in the total exports (imports).

The increase in short-term credit during the crisis was largely due to a gas pipeline project under the Black Sea.

Some observers argue that the sharp decline in import volumes and the slowdown in export volume growth are closely related to the collapse of trade financing, as external outstanding short-term credit to crisis countries fell by 20 percent in real terms compared to precrisis levels. However, the decline in trade financing seems to have had little effect on export volumes, while the fall in import volumes could have been caused by the sharp real devaluation and fall in domestic demand that followed each crisis (Table 2.1 and Figure 2.2).

Figure 2.2.Real Effective Exchange Rates

(Observations centered around the crisis year, in percent of exports)

Source: IMF, International Financial Statistics.

1Average of the 10 countries.

In addition to trade finance, other factors may have affected trade volumes, including exchange rates, relative prices, and domestic and external demand. To control for the various factors that may have affected trade flows during crises, we estimate export and import volume equations including trade financing as an explanatory variable.

This chapter describes the data used and their limitations, discusses model specification and econometric estimation, and presents the estimation of the export and import volume equations and the trade volume-to-trade finance elasticities.

Data

Table 2.2 presents the definitions of variables used in this chapter. We used as a proxy for trade financing flows the change in outstanding short-term credit in U.S. dollars as reported in Global Development Finance (GDF),13 which includes short-term credit for trade as reported by the Organization for Economic Cooperation and Development (OECD) and the international banks’ short-term claims as reported by the Bank for International Settlements (BIS). However, using GDF short-term credit as a proxy for trade financing has limitations: it excludes trade financing associated with intrafirm trade by multinational corporations (including most processing trade), and trade related to foreign direct investment.14 Also, trade financed by domestic banking sources may not be responsive to external trade financing reported in the BIS statistics. We used a dummy variable for domestic banking crisis as trade financing supply is also related to the ability of domestic banks to intermediate foreign trade financing.

Table 2.2.Summary of the Variables1
VariablesDescription
logXj,tLogarithm of Export volumes of country j at time t
logMj,tLogarithm of Import volumes of country j at time t
logXWj,tLogarithm of World trade volume index
logYWj,tLogarithm of World GDP index
logYj,tLogarithm of GDP of country j at time t
logRELPXj,tLogarithm of Relative price index of exports
logRELPXj,tLogarithm of Relative price index for imports
FINj,tFirst difference of logarithm of outstanding short-term credit to country j at time t, Dj,t
DUMMYj,t"1" for domestic banking crisis, and "0" otherwise for Argentina, Indonesia, Mexico and Russia

See Appendix 2.1 for sources and definition of variables.

See Appendix 2.1 for sources and definition of variables.

The panel data consists of 10 countries over 10 years, which yields a sample of 100 observations. We were constrained to use annual data, as most of the variables have annual frequency. Also, we did not include more annual observations, as we are interested in the trade finance effects on trade around the crisis year, and we would expect that observations far away from the crisis year would add little information on trade finance on trade flows during crisis.

We tested all variables for each country (Table 2.3) for unit roots and found a fair amount of disagreement among the different tests, which may be partly due to the sample period being relatively short.15 There is some evidence that most variables are nonstationary in levels but for FIN.16

Table 2.3.Unit Root Tests1
 LevelsFirst difference
VariablesIm, Pesaran and ShinProb.PP—Fisher Chi-squareProb.Im, Pesaran and ShinProb.PP—Fisher Chi-squareProb.
logXj,t−0.62720.265322.88770.2943−0.07350.470734.83960.0210
logMj,t0.53130.702411.37020.93610.25630.601239.51700.0057
logXWj,t0.16050.563728.68420.0942−0.73090.232458.46480.0000
logYWj,t0.00230.500981.38170.0000−0.75700.224534.23040.0246
logYj,t0.73780.769714.93830.7799−0.13490.446354.14830.0001
logRELPXj,t0.13550.553923.61170.2598−1.22540.110255.94340.0000
logRELPMj,t0.05490.521924.21120.2333−0.81120.208649.28450.0003
FINj,t−0.91600.179891.95350.0000−0.53370.2968111.0730.000

Null hypothesis: Unit root assuming individual unit root process (see Kim and Maddala, 2001, pp. 134–37).

Null hypothesis: Unit root assuming individual unit root process (see Kim and Maddala, 2001, pp. 134–37).

An overview of the data shows that in most countries, external short-term credit fell significantly in real terms following the crisis year (Figure 2.3), while export volumes continued to grow (Figure 2.4) and import volumes fell (Figure 2.5).

Figure 2.3.Real Outstanding External Short-Term Credit

(Observations centered around the crisis year, in billions of 2000 U.S. dollars)1

Source: IMF, World Economic Outlook.

1 Deflated by U.S. whole industrial price.

Figure 2.4.Export Volume Indexes

(Observations centered around the crisis year = 100)

Source: IMF, World Economic Outlook.

1Average of the 10 countries.

Figure 2.5.Import Volume Indexes

(Observations centered around the crisis year = 100)

Source: IMF, World Economic Outlook.

1Average of the 10 countries.

However, the sharp depreciation of national currency increased export and import relative prices (Figures 2.6 and 2.7), which may have boosted exports and weakened imports.

Figure 2.6.Export Relative Price Indexes

(Observations centered around the crisis year = 100)

Source: IMF, World Economic Outlook.

1Average of the 10 countries.

Figure 2.7.Import Relative Price Indexes

(Observations centered around crisis year = 100)

Source: IMF, World Economic Outlook.

1Average of the 10 countries

Also, at the time of the crisis, the countries did not face a fall in world demand, as world gross product continued increase (Figure 2.8). This certainly contributed to support exports, while most of the sample countries faced a sharp contraction of their GDP, leading possibly to lower demand for imports (Figure 2.9).

Figure 2.8.World Real GDP Index for Each Country at the Time of its Crisis

(Observations centered around the crisis year)

Source: IMF, World Economic Outlook.

1Average of the 10 countries.

Figure 2.9.Countries' Real GDP Index

(Observations centered around the crisis year = 100)

Source: IMF, World Economic Outlook.

1Average of the 10 countries.

Model Specification and Econometric Estimation Methodology

The data presented in the previous section suggest that other factors may have affected export and import volumes during crises in addition to trade finance. To control for the various factors that may have affected trade flows during crises, we estimate export and import volume equations, including trade financing as an explanatory variable, using panel data with observations for each variable centered on the crisis year.

Our basic equations have the following simple specifications:

where t- time annual observations centered around the crisis year (t = −4, −3, −2, −1, 0, +1, +2) and t = 0 year in which crisis began, where j—country (Argentina, Brazil, Indonesia, Malaysia, Philippines, Russia, South Korea, Thailand, Turkey, and Mexico), where M and X are import and export volumes, RELPX and RELPM are the export and import relative price indexes, FIN is trade-related finance, Y domestic demand, XW world trade volume index, and DUMMY is a dummy for domestic banking crisis (equal 1 for banking crisis and 0 if is not the case). 17

The error terms u and v are assumed to have zero mean and constant variance and not autocorrelated. The expected coefficient signs for the export equation are: a1>0, a2>0, a3>0 anda4<0. The expected coefficient signs for the import equation are: b1> 0, b2< 0, b3> 0, andb4< 0,.

As the unit root tests suggest that most of variables are nonstationary in levels (Table 2.3), we estimated the first difference of equations (1) and (2), including two lags for each first differenced variable.

We estimated equations (1) and (2) using generalized least squares (GLS), instrumental variables (IV), both with fixed effects and generalized method of moments (GMM). The GLS recognizes the nonsphericalness of the error terms u and v and is more efficient than LS, particularly in the case of heteroskedasticity. The IV and GMM estimation addresses simultaneity and errors in variable measurement. In particular, measurement error in the trade finance variable may be serious, as there is no reliable data source. Finally, we tested all restrictions on the coefficients of equations (1) and (2) by means of Wald tests to determine a more parsimonious model specification, including the fixed effects assumption.

Estimation Results

Estimation of equations (1) and (2) suggests that trade finance affects both export and import volumes in addition to relative prices and income. Trade financing explains a relatively small part of the fall of trade flows in recent crisis, as trade volume elasticities to trade financing are small, while a fall in trade financing in connection with domestic banking crisis can lead to a substantial loss of trade.

Tables 2.4 and 2.5 summarize the estimation results.18 Overall, all variables have the expected signs and most of the coefficients are significant at 1 and 5 percent levels. IV and GMM estimates do not differ significantly from GLS estimates, indicating that the results are relatively robust. The statistic Durbin-Watson suggests there is no autocorrelation, and the common intercept hypothesis is rejected at 5 percent.

Table 2.4.Export Volume Equations

(Dependent Variable: ΔlogXj,t)

Explanatory VariablesGLS Fixed Effects (1)IV1 Fixed Effects (2)GMM1 Fixed Effects (3)GMM1 Fixed Effects (4)
ΔlogXWjt0.2731 ***
ΔlogXWjt−10.2892 *0.4133 ***0.2953 ***
ΔlogRELPXjt0.0377 ***0.0420 **0.0180 ***0.0638 ***
ΔFINjt0.0177 ***0.0135 ***
ΔFINjt−10.0387 **
Dummy−0.0666 ***−0.0752 ***−0.0170−0.0550 ***
AR(1)0.1167 ***0.1170 ***
Number of observations80808080
R-squared0.750.750.310.30
Durbin-Watson stat.2.0312.3242.3342.072
Common intercept F-test 24.020 **3.000 **
Notes: (*) significant at 10 percent level, (**) significant at 5 percent level, and (***) significant at 1 percent level. We defined ΔlogZt = ΔlogZt—ΔlogZt−1.

Instruments: lagged world demand, real domestic credit and real exchange rate, and dummy for banking crisis.

The common intercept restriction rejected at (**) 5 percent level.

Notes: (*) significant at 10 percent level, (**) significant at 5 percent level, and (***) significant at 1 percent level. We defined ΔlogZt = ΔlogZt—ΔlogZt−1.

Instruments: lagged world demand, real domestic credit and real exchange rate, and dummy for banking crisis.

The common intercept restriction rejected at (**) 5 percent level.

Table 2.5.Import Volume Equations

(Dependent Variable: ΔlogMjt)

Explanatory VariablesGLS

Fixed Effects (5)
IV

1 Fixed Effects (6)
GMM

Fixed Effects (7)
GMM

2 Dynamic (8)
ΔlogMt−1−0.0489 *
ΔlogYj,t2.4571 ***1.7757 ***1.9086 ***1.8337 ***
ΔlogRELPMjt−0.2024 ***−0.1782 **−0.1453 *−0.1267 ***
ΔFINjt0.14200.1213 *0.10260.0798 ***
Dummy−0.0891 *−0.1141 **−0.1106 **−0.1138 ***
Number of observations90909080
R-squared0.770.690.690.5300
Durbin-Watson stat1.9982.2442.2512.9690
Common intercept F-test34.500 **3.0600 **
Notes: (*) significant at 10 percent level, (**) significant at 5 percent level, and (***) significant at 1 percent level. We defined ΔlogZt = ΔlogZt—ΔlogZt−1.

Instruments: lagged world demand, real gross domestic product, real domestic credit, real exchange rate, trade finance, and dummy for banking crisis.

Linear dynamic panel data estimation (Arellano-Bond, 1991).

The common intercept restriction rejected at (**) 5 percent level.

Notes: (*) significant at 10 percent level, (**) significant at 5 percent level, and (***) significant at 1 percent level. We defined ΔlogZt = ΔlogZt—ΔlogZt−1.

Instruments: lagged world demand, real gross domestic product, real domestic credit, real exchange rate, trade finance, and dummy for banking crisis.

Linear dynamic panel data estimation (Arellano-Bond, 1991).

The common intercept restriction rejected at (**) 5 percent level.

Trade financing affects both export and import volumes positively, as expected, but its coefficients are relatively small. The elasticity of export volume with respect to trade financing is estimated between 0.02 and 0.04, and is statistically significantly different from zero, while the elasticity of import volume with respect to trade financing is about 0.10, and statistically significantly different from zero in two of the four regressions. The coefficient of the dummy variables for domestic banking crises is relatively large and significant in both equations. The dummy variable explains about 7 percent and 10 percent of the fall in export and import volumes, respectively, compared with precrisis volumes in those countries affected by domestic banking crisis.

Conclusions

Our results suggest that trade finance positively affects both export and import volumes as well as relative prices and income in the short run. A fall in trade financing in connection with a domestic banking crisis can lead to a substantial loss of trade, while trade financing (narrowly defined here as externally provided short-term bank credit) explains a relatively small part of the fall of trade flows in recent crises, as elasticities of trade volumes with respect to trade financing are small.

Table 2.6 summarizes trade financing effects on export and import volumes. A domestic banking crisis has a large effect on exports and imports, possibly because domestic banks are not able to intermediate foreign trade financing. The domestic banking crisis dummy explains a fall in exports of about 6 percent and in imports of about 10 percent compared with precrisis levels. In contrast, the estimated elasticities are small, and a fall of 20 percent in trade finance—as shown in Table 2.1—explains only a decline of about 1 percent in exports and 2 percent in imports. The low elasticities of trade volumes with respect to trade financing reflect the fact that a large part of exports is financed either from domestic sources or outside the banking system. As a result, export volumes are not very sensitive to changes in available data on externally provided short-term bank trade credit.

Table 2.6.Summary of Trade Financing Effects on Trade
 Export VolumesImport Volumes
Elasticity of trade volumes with respect to trade financing

(centered at the crisis year)1
0.030.08
Dummy variable for domestic banking crisis

(change in percent compared with pre-crisis volumes)
−5.5−11.0
Sources: Tables 2.4 and 2.5.

εt=0 ≈α [1−(ΔDt/Dt−1/(ΔDt+1 /Dt)], where ε is the coeficient of FIN in the regressions, D is the outstanding stock of short-term debt, and t=0 is the crisis year.

Sources: Tables 2.4 and 2.5.

εt=0 ≈α [1−(ΔDt/Dt−1/(ΔDt+1 /Dt)], where ε is the coeficient of FIN in the regressions, D is the outstanding stock of short-term debt, and t=0 is the crisis year.

These results provide some justification to policies aimed at supporting trade financing during a crisis, particularly when domestic banks are in distress and are unable to intermediate foreign trade financing. At the same time, they point out that disruptions in trade financing explain part of the total fall of trade flows in recent crises, and that other policies are needed to address each country’s external vulnerabilities, in particular large macroeconomic imbalances, banking system distress, low external reserves, and unsustainable external debt.

Appendix 2.1. Variables and Data Sources

Import and export in U.S. dollars as reported in International Financial Statistics (IMF).

Import and export volume indexes (Mj, t and Xj, t) as reported in World Economic Outlook (IMF).

Real effective exchange rates (ER) and nominal exchange rate (E), national currency per U.S. dollar as reported in International Financial Statistics (IMF).

Export price indexes (PX)- Price deflator for exports of goods as reported in World Economic Outlook (IMF) for each country j.

Import price indexes (PM) - Price deflator for exports of goods as reported in World Economic Outlook (IMF) for each country j.

Wholesale price index (WPI) as reported in International Financial Statistics (IMF).

External short-term credit (FIN) as reported in Global Development Finance (World Bank). This variable was used as proxy for trade finance. The stock of short-term credit in the GDF is calculated adding information on banks’ short-term claims by country from the Bank for International Settlements (BIS) and the short-term credit for exports from the Organization for Economic Cooperation and Development (OECD). As the BIS data are reported in terms of remaining maturity, the GDF adjusts the BIS data to obtain an estimate of banks’ claims of one-year maturity. Both institutions report short-term claims/credit in U.S. dollars.

Domestic demand (Y)—Gross domestic product, constant prices as reported in World Economic Outlook (IMF) for each country j.

World demand (YW)—Trade weighed demand as reported in World Economic Outlook (IMF) for each country j. This variable was used as an instrument in the IV estimation of the chapter’s equations.

World trade index (XW)—Volume of exports of goods & services as reported in World Economic Outlook, (IMF) for each country j

Relative export price index (RELPX) is defined as export price index divided by wholesale price index and multiplied by the exchange rate: RELPX = (PX/WPI)*E

Relative import price (RELPM) is defined as import price index divided by wholesale index and multiplied by the exchange rate: RELPM = (PM/WPI)*E

Real domestic credit (DC) is defined as the nominal domestic credit as reported in International Financial Statistics (IMF) deflated by the consumer price index. This variable was used as an instrument in the IV estimation of the chapter’s equations.

References

    ArellanoM. and S.R.Bond1991“Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,”Review of Economic StudiesNo. 58 pp. 27797.

    Bank for International Settlements1995Guide to the BIS Statistics on International BankingBasle.

    CastroAlexandre Samy de and Marco AntonioCavalcanti1998“Estimação de equações de exportação e importação para o Brasil—1955/95,”Pesquisa e Planejamento EconômicoVol. 28 (April) pp. 168.

    FingerandK. Michael and LudgerSchuknecht1999“Trade Finance and Financial Crisis” Special Studies No. 3 (Geneva: World Trade Organization).

    FullertonThomas M.W. CharlesSawyer and Richard L.Sprinkle1999“Latin American Trade Elasticities,”Journal of Economics and FinanceVol. 23 (Summer) pp. 14356.

    GoldsteinMorris and Mohsin S.Khan1985“Income and Price Effects in Foreign Trade,” in Handbook of International EconomicsVol. IIed. by R.W.Jones and P. B.Kenen (Amsterdam: Elsevier Science Publishers) pp. 1041105.

    ImKyungsoM. HashemPesaran and ShinYongcheol2003“Testing for Unit Roots in Heterogeneous Panels,”Journal of EconometricsVol. 155 pp. 5374.

    International Monetary Fund2003“Trade Finance in Financial Crisis—Assessment of Key Issues,”SM/03/395 (December) unpublished.

    KimIn-Moo and G.S.Maddala2001Unit Roots Cointegration and Structural Change (Cambridge: Cambridge University Press).

    MaddalaG.S. and S.Wu1999“A Comparative Study of Unit Tests with Panel Data and New Simple Test,”Oxford Bulletin of Economics and StatisticsVol. 61 pp. 63152.

    PortugalMarcelo1993“A instabilidade dos parâmetros nas equações de exportação brasileiras,”Pesquisa e Planejamento EconômicoVol. 23 (August) pp. 31348.

    ResendeMarco Flavio2001“Crescimento econômico, disponibilidade de divisas eimportações no Brasil: um modelo de correção de erros,”Pesquisa e Planejamento EconômicoVol. 31 (August) pp. 289330.

    ResendeMarco Flavio1997“Disponibilidade Cambial e Especificação da Função de Demanda de Importações para o Brasil” Texto para Discussão No. 506 (Rio de Janeiro: Instituto de Pesquisa Economica Aplicada).

    Serge JeanneauSerge and MarianMicu2002“Determinants of International Bank Lending to Emerging Market Countries,”BIS Working Paper No. 112 (Basle: Bank for International Settlements).

    WangY.K1983“The Econometric Models of South Korea and Indonesia” (Taipei: The Institute of Economics, Academia Sinica) pp. 30383.

    Other Resources Citing This Publication