The Dynamics of the Trade Balance: An Empirical Investigation of the Marshall-Lerner Condition and J-curve Hypothesis in Trinidad and Tobago (Working Papers, WP05/2021)

Working Papers

WP 06/2021 November 2021

The Dynamics of the Trade Balance: An Empirical

Investigation of the Marshall-Lerner Condition and

J-curve Hypothesis in Trinidad and Tobago

Ashley Bobb and Lauren Sonnylal

Research Department

This paper investigates whether an exchange rate depreciation would improve the trade balance for Trinidad and Tobago, within the theoretical underpinnings of the Marshall-Lerner (ML) condition and the J-curve Hypothesis, using data from 1991 to 2020. The ML condition posits that improvement in a country's trade balance depends on the combined responsiveness of demand for imports and exports, in response to a currency depreciation. This economic concept is complemented by the J-curve hypothesis, which suggests that a depreciation in the exchange rate may not result in an immediate increase in the trade balance but rather it worsens in the short run and gradually improves thereafter. Using an Auto-regressive Distributed Lag model, cointegration analysis, and error correction testing, the validity of these theories was examined with specific emphasis on the total and non-energy bilateral trade relationships between Trinidad and Tobago and its main trading partner, the United States (US). The empirical results confirmed the existence of the ML condition for Trinidad and Tobago in the long run; however, the improvement in the bilateral total trade balance was minimal and not sustained, thereby rejecting the J-curve hypothesis. In contrast, the combined responsiveness of the bilateral non- energy imports and exports did not satisfy the ML condition and was further corroborated by the rejection of the J-curve hypothesis. Based on the findings of the J-curve hypothesis in particular, it suggests that leveraging the exchange rate will not result in significant or sustained trade balance improvements.

JEL Classification Numbers: F10, F14, F41

Keywords: Marshall-Lerner Condition, J-curve Hypothesis, Depreciation, Imports, Exports, Trade Balance

The Working Papers Series includes papers that are primarily written by Central Bank of Trinidad and Tobago research economists to solicit comments from interested readers and stimulate discussion. The views expressed are those of the authors and not necessarily those of the Central Bank. Please send comments to [e-mail].

© Central Bank of Trinidad and Tobago, 2021

The Dynamics of the Trade Balance: An Empirical Investigation of the Marshall-

Lerner Condition and J-curve Hypothesis in Trinidad and Tobago

Ashley Bobb and Lauren Sonnylal

1.0 Introduction

Trinidad and Tobago is categorised as a small, open economy where exogenous shocks could dominate developments within its external accounts. Recent economic crises impacting the global community have underscored this vulnerability, leading to a weaker performing merchandise trade balance and by extension the current account balance. In the wake of the novel coronavirus (COVID-19) pandemic, which emerged in late-2019, Trinidad and Tobago's current account balance recorded a decline of approximately 99.0 per cent over 2020, when compared to 2019, most of which was attributed to a falloff in the net goods trading position. In addition, Trinidad and Tobago's current account recorded a deficit in the third quarter of 2020- its first deficit in approximately four years.

Apart from this ongoing shock, other negative factors that have weighed on the external accounts range from volatile commodity prices in international energy markets, subdued domestic production (particularly from the energy sector), and waning foreign demand for related products due to a growing global shift to green energy sources. The amalgamation of these circumstances, against the backdrop of a steady decline in the stock of official reserves and downward pressures on the exchange rate, has fuelled renewed interest in investigating whether depreciating the domestic currency would manifest in a favourable impact on the external current account, through a reduction in the demand for imports and an increase in exports as postulated by theory.

Currency depreciation is a conventional and text-book policy anecdote to addressing balance of payments imbalances and boosting export competitiveness. Moreover, management of a country's exchange rate is another consideration as it creates a challenge for policy-makers when formulating targeted strategies for economic development and sustainability. Therefore, the question of the role of the exchange rate, in relation to the external accounts, is a pertinent consideration for an economy such as Trinidad and Tobago that is broadly described as having a stabilised exchange rate arrangement (IMF, 2018).

Particularly noteworthy is the significant contribution of merchandise trade to the overall current account position, which is primarily supported by exports of energy commodities. In fact, energy export earnings have averaged 80.0 per cent of total exports over the period 1991 to 2020. On a trading partner basis, the United States (US) accounts for a sizeable portion of Trinidad and Tobago's trade flows, making this economy a major influencer of the domestic goods balance. Over the 30-year period, exports to the US have averaged 45.5 per cent of overall export earnings, and 34.2 per cent of foreign goods demanded were imports from the US. Meanwhile, the TT-US exchange rate has ranged from TT$5.67 in the second quarter of 1993 (the rate was floated in April 1993) to TT$6.75 at the end of 2020. Despite the gradual depreciation in the currency over the reference period, it did not induce concurrent improvements in the trade balance. The deviation in occurrences could be symptomatic of a delayed response between changes in the exchange rate and the current account.

Consequently, this paper seeks to primarily explain how Trinidad and Tobago's trade balance responds to movements in the exchange rate. Theoretically, the foundation for understanding the co-movement among these variables is derived from the Marshall-Lerner (ML) condition which postulates that the improvement of a country's trade balance, upon depreciation of the currency, depends on the combined responsiveness of the demand for imports and exports. Econometrically, this means that the absolute sum of the price elasticities must be greater than unity for the trade balance to improve. This economic concept is further complemented by an analysis of the

Central Bank of Trinidad and Tobago Working Papers-WP 06/2021 November 2021

Page 1

J-curve effect, which suggests that improvements in the trade balance, in response to a depreciation, may not occur instantaneously but it deteriorates before displaying an improvement.

This research aims to build on the work of Caribbean authors by narrowing focus to the impact of exchange rate adjustments on the trade balance of solely Trinidad and Tobago using a holistic approach of complementing the ML condition with the J-curve hypothesis. Particular emphasis is also placed on examining the relationship with the country's main bilateral trading partner- the US to eliminate aggregation bias. Higher frequency data, which covers a longer time period, is employed compared to previous studies. In this vein, the paper highlights two distinct bilateral trade balance equations to provide a thorough analysis of the depreciation effects on total US trade and also a segment of this trade balance, namely, total imports and exports, and non-energy imports and exports. The results of this individual country case could provide policymakers with insights on the effectiveness of differing exchange rate and trade policies.

The structure of the paper is as follows; Section 2.0 provides an account of previous studies, conducted at both an international and regional level, on this topic. This is followed by Section 3.0 which highlights some key characteristics of the economy of Trinidad and Tobago. Section 4.0 defines the data and methodology used to examine the ML condition and J-curve hypothesis, while an examination of the empirical results stemming from the econometric techniques employed can be found in Section 5.0. This section also expounds on an analysis of the results in reference to domestic conditions. The paper concludes in section 6.0 which summarises the main findings.

2.0 Literature Review

In the late-nineteenth century, neo-classical economist, Alfred Marshall (Marshall, 1890) laid the foundation for discussing elasticities in trade economics. This work was further expounded on by Abba Lerner (Lerner, 1952) during the twentieth century, which gave birth to the Marshall-Lerner (ML) condition. Since then, this concept has become a fundamental tenet in international economics and has been widely referred to when examining international trade relationships. Further complementing this economic framework is the existence of the J-curve effect, which was first observed by Stephen Magee in 19731. These concepts have been applied to country experiences both internationally and regionally due to its usefulness in devising trade policies and regulations for individual economies.

The ML condition seeks to determine the impact of a currency depreciation or devaluation on the trade balance of a country's balance of payments (BOP). The academic literature states that the ML condition is satisfied when the aggregate elasticities of a country's export and import demand is greater than one. In simpler terms, this economic concept theorises that an economy's trade will improve, following a currency depreciation, if the sum of the foreign elasticity of demand for exports and the home country elasticity of demand for imports is greater than one (Pilbeam, 2013). However, if the sum of these two elasticities is less than unity, then a depreciation will deteriorate the trade balance (Pilbeam, 2013). Research has also highlighted that the ML condition does not always hold in the short run, however evidence points to its existence in the medium to long run.

Depreciation or devaluation is one approach that can possibly strengthen the trade balance, but the impact of this exchange rate adjustment usually entails a time lag before showing any impact on the trade balance, resulting in varying effects in the short and long run. One such pattern created is referred to as the J-curve, which describes an initial worsening of the trade balance followed by a later improvement. Junz and Rhomberg (1973) identified five time lags: recognition, decision, delivery, replacement and production, while Magee (1973) further explains

1 Magee, Stephen P. 1973. "Currency Contracts, Pass-Through, and Devaluation." Brookings Papers on Economic Activity 4 (1): 303-325.

Central Bank of Trinidad and Tobago Working Papers-WP 06/2021 November 2021

Page 2

that these adjustment lags arise from currency-contracts signed prior to devaluation, newer currency-contracts signed after devaluation and sluggish quantity adjustments.

The transmission channel from a depreciation or devaluation of the exchange rate to the country's trade balance has been argued in the literature. Three main approaches emerged namely, the elasticities approach (EA), the absorption approach (AA) and the monetary approach (MA). The ML condition is grounded in the EA which suggests that a nominal devaluation, by exerting a negative impact on the real exchange rate of a nation's currency, will improve the global competitiveness of its tradable goods (Rawlins, 2009). Proponents of the AA focus on how devaluation may change the terms of trade, increase production, switch expenditures from foreign to domestic goods or have some other effect in reducing absorption relative to production and thus improve the trade balance (Straughn 2003). Meanwhile, in the MA, a devaluation leads to a reduction in real balances, a fall in expenditures and improvement in the trade balance (Rawlins, 2009).

The ML condition has been econometrically tested by many researchers giving rise to a vast body of literature based on country examples. However, no consensus has been reached. Among the international literature, empirical work conducted by Boyd, et al. (2001) focused on the applicability of the ML condition for eight2 Organisation of Economic Co-operation and Development (OECD) countries employ structural cointegrating vector autoregressive distributed lag (VARDL) modelling techniques. Econometric evidence point to mixed results for the countries, with five of the eight showing statistically significant responses to exchange rate effects in the long run, while J-curve effects were observed for six of the eight countries. Moreover, strong evidence was presented to establish that the real exchange rate has a significant impact on the trade balance which also supports a case for currency depreciation by policy makers.

Boyd, et al (2001) expounded on the use of aggregate trade data, however a strand of studiesalso utilised bilateral trade data. For instance, Bahmani-Oskooee and Brooks (1999) postulated that a country's trade balance could show signs of improvement with one trading partner while simultaneously deteriorating with another. Furthermore, this contradiction can also be the result of the use of the real exchange rate as opposed to the bilateral exchange rate. Therefore, an analysis on the bilateral data level would contribute positively to the topic by reducing aggregation bias. In the 1999 study by Bahmani-Oskooee and Brooks, the trade relationships between the US and six partner countries were examined and revealed varied findings. The ML condition was satisfied between the US and Japan, United Kingdom, France and Italy, while no evidence was found to support the existence of the condition for US trade with Canada and Germany. In such an instance, a depreciation of the domestic currency would favourably impact on trade balances with particular economies, therefore a blended policy response would be appropriate.

Among studies of the ML condition and J-curve hypothesis is an extensive statistical investigation conducted by the International Monetary Fund (IMF) in 2006 into the exchange rate and trade balance adjustment policies of 46 middle-income and emerging market economies. An important a priori expectation stated that the responsiveness of exporters to an exchange rate depreciation will depend on the nature of the economy's main export, therefore countries were categorised according to their dominant export; oil, non-oil commodities, or manufacturing (IMF, 2006). Results highlighted that the sensitivity of the trade balance to a depreciation in the nominal exchange rate differed among the varying exporter groups. In the short run, countries with an export-to-gross domestic product (GDP) ratio of 40.0 per cent experienced the following movement in their trade balance: a 10.0 per cent nominal exchange rate depreciation led to a 0.3 per cent of GDP improvement in the trade balance of manufactured goods exporters, compared to a 1.5 per cent of GDP improvement for commodity exporters. Over the medium term, a

10.0 per cent depreciation will result in an improvement in the trade balance of 1.5 to 2.0 per cent of GDP depending on the class of exporter. It should be noted that most of the improvement in the trade balance occurred over a 3 to 5-year period after the exchange rate depreciation. These findings indicate that the sensitivity of an economy's

2 The eight countries included in the study are: Canada, France, Germany, Italy, Japan, the Netherlands, UK and US.

Central Bank of Trinidad and Tobago Working Papers-WP 06/2021 November 2021

Page 3

trade balance dynamics to an exchange rate depreciation depends on the market structure of its exports. There is also merit in examining exchange rate policy to achieve an improved current account position.

Investigations on the ML condition were also performed on developing economies such as India (Pandey, 2013) and Pakistan (Hussain and Bashir, 2013), which both incorporated error correction models. Empirically tested conclusions drawn from both papers suggest that the ML condition held for both India and Pakistan. It should be noted that the latter tested the applicability of the ML condition and J-curve effect on disaggregated bilateral trade data between Pakistan and two prominent trading partners, the US and UK. Ultimately, the short-run and long-run period analysis pointed to the existence of both concepts with these trading partners. Meanwhile, Narayan and Narayan (2004) tested for the J-curve in Fiji using long run elasticities estimated from an ARDL, dynamic ordinary least squares (DOLS) and Fully Modified Ordinary Least Squares (FM-OLS) approaches. Similar to Pakistan, the authors found evidence of a J-curve for Fiji. In addition, the trade balance was negatively affected by domestic income and positively related to foreign income.

On the African continent, research was conducted on the import-dependent, petroleum exporting country of Nigeria by Loto (2011) and Danmola et al (2013). For the former, which used non-oil trade balance, the ML condition did not hold as the sum of the demand elasticities for imports and exports was less than unity at 0.7851, concluding that a depreciation in the Nigerian naira does not improve the trade balance. Furthermore, Loto (2011) inferred that import-dependent economies, like Nigeria, can hardly benefit from a depreciation of its currency. Meanwhile, the latter study investigated the J-curve hypothesis and confirmed its existence as the Granger causality test indicated a short-run relationship between the exchange rate devaluation and the trade balance. These authors concluded that trade needs to diversify the sources of foreign exchange apart from the petroleum sector so as to benefit from the initial devaluation. For Uganda, Kamugisha and Assoua (2020) devaluation only had a significant effect on the trade balance in the short run and thus concluded that devaluation may not be appropriate to sustainably improve its trade balance.

Regionally, Hsing (2008) studied the J-curve for seven Latin American countries using bilateral trade data and found evidence supporting its effect in Chile, Ecuador and Uruguay while evidence for Argentina Brazil, Colombia and Peru was lacking. As such, the author determined that conventional theory that supports the pursuit of real depreciation to improve the trade balance may not apply to some countries. These countries may need to exert caution in determining exchange rate policy. However, a previous study on Colombia by Rincon (1999) indicated that there was evidence of a J-curve and devaluation does improve its trade balance, which is consistent with the ML condition, and is enhanced if accompanied by reduction in the money stock and/or an increase in income. Rawlins (2009) investigated the relationship between bilateral currency depreciations, income levels and trade balance for a panel of 12 countries in Central America and the Caribbean3 and four industrialised countries: US, UK, France and Japan. Results from the Ordinary Least Squares and the Fisher-Johansen Panel Cointegrating techniques were generally mixed. All of the contemporaneous real exchange rates carried positive coefficients indicative of improvement in the trade balance in the year the devaluation was implemented, except US and the UK. However, results were tempered by insignificant t-statistics. Moreover, his study suggests that domestic and foreign income have a greater impact on the trade balance than the real exchange rate.

In terms of exclusively Caribbean studies, Boyd and Smith (2003), despite using several econometric techniques, found little satisfaction for a cointegrating BOP equation or for the ML condition in 10 Caribbean countries over the period 1980 to 2001. Countries that fulfilled the ML condition were Belize, and Trinidad and Tobago via the ARDL technique, however only at the 10.0 per cent level, and Barbados and Belize in the cointegrating Vector AutoRegression (VAR). It must be noted that the J-curve was demonstrated for St. Kitts, Jamaica and Trinidad and Tobago. The authors allude to the import and export structures of the various countries under investigation, and the combination of the short time series and noisy measures as possible reasons why improvements in the BOP

3 The 12 Central American and Caribbean economies are: Barbados, Belize, Costa Rica, Dominican Republic, El Salvador, Guatemala, Honduras, Jamaica, Nicaragua, Panama, St. Kitts and Nevis, and Trinidad and Tobago.

Central Bank of Trinidad and Tobago Working Papers-WP 06/2021 November 2021

Page 4