Abstract
We propose a novel probability approach to examine the sustainability of the current account balance by generating density forecasts and calculating the probability that the current account balance will be lower than a specified threshold. We define a current account as sustainable by having a low probability of the current account deficit exploding. We use a vector autoregressive model to generate density forecasts up to five years ahead. We apply the method to ten countries that had high current account deficits in the past and find cases with both high and low probability of sustainability. We analyse historical episodes to illustrate the predictive capability of our framework and find that our method would have worked well in the past. We further find that the sustainability risk does not relate to whether the government or the private sector is the main driver of the deficit.
1 Introduction
The position of a current account balance reflects macroeconomic developments and the efficacy of trade and macro policies for open economies (Bahmani-Oskooee – Rhee 1997). Persistent current account imbalances are often viewed as a problem that requires policy action (Milesi-Ferretti – Razin 1996). While both excessive current account surpluses and deficits might reflect imbalances, deficits pose greater risks for economic stability. Persistent and excessive current account deficits, which are financed by financial account surpluses, expose countries to the risk of sudden stops and reversals in capital flows, which can lead to significant financial instability and may lead to painful and prolonged macroeconomic adjustments (Darvas et al. 2015). The literature provides evidence that countries that have experienced sudden declines in capital inflows and/or abrupt current account deficit reversals have suffered significant reductions in their rates of economic growth (Edwards 2005; Milesi-Ferretti – Razin 1996; Ghosh – Ramakrishnan 2018). Thus, in this paper, we focus on the sustainability of current account deficits.
The main challenge in analysing the (un)sustainability of a current account deficit is to determine levels of appropriate or excessive deficit relative to the country's fundamentals and desired policies over the medium term (International Monetary Fund 2019). Milesi-Ferretti and Razin (1996) posit that a current account deficit above 5% of Gross Domestic Product (GDP) warrants attention. However, the evidence is mixed. Historical episodes suggest that several countries, such as Australia, Ireland, Malaysia, and Korea, have been able to sustain current account deficits above 5% of GDP for several years, while other countries such as Chile, Hungary, Greece and Mexico have not been able to do so and have suffered severe external economic crises.
Most studies have defined the concept of current account sustainability in the form of intertemporal solvency, with the key assumptions that economic agents have perfect foresight about the economic environment and that the markets are complete (Trehan – Walsh 1991; Husted 1992; Roubini – Wachtel 1999; Obstfeld – Rogoff 1995; Hudson – Stennett 2003; Dülger 2016). However, this methodology does not consider economic uncertainty.
Milesi-Ferretti and Razin (1996) argue that the intertemporal solvency approach is not appropriate to define and gauge sustainability, because it assumes the debtor is “willing to pay” and the creditor is “willing to lend”. Their argument implies that this may not always be the case. The authors add that the framework fails to consider the role of asymmetric information in international financial markets. Moreover, the empirical evidence for using the intertemporal framework to examine the sustainability of the current account balance is, at best, mixed (Fountas et al. 1999; Dülger – Ozdemir 2005; Yol 2009; Hatzinikolaou – Simos 2013; Donoso et al. 2014; Callamand et al. 2018).
Our main argument is that the key assumption of intertemporal solvency, that economic agents have perfect foresight about the economic environment, is problematic. In a sense, economic uncertainty does exist, and its existence weakens the realism of intertemporal solvency theories. It is well known that economic uncertainty plays a role in economic life (Obstfeld – Rogoff 1995; Anzuini et al. 2020). Nonetheless, the existence of economic uncertainty should not be a controversial issue, but as noted by Cochrane (2021), the question should be, how much uncertainty is there? One way of measuring uncertainty is to derive the probability distribution of a forecast based on a forecasting model (Rossi 2014).
Current account imbalances reflect interactions between savings and investment decisions of government and domestic private sector agents, as well as the lending decisions of foreign investors. These interactions can be influenced by macroeconomic, financial, and fiscal uncertainty. For example, if government foreign borrowing is increasing rapidly under its current policy, foreign investors' perceptions may shift on whether the debt can be repaid, leading to expectations that an abrupt policy reversal might occur, which could take the form of a debt default, a large currency devaluation, or a fiscal adjustment (Milesi-Ferretti – Razin 1996). In this regard, one can argue that a question of current account sustainability does not entail a yes or no answer, but rather be viewed as a probabilistic concept.
We contribute to the existing literature by employing a novel probability approach, defining current account sustainability as having a low probability of the current account deficit exploding. Conversely, a high probability indicates unsustainability. We implement this new probabilistic concept by using density forecasts, that is, predicting the distribution of current account deficits over a five-year forecast horizon based on vector autoregressive models. This approach can be seen as a way to determine the risk of unsustainability or an early warning signal to avoid abrupt policy shifts that could lead to a reversal of capital inflows. More importantly, unlike previous literature, our approach integrates the role of uncertainty and offers a framework that allows us to quantify uncertainty surrounding the future values of the current account balance.
Our empirical work studies three issues. First, the historical analysis. We apply our method to selected historical episodes of current account deficits, some of which were followed by upward shifts in the current account, while others were not. Specifically, we study whether our method would have predicted the subsequent path of the current account correctly in the past. Second, the assessment of the latest current account data. By using the latest available 2023 data for selected countries, we evaluate whether the current account balances are sustainable. Third, the sectoral (private versus public) drivers of the current account balance. The primary rationale for this analysis is that government-driven and private sector-driven deficits might adjust differently. Analysing sectoral contributions to the current account deficit helps draw inferences as to whether the current account deficit driven by the government sector is more or less likely to be sustainable compared to a case where the most significant contributor is the private sector.
The subsequent sections introduce our methodology, discuss our data, and present our empirical analysis. The last section offers concluding remarks.
2 Methodology
We measure the uncertainty surrounding the future values of the current account balance by deriving the probability distribution of forecasts using a forecasting model. Regarding the model, the abrupt response of current account dynamics to capital flow shocks (or vice-versa), the role of uncertainty in those dynamics, and the possibility that government-driven and private sector-driven deficits may adjust differently, all point to potential nonlinearities in the data-generating process. Such dynamics are often complex, and linear models may not fully capture the varying impacts depending on the size or nature of the shocks. Therefore, we begin by testing for nonlinearity before specifying our model.
2.1 Testing nonlinearity
We employ the Brock-Dechert-Scheinkman (BDS) Test, proposed by Brock, Dechert, Scheinkman and LeBaron (1996) to test the presence of nonlinearity in residuals. The BDS method tests the null hypothesis of no significant nonlinear dependence in residuals, against the alternative of significant nonlinear dependence. Table 1 presents the test results, showing the dimension (which represents the number of time lags considered in detecting the dependence) and the corresponding P-values for each country in our sample.
The Brock-Dechert-Scheinkman (BDS) test results
| Dimension | Australia | Chile | Greece | Hungary | Ireland | Korea | Malaysia | Mexico | South Africa | United States |
| 2 | 0.1550 | 0.5798 | 0.4013 | 0.2007 | 0.1515 | 0.3263 | 0.1333 | 0.0301 | 0.0032 | 0.2500 |
| 3 | 0.5980 | 0.5018 | 0.9645 | 0.9425 | 0.3398 | 0.0753 | 0.7336 | 0.1252 | 0.0002 | 0.7116 |
| 4 | 0.0295 | 0.0534 | 0.2407 | 0.5152 | 0.0092 | 0.1155 | 0.5385 | 0.1673 | 0.0024 | 0.9457 |
| 5 | 0.0005 | 0.0180 | 0.0167 | 0.4167 | 0.1043 | 0.2323 | 0.2311 | 0.0568 | 0.0001 | 0.7951 |
| 6 | 0.0338 | 0.0045 | 0.0622 | 0.4593 | 0.0515 | 0.5461 | 0.0821 | 0.0013 | 0.0000 | 0.6089 |
Source: author.
Notes: Table 1 displays the P-values at the 5% significance level, along with the corresponding dimensions.
Our findings show mixed evidence of nonlinearity for Australia, Chile, Greece, Ireland, Korea, Mexico, and Malaysia. For the United States and Hungary, we fail to reject the null hypothesis of no significant nonlinear dependence, suggesting no compelling evidence of nonlinearity. However, for South Africa, we reject the null hypothesis in favour of the alternative, indicating the presence of nonlinearity.
Overall, the BDS test results suggest that for most countries, there is no overwhelming evidence to justify that a linear model is inadequate for capturing the dynamics of the variables in our analysis, and to sufficiently capture the complexity of the data. Thus, we proceed with a linear model.
We determine whether our current account forecasts suggest a sustainable path by calculating the probability that the future value of the current account balance will be lower than a specified threshold. When this probability is low, the current account deficit is unlikely to explode, indicating sustainability. Conversely, a high probability throughout the forecast horizons, or a considerably increasing probability as the forecast horizon increases, indicates unsustainability.
We conducted an extensive exercise on current account reversals for 192 countries representing various income groups. Our findings suggest that current account reversal episodes are often preceded by significant current account deficits, averaging around −8% of GDP (see Fig. 1). Thus −8% was used as the threshold. Additionally, we also include the −5% threshold which has been popular in the literature.
The average performance of current account balance pre- and post-reversal and non-reversal episodes
Source: author's calculations, based on IMF World Economic Outlook Database.
Notes: To identify the reversal episodes, we adopted criteria similar to those proposed by Ferretti and Razin (1996). Defining reversal episodes: the current account deficit must narrow by at least 3 percentage points of GDP on average from the three years preceding the reversal episode to the four-year period consisting of the reversal year and the subsequent three years. Defining non-reversal episodes: the current account deficit must not narrow by more than 1.5 percentage points of GDP on average from the three years preceding the reversal episode to the four-year period consisting of the reversal year and the subsequent three years. Reversal and non-reversal episodes were identified in the sample period from 1983 to 2023. The reversal and non-reversal years are denoted by t. The three years before reversal/non-reversal years are denoted as
Citation: Society and Economy 2025; 10.1556/204.2024.00019
It is important to highlight that the use of an absolute value of the threshold, for example, −8% is simply a practical tool for illustrating our proposed probability approach. The key takeaway of our argument is not the absolute value of the threshold itself, but the insights gained from understanding how probabilities change over the forecast horizon. Thus, the focus in this paper should remain on the proposed probability approach rather than the specific threshold used. We would need country-specific thresholds if our focus is on the absolute value of the probability. We would also need a good justification for the selection of these thresholds, but this is not the case here, as this paper focuses on how these probabilities evolve over different forecast horizons.
Instead of approximating the probability distribution of the forecasts with an analytical formula, which might be sensitive to whether the VAR variables have unit roots or not,1 we derive forecast uncertainty from past forecast errors based on a recursive out-of-sample forecasting exercise. That is, we retain the first n years of the sample period for an initial estimation of the model, create forecasts up to 5 years ahead, and compare these forecasts to actual outcomes to calculate the forecast errors for 1, 2, …, 5 years ahead. For example, for the analysis of the sustainability of the latest 2023 current account deficits, we first estimate the VAR model on 1980–2004 data and calculate forecasts for 2005–2009. Next, we increase the estimation period by one year (1980–2005) and calculate forecasts and forecast errors for the subsequent five years ahead (2006–2010). And so on, we continue up to the last observation of the sample.
We thus estimate a series of forecast errors for the variables included in the VAR model over five forecast horizons, 1, 2,…, 5 years, which we use to perform bootstrapping in further analysis. There are two key reasons for choosing the bootstrapping method instead of assuming a normal distribution in errors. First, diagnostic tests following the linear VAR model estimation, suggest weak evidence of a normal distribution in the data generating process for most countries in our sample. Second, the literature reveals that abrupt adjustments in the current account are not rare (Adalet – Eichengreen 2005). In this regard, assuming a Gaussian distribution may not be appropriate. To proceed, we resample one million random draws using the forecast errors for each forecast horizon separately. These random draws are added to our point forecasts, to create a simulated series, which we then use to characterise the empirical density function of our forecasts. We then use these forecast distributions to calculate the probability that the future values of the current account balance will fall below a specified threshold.
This process helps us to understand the distribution of probable future current account values and capture the level of uncertainty associated with our forecasts.
3 Data
We use annual data collected from the International Monetary Fund, World Bank, and Bruegel Institute. Our list of variables includes the current account balance (% of GDP), GDP growth rate, trade openness (% of GDP), CPI-based real effective exchange rate index, fiscal balance (% of GDP), and old age dependency ratio. These variables are often used in the literature to understand current account dynamics. For example, the old age dependency ratio is a structural fundamental that has been used to understand the effect of demographic dynamics on the current account balance (see International Monetary Fund 2022).
Our data period spans from 1960 to 2023. We use a set of 10 countries for the historical validation of our method and the assessment of the latest data, while for the sector analysis, we are bound to use a different set of countries due to data availability.
For the historical validation of our method and the assessment of the latest data, we consider countries that in the past had persistent current account deficits, some of which adjusted, while others did not:2
- (i)Episodes in which sustained current account deficits have not triggered a sharp current account or policy reversal. This includes countries such as Australia (1981–2018), Malaysia (1979–1986), South Africa (2003–2019) and the United States (1982–2023, except 1991).
- (ii)Episodes in which external or domestic factors have caused a sharp policy reversal to shield the current account balance, but have not led to an external economic crisis. This includes experiences of countries such as Ireland (1979–1990), and Korea (1978–1988).
- (iii)Episodes in which persistent current account deficits have been followed by an external crisis, resulting in debt rescheduling, renegotiations, or a massive bailout. This includes countries such as Chile (1977–1982), Mexico (1991–1994), Hungary (2008–2010) and Greece (2010–2018).
Although the list may not be exhaustive, the selected countries constitute a diverse set to study the sustainability of their historical and 2023 current account deficits.
To analyse the impacts of sectoral drivers of current account deficits, we had to use a different set of countries due to data availability at sectoral level. Due to the availability of net financial balance data at the sectoral level, we use a sample of ten European Union countries that have had protracted current account deficits in the past. The annual data for these countries is sourced from Eurostat for the 1980–2023 period.
4 Results
4.1 The validation of our method based on historical current account episodes
We analyse the historical episodes of current account imbalances for selected countries to test whether our methodology would have predicted the (un)sustainability of current account deficits in subsequent years. For example, the United States had a persistent current account deficit for decades, which fluctuated little without any major upward movements, suggesting that the deficit was sustainable. On the contrary, the Greek current account underwent a major upward adjustment after 2010, suggesting that the earlier deficit was not sustainable.
We therefore estimate our VAR model using data up to 2009 for Greece and calculate out-of-sample forecasts for 2010–2014. We quantify the uncertainty of these forecasts by calculating historical out-of-sample forecast errors for 1995–2009. We use these forecast errors to perform a bootstrapping method for each forecast horizon from one year to five years to quantify the uncertainty of our out-of-sample forecasts for 2010–2014. We then calculate the probability that the 2010–2014 current account balances will be lower than the threshold of −8%. If this probability is high, or if it is low for short forecast horizons but increases substantially at longer forecast horizons, then this indicates a high likelihood that the current account had to be adjusted. Conversely, there would be no need for an adjustment if the probability is low. We then examine whether the current account did indeed adjust after 2009 and if our model foresaw this adjustment accurately. This entire process is applied to all other countries in the sample, analysing their respective episodes of current account imbalances.
Table 2 shows that Greece and Hungary had an average of over 80% probability of their deficits falling below the threshold of −8% in the periods ahead, indicating unsustainability. Conversely, the rest of the countries exhibited relatively low probabilities across different episodes and forecast horizons, indicating sustainability, except South Africa, for which the 2013 episodes suggested a low probability at the one-year forecast horizon, but a high probability (approximately over 60%) over the three-year to five-year forecast horizons.
The probability that the current account balance will be lower than a threshold value
| Threshold: | −8 | Forecast period | ||||
| 1 | 2 | 3 | 4 | 5 | ||
| Country | Observed CA balance | Probability | Probability | Probability | Probability | Probability |
| Australia (2018) | −2.22 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| Greece (2009) | −10.87 | 0.8668 | 0.5706 | 0.7696 | 0.9171 | 1.0000 |
| Hungary (2007) | −7.37 | 0.7502 | 0.8175 | 0.7999 | 1.0000 | 1.0000 |
| South Africa (2013) | −5.33 | 0.0665 | 0.2854 | 0.6157 | 0.5830 | 0.6371 |
| South Africa (2019) | −2.6 | 0.0473 | 0.4997 | 0.6843 | 0.7774 | 0.8233 |
| United States (2008) | −4.72 | 0.0000 | 0.1768 | 0.6262 | 0.8667 | 0.8572 |
| United States (2012) | −2.57 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.2223 |
Source: author's calculations.
Notes: The figures in the parenthesis represent the year for which we calculate the probability of the current account balance falling below the threshold of −8%. For Hungary, the VAR model excludes trade openness and fiscal balance due to data limitations. Likewise, Chile, Ireland, Korea, Malaysia, and Mexico were excluded due to a lack of data availability pre-1980.
Focusing on the adjustment of the selected historical episodes, we observe the following. For Greece, the high probabilities generated by our model justify the subsequent significant adjustments in their current account deficit. Specifically, before 2009, Greece consistently recorded large deficits, which adjusted sharply from −10.1% in 2011 to −2.5% in 2012. Between 2013 and 2014, the current account balance stabilised, ranging between −2.5% and −3%. Similarly, Hungary's current account deficit reversed sharply from −7.4% in 2007 to −0.8% in 2009 and turned into a surplus between 2010 and 2012.
For South Africa, after 2013, there was a modest adjustment with the current account deficit declining from −5.3% in 2013 to −2.4% in 2017. Our model suggested a low probability of adjustment in the short-term, but a rather high probability (around 60%) over the medium- to longer- term, so the upward adjustment is largely consistent with the model prediction. For 2019, the probabilities are surprisingly high for the most part of the forecast horizon, given a relatively small deficit of −2.6%. In fact, the country's current account balance flipped to a surplus of 1.9% and 3.7% in 2020 and 2021, respectively. Thereafter, it slipped back into a deficit of −0.5% and −1.6% in 2022 and 2023, respectively, with an IMF projection of −1.8% for 2024.
The unexpected surplus in the current account balance in 2020–2021 could have been explained by lower domestic demand resulting from the recession caused by the COVID-19 pandemic – an unpredictable event. On the other hand, the easing of COVID-19 lockdown restrictions and the related rebound in global economic activity supported the bouncing back of South Africa's export volumes in 2021 (South African Reserve Bank 2021). This highlights the need for careful assessment when the model prediction does not materialise.
The United States' current account deficit persisted without significant adjustments, consistent with our low probabilities that the country's current account deficit may fall below the threshold of −8% in the near term, except in the much longer term, such as 4 or 5 years ahead (over 60% for the 2008 episode). The country recorded a deficit of −4.7% in 2008 and -2.6% in 2012 and has since remained around −2.5%. Overall, this analysis demonstrates the model's effectiveness in predicting the need for adjustments in current account balances and underscores the importance of monitoring economic conditions to inform policy decisions.
4.2 The assessment of 2023 current account balances
We find that for the majority of countries in our sample, the probability that the current account deficit will fall below the threshold of −8% is extremely low, except for Chile (Table 3). For instance, Australia has practically zero probability (rounded to four digits after the decimal) that the country's current account balance will be lower than the threshold throughout the forecast horizons. This is not surprising as since 1980 the country's deficit has hovered around 6% (which is above the threshold of 8%) for over three decades before switching to a surplus in 2019. In 2023, the current account had a surplus of 1.2% and thus it is extremely unlikely that it will fall below −8%.
The probability that the current account balance will be lower than a threshold value
| Threshold: | −8 | Forecast period | ||||
| 1 | 2 | 3 | 4 | 5 | ||
| Country | 2023 CA Balance | Probability | Probability | Probability | Probability | Probability |
| Australia | 1.22 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| Chile | −3.55 | 0.1574 | 0.4437 | 0.5882 | 0.6248 | 0.7332 |
| Greece | −6.94 | 0.3153 | 0.3340 | 0.3529 | 0.2500 | 0.2000 |
| Hungary | 0.26 | 0.0524 | 0.2216 | 0.1769 | 0.2506 | 0.3331 |
| Ireland | 9.87 | 0.0527 | 0.1112 | 0.1179 | 0.1873 | 0.2001 |
| Korea | 2.07 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| Malaysia | 1.20 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.1331 |
| Mexico | −0.32 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| South Africa | −2.97 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| United States | −1.62 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
Source: author's calculations.
Notes: For each forecast horizon, we calculate the probabilities according to equation (2).
Some of the main characteristics of the Australian external position may explain the low probability of the country's current account balance falling below the threshold. Australia's external debt has been denominated in domestic currency, the Australian dollar. This created a buffer against fluctuations in the exchange rate and shielded the economy from increases in the domestic currency value of external liabilities in the event of a currency depreciation. In addition, foreign direct investment (FDI) has been stronger than portfolio investment (PI) flows in the last two decades. FDI is a more stable source of financing than PI (Darvas et al. 2015), thereby making the Australian economy less prone to external risks.
Similarly, Korea, Malaysia, Mexico, South Africa and the United States also have extremely low probability that their respective current account balances of 2.07%, 1.20%, −0.32%, −2.97% and −1.62% will fall below an 8% deficit. This result suggests that these economies are on sustainable paths. Notably, these low probability values are broadly justified by their latest balances of 2023, which are indeed small deficits, and some are in surplus. Moreover, it can be argued that the salient features of the US economy, including the dollar's reserve currency status and the size of the economy, play a role in sustaining its persistent deficit.
In stark contrast, Chile faces a markedly different outlook, with a probability exceeding 60% that its current account balance will fall below the −8% threshold, particularly on longer forecast horizons. This trajectory suggests a high risk of unsustainability in the long term. While Greece also shows substantial probabilities of unsustainability, these probability values appear to be gradually declining over the forecast horizon. This indicates that Greece's current account balance might improve in the years ahead, a trend supported by the projections reported by the International Monetary Fund (2024).
Ireland's substantial 2023 surplus of 9.87% reflects low short-term probabilities of unsustainability, but these probabilities increase as the forecast period extends, with an estimated 12% risk in three-year forecasts, 19% in four years, and 20% in five years. These results indicate potential vulnerabilities in the long-term outlook, which are consistent across both the baseline and −5% deficit thresholds. Large past forecast errors and resulting high uncertainty in current account forecasts likely explain why these probabilities remain significantly above zero, even if the 2023 current account balance was in a substantial surplus.
For Hungary, which had a near-balanced position in 2023, the probabilities are low in the short-term but increase considerably through the end of the forecast period. Specifically, the probability of an unsustainable current account balance is just 5% one year ahead but increases to 33% by the five-year mark. This pattern is also consistent at the −5% threshold. The forecast errors for Hungary were larger, more negative, and more volatile than for Mexico, another country with a near-balanced current account in 2023. This volatility and negativity in Hungary's forecast errors likely explain why the probabilities for Hungary are significantly higher than those for Mexico.
It appears that Ireland and Hungary may be able to manage their current account balances in the short-term. However, there are underlying vulnerabilities that could pose challenges over the longer term, as the probabilities of unsustainable deficits are gradually increasing over the forecast horizon.
Notably, larger initial deficits correlate strongly with higher long-term unsustainability risks, as seen with Greece and Chile (Tables 3 and 4). The exception is Ireland, which has a large surplus in its last observation. These results indicate that while large deficits contribute to the risk of unsustainability, country-specific circumstances may matter, as seen in the case of Ireland. Ireland's situation highlights the complexity of current account dynamics: even substantial surpluses can carry risks of unsustainability, illustrating the need to account for individual national contexts. Therefore, in addition to observing the current account balance itself, it is important to consider unique national circumstances to better understand its trajectory and the associated potential risks. Similar inferences can be drawn for Hungary.
The probability that the current account balance will be lower than a threshold value
| Threshold: | −5 | Forecast Period | ||||
| 1 | 2 | 3 | 4 | 5 | ||
| Country | 2023 CA balance | Probability | Probability | Probability | Probability | Probability |
| Australia | 1.22 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| Chile | −3.55 | 0.5789 | 0.7775 | 0.8823 | 1.0000 | 0.9334 |
| Greece | −6.94 | 0.7367 | 0.5555 | 0.4121 | 0.4368 | 0.4666 |
| Hungary | 0.26 | 0.1582 | 0.4448 | 0.4714 | 0.5008 | 0.4665 |
| Ireland | 9.87 | 0.0524 | 0.1110 | 0.2347 | 0.1873 | 0.2001 |
| Korea | 2.07 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| Malaysia | 1.20 | 0.0525 | 0.0000 | 0.0592 | 0.1877 | 0.4000 |
| Mexico | −0.32 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 |
| South Africa | −2.97 | 0.0000 | 0.0555 | 0.2358 | 0.1253 | 0.0665 |
| United States | −1.62 | 0.0000 | 0.0000 | 0.1180 | 0.0625 | 0.0000 |
Source: author's calculations.
Notes: For each forecast horizon, we calculate the probabilities according to equation (2).
4.3 The probability of current account sustainability: government-driven versus private sector-driven current account balance
We find no strong evidence that a government-driven current account balance is inherently more or less sustainable than those driven by the private sector (Tables 5 and 6). For instance, while Bulgaria and Cyprus, both with private-sector-driven current account balances, exhibit relatively high probabilities of unsustainability, countries with government-sector-driven balances, such as France, Italy, and Spain, show significantly lower risk probabilities (Table 5).
The probability that the current account balance will be lower than a threshold value
| Threshold: | −8 | Forecast period | |||||
| 1 | 2 | 3 | 4 | 5 | Largest contributing sector to the CA deficit | ||
| Country | 2023 CA balance | Probability | Probability | Probability | Probability | Probability | |
| Bulgaria | 0.29 | 0.0000 | 0.1114 | 0.1177 | 0.0628 | 0.0000 | NFC |
| Cyprus | −9.29 | 0.7894 | 0.7779 | 0.8235 | 0.7496 | 0.6666 | HH |
| France | −0.75 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | GOV |
| Greece | −6.94 | 0.3153 | 0.3332 | 0.3527 | 0.2504 | 0.2003 | GOV |
| Hungary | 0.26 | 0.0529 | 0.2778 | 0.4715 | 0.4373 | 0.4667 | GOV |
| Ireland | 9.87 | 0.0526 | 0.1114 | 0.1178 | 0.1880 | 0.2003 | NFC&GOV |
| Italy | 0.16 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | GOV |
| Portugal | 1.37 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | NFC&GOV |
| Romania | −7.10 | 0.5264 | 0.4444 | 0.4698 | 0.4380 | 0.3997 | GOV |
| Spain | 2.60 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | GOV |
Source: author's calculations.
Notes: We have excluded some of the EU deficit countries due to data and estimation limitations. The “Largest contributing sector to the CA deficit” includes households (HHs), the government (GOV), nonfinancial corporates (NFCs) and the financial corporate sectors (FCs).
The probability that the current account balance will be lower than a threshold value
| Threshold: | −5 | Forecast period | |||||
| 1 | 2 | 3 | 4 | 5 | Largest contributing sector to the CA deficit | ||
| Country | 2023 CA balance | Probability | Probability | Probability | Probability | Probability | |
| Bulgaria | 0.29 | 0.0523 | 0.1672 | 0.1176 | 0.0626 | 0.0664 | NFC |
| Cyprus | −9.29 | 0.8950 | 0.8889 | 0.8236 | 0.7500 | 0.7997 | HH |
| France | −0.75 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | GOV |
| Greece | −6.94 | 0.7367 | 0.5557 | 0.4111 | 0.4371 | 0.4672 | GOV |
| Hungary | 0.26 | 0.2635 | 0.4994 | 0.5880 | 0.4999 | 0.5326 | GOV |
| Ireland | 9.87 | 0.0524 | 0.1107 | 0.2353 | 0.1872 | 0.2002 | NFC&GOV |
| Italy | 0.16 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | GOV |
| Portugal | 1.37 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | 0.0000 | NFC&GOV |
| Romania | −7.10 | 0.8420 | 0.7778 | 0.5890 | 0.5621 | 0.5334 | GOV |
| Spain | 2.60 | 0.0000 | 0.0556 | 0.1174 | 0.0627 | 0.0000 | GOV |
Source: author's calculations.
Notes: We have excluded some of the EU deficit countries due to data and estimation limitations. The “Largest contributing sector to the CA deficit” includes households (HHs), the government (GOV), nonfinancial corporates (NFCs) and the financial corporate sectors (FCs).
In detail, France, Italy, and Spain, where government sector surpluses dominate, show minimal probabilities of falling below the −8% current account threshold. The probabilities range from 0% to 7% across all forecast horizons, indicating a stable outlook with low risk of unsustainability. In contrast, Bulgaria and Cyprus, with private sector-driven deficits, exhibit higher probabilities of unsustainability. For Cyprus, which has a substantial current account deficit of −9.29% in 2023, shows considerably higher probabilities of unsustainability, between 30% and 80%, which we find consistent with its larger deficit. For Bulgaria, the probabilities range from 11% to 22%, somewhat justified by its small surplus of 0.29%.
Greece and Romania, both with government-sector-driven deficits of −6.9% and −7.1%, respectively, present contrasting risk profiles. Romania's unsustainability risk is notably higher, while Greece's probability values hover at moderate levels, mostly around one-third, indicating a relatively lower risk than Romania, despite comparable deficit levels.
Ireland's current account balance, driven primarily by non-financial corporates and the government sector, demonstrates low probabilities of unsustainability, supported by its substantial surplus of 9.87%. Nevertheless, Ireland's probability values increase steadily over the forecast horizon, suggesting an emerging vulnerability highlighted earlier in Tables 3 and 4. Portugal, which also holds a surplus (1.37%), aligns with lower probabilities of unsustainability, further supporting the alignment of probability outcomes with current account balances in the short to medium term.
Overall, we find that the probability outcomes align well with the most recent current account balances, particularly in the short to medium term. More importantly, we find strong evidence that sectoral differentiation (government versus private) does not significantly influence the sustainability assessment of current account balances.
5 Concluding remarks
The paper studied the sustainability of current account deficits with a novel method. The existing research has largely relied on an intertemporal solvency framework. A major shortcoming of this framework is the lack of consideration of economic uncertainty for which the economic agents are subject to. We were the first to use a probability approach, which allows us to quantify uncertainty and study the probability distribution of the current account forecasts. We define a current account as sustainable by having a low probability of the current account deficit exploding.
We used a vector autoregressive model to generate forecasts up to five years ahead for the selected countries. We quantified the uncertainty of these forecasts based on historical out-of-sample forecasting errors. To test our methodology, we applied it to several past episodes of current account deficits and found that our model would have done a good job in detecting unsustainable developments in the past for our selected sample.
By analysing the sustainability of the 2023 current account balance of selected countries, we found cases with both high and low probability of sustainability. Ireland and Hungary show short-term sustainability in current account balances but may face potential long-term vulnerabilities, emphasizing the need for continual monitoring and analysis of current account balances to detect unsustainability risks. Additionally, our results suggest that country-specific factors, like the case of Ireland, may influence the trajectory of current account adjustments. We further found no strong evidence that a government-driven current account deficit is more or less likely to be sustainable than one driven by the private sector.
Our approach provides a nuanced understanding of current account sustainability, integrating the role of uncertainty and offering a probabilistic framework that allows us to quantify uncertainty surrounding the future values of the current account.
Further research could refine our method, for example, by using alternative forecasting models. Additionally, our mixed results from the nonlinearity test warrant considering nonlinear models to capture any aspects of the data generating process that would be better explained by such a model.
Acknowledgements
I extend my heartfelt gratitude to my supervisor, Dr. Zsolt Darvas, for his invaluable guidance, constructive feedback, and unwavering support throughout the development of this paper. I also thank the participants at the XII PhD Workshop, hosted by the University of Pécs in April 2024, and the 8th International Conference on Applied Theory, Macro, and Empirical Finance, organised by the University of Macedonia in April 2024. Their insightful comments and suggestions have significantly enriched this work.
References
Adalet, M. – Eichengreen, B. (2005): Current Account Reversals: Always a Problem? In: Clarida, R. H. (ed): G7 Current Account Imbalances: Sustainability and Adjustment. Chicago: University of Chicago Press.
Anzuini, A. – Rossi, L. – Tommasino, P. (2020): Fiscal Policy Uncertainty and the Business Cycle: Time Series Evidence from Italy. Journal of Macroeconomics 65: 103–238.
Bahmani-Oskooee, M. – Rhee, H. J. (1997): Are Imports and Exports of Korea Cointegrated? International Economic Journal 11(1): 109–114.
Callamand, O. D. – Melo-Velandia, L. F. – Arana, M. V. (2018): Asymmetric Behaviour of Current Account Sustainability in Latin America. Internal Finance 21(1): 2–22.
Cochrane, J. H. (2021). Rethinking Production under Uncertainty. The Review of Asset Pricing Studies 11(1): 1–59.
Darvas, Z. – Hüttl, P. – Merler, S. – De Sousa, C. – Walsh, T. (2015): Analysis of Developments in EU Capital Flows in the Global Context. Study for the DG MARKT of the European Commission.
Donoso, V. – Martin, V. (2014): Current Account Sustainability in Latin America. Journal of International Trade and Economic Development 23(5): 735–753.
Dülger, F. (2016): The Sustainability of Current Account in the Presence of Endogenous Multiple Structural Breaks: Evidence from Developed and Developing Countries. Panoeconomicus 63(3): 339–358.
Dülger, F. – Ozdemir, Z. A. (2005): Current Account Sustainability in Seven Developed Countries. Journal of Social Economics Research 7(2): 47–80.
Edwards, S. (2005): Is the US Current Account Deficit Sustainable? and if Not, How Costly Is Adjustment like to Be? NBER Working Paper No 11541.
Fountas, S. – Wu, J. L. (1999): Are the US Current Account Deficits Very Sustainable? International Economic Journal 13(3): 51–58.
Ghosh, A. – Ramakrishnan, U. (2018). Current Account Deficit: Is There a Problem? Finance and Development. https://www.imf.org/external/pubs/ft/fandd/basics/pdf/ghosh_current-account-deficits.pdf, accessed 21/11/2024.
Hamilton, J. D. (1994). Time Series Analysis. Princeton, New Jersey: Princeton University Press.
Hatzinikolaou, D. – Simos, T. (2013): A New Test for Deficit Sustainability and its Application to US Data. Empirical Economics 45(1): 61–79.
Hudson, S. – Stennett, R. (2003): Current Account Sustainability in Jamaica. Bank of Jamaica Working Paper 2: 1–43.
Husted, S. (1992): The Emerging US Current Account Deficit in the 1980s: a Cointegration Analysis. The Review of Economics and Statistics 1992: 159–166.
International Monetary Fund (2019): External Balance Assessment Methodology 2018 Update. Working Paper No. 2019/065.
International Monetary Fund (2022): External Balance Assessment Methodology 2022 Update. Working Paper No. 2023/047
International Monetary Fund (2024): World Economic Outlook April 2024. Washington DC: IMF.
Miles-Ferretti, G. M. – Razin, A. (1996): Sustainability of Persistent Current Account Deficits. NBER Working Papers No 5467.
Obstfeld, M. – Rogoff, K. (1995): The Intertemporal Approach to the Current Account. Handbook of International Economics 3: 1731–1799.
Rossi, B. (2014): Density Forecasts in Economics, Forecasting and Policymaking. Barcelona: The Centre de Recerca en Economia International.
Roubini, N. – Wachtel, P. (1999): Current–account Sustainability in Transition Economies. In: Blejer, M. I. – Škreb, M. (eds): Balance of Payments, Exchange Rates, and Competitiveness in Transition Economies. Boston, MA: Springer.
South African Reserve Bank (2021): South African Reserve Bank Report 2021. https://www.resbank.co.za/content/dam/sarb/publications/quarterly-bulletins/current-account/2021/Current%20account%20release%20September%202021.pdf, accessed 21/11/2024.
Trehan, B. – Walsh, C. E. (1991): Testing Intertemporal Budget Constraints: Theory and Applications to US Federal Budget and Current Account Deficits. Journal of Money, Credit and Banking 23(2): 206–223.
Yol, M. A. (2009): Testing the Sustainability of Current Account Deficits in Developing Economies: Evidence from Egypt, Morocco, and Tunisia. The Journal of Developing Areas 43(1): 177–197.
Note that a VAR can be estimated consistently, and used for forecasting, irrespective of whether the variables have unit roots or not; see Hamilton (1994).
We borrowed this categorisation from Milesi-Feretti and Razin (1996). The various experiences are characterised in terms of the macroeconomic policy stance taken, macroeconomic and the structural features of the country's economy, and the external shocks affecting the economy (Milesi-Feretti – Razin 1996). The years in brackets indicate the specific episodes.
