Policy Responses to Balance-of-Payments Crises: The Role of Elections

Governments have a number of policy tools that can be used to address pressure on the balance of payments, threatening an undesirable decline in the relative value of the national currency. They can: (1) sell reserves, (2) raise interest rates, (3) impose capital controls, (4) apply trade restrictions, or (5) depreciate the currency. While researchers typically analyze these policies in isolation from one another, we treat them as a menu of options available to election-minded politicians. We analyze the use of these five policy responses to payments difficulties for a large sample of countries since the early 1970s. We argue that governments try to minimize political costs by adopting less transparent policies first and only moving to more visible policies as necessary, delaying the most visible and politically costly policies until after elections. The evidence is consistent with these claims: governments are more likely to draw down reserves and impose capital controls before other options. If these policies do not succeed, they tend to raise interest rates. If further action is needed, they delay devaluations and trade protection until after elections.

during the Great Depression can be understood as a macroeconomic policy response induced by the Btrilemma^of the gold standard.
History thus illustrates that there is wide variation over time and across countries in the policies governments adopt to deal with payments crises. The existing literature, however, is ill-suited to account for this variation. While many researchers recognize that the Bpolicy trilemma^imposes constraints on crisis policy choice, few studies analyze government choices across the full spectrum of policy options that they have at their disposal. 4 There is a large literature on the political economy of exchange rates, for example, but it is largely disconnected from the literatures on trade policy, capital controls, interest rates, and reserves. 5 Apart from Forbes and Klein (2015) the paper that pioneered the analysis of the causes and consequences of different policy responses to payments crisesscholars either study policies in isolation from one another or investigate connections between a limited number of policy areas. For example, several recent papers explore the relationship between exchange rates and trade policy (Broz and Werfel 2014;Bown and Crowley 2013;Copelovitch and Pevehouse 2013). In a similar vein, Stefanie Walter explores why some countries devalue their currencies during a balance-ofpayment crisis while others opt for Binternal devaluations^-the combination of high interest rates, fiscal austerity, and structural reforms that seek to regain competitiveness by lowering wage costs and increasing productivity (Walter 2015;Walter 2013). While moving in the right direction, these previous works neglect to consider all the policy instruments that authorities have at their disposal during crises. 6 Since multiple policies can help restore external equilibrium, we follow Forbes and Klein (2015) and analyze them jointly. Yet our paper is distinctive in a number of ways. First, and most important, whereas Forbes and Klein (2015) assess the economic covariates of policy responses to sudden stops in international capital flows, we take an explicitly political economy approach to this topic by focusing on how elections shape the choice and sequencing of policies. Second, while Forbes and Klein (2015) concentrate on four crisis policy responses (i.e., reserve sales, currency depreciations, interest rate increases, and capital controls), we add trade policy as a fifth potential response to crises. As Keynes (1931) recognized during the Great Depression, trade policy can substitute for exchange rate policy because the trade impact of devaluation is equivalent to a uniform tariff on all imports plus a uniform subsidy on all exports. Beyond this equivalence, we also base our arguments on the trilemma: devaluation allows monetary policy to be used to reduce interest rates and increase domestic demand while trade measures do not free up interest rate policy. Third, we document and analyze crisis responses back to the early 1970s, greatly expanding the historical coverage of Forbes and Klein (2015), who focus on crisis responses from 1997 to 2011.
Our focus on elections is related to the literature on opportunistic political business cycles which emphasizes how electoral calendars shape the policy choices of officeseeking politicians. 7 One strand of this literature shows that incumbent politicians 4 The trilemma-that fixed exchange rates and open capital markets mean a loss of monetary policy autonomyis explored empirically by Aizenman et al. (2010) and Obstfeld et al. (2005). 5 For a survey of the exchange-rate policy literature, see Broz and Frieden (2008). Rodrik (1995) provides a review of the trade policy literature. 6 We do not consider fiscal austerity and structural reforms in this paper because these policies tend to be more difficult to implement in the short-run context of a payments crisis. However, an analysis of medium-term responses to crises would include these policies in the menu of adjustment options open to policymakers. 7 For reviews, see Franzese (2002) and Drazen (2000). manipulate fiscal and monetary policies before elections to signal their competence to voters (Rogoff 1990;Rogoff and Sibert 1988). Another strand establishes that the proximity of elections induces politicians to become more protectionist (Conconi et al., 2014;Karol 2007). We show that electoral calendars also affect politicians' choice of balance-of-payments policies during crises.
The paper is organized as follows. In the next section, we present our theoretical expectations regarding which policy instrument authorities will turn to during a crisis, as well as the order in which these policies are implemented. Here we focus on delineating how the timing of elections influences the choice of policy instruments, and we base our predictions on the relative unpopularity and transparency of these instruments to voters. With the option to BPick their Poison^-the title of Forbes and Klein's (2015) paperwe expect governments to first select temporizing and opaque policies, such as reserves drawdowns and capital controls, prior to elections, and to postpone more transparent and unpopular policies, like devaluations and import surcharges, until after elections. In Section 3, we describe how we measure Bpolicy responses,^which follows Forbes and Klein (2015), and provide descriptive statistics of the patterns of responses that we find in the data. Section 4 provides statistical tests of our political economy arguments and Section 5 concludes.

Theory: Explaining the Choice of Responses
Policymakers faced with pressure on the payments balance have a variety of options available to them. While all may help relieve payments pressure, each option operates differently and has somewhat different economic effects. As a result, each option may also have different political economy implications. In this section, we focus on how elections relate to each policy alternative, highlighting how politicians choose among policies that vary in their unpopularity with, and transparency to, voters. Since any policy response to a crisis is likely to have some negative effects, the consequence is that countries must Bpick their poison^ (Forbes and Klein 2015). Building on this insight, we expect policymakers to choose less unpopular and/or more opaque policies before elections and delay the more unpopular and transparent policies until after elections.
In a crisis compelled by a sudden stop of international capital flows, all policy alternatives have negative macroeconomic effects. For example, a policy-induced increase in the short-term nominal interest rate that aims to restore balance-of-payments equilibrium causes aggregate output and employment to fall, while a devaluation of the currency for the same purpose reduces national purchasing power and aggravates inflation as imports become more expensive. Inasmuch as all policy options have negative macroeconomic consequences, our focus differs from that of the political business cycle literature, which assumes that incumbent politicians can expand the economy in the pre-election period by way of monetary stimulus (Nordhaus 1975;McRae 1977) or fiscal (Keech and Pak 1989;Drazen 2000). In a balance-of-payments crisis, policymakers are restricted to choosing the least politically costly policy from a set of bad options.
We assume that politicians will try to minimize the political costs of dealing with a payments crisis. We expect them to try options that have lower political cost first, only moving to higher-cost options if these options fail to stem the crisis. Furthermore, we expect that politicians will delay higher-cost options until after elections.
BPolitical cost^is a function of the transparency and immediacy of the policies' negative economic impact. In this section, we assess each of the five policy options in terms of their observability to voters and the speed and intensity of their impact on voters. We reason that before elections, politicians will prefer policies whose negative consequences are less observable to voters or are delayed until after elections take place. When more than one policy is used, we argue that politicians will implement the least visible and least objectionable policies first, before they move on to more unpopular policies.

Selling Reserves
We think governments will turn first to foreign exchange intervention, in which the central bank seeks to counter a currency crisis by selling reserves. Exchange market interventions by the central bank are generally unobserved by the electorate and can be an effective means of reducing pressure on the exchange rate. But a central bank has finite stocks of foreign exchange reserves that can be depleted rapidly. Hence, the sale of reserves might end up being a transitory measure with no lasting effect on currency values, unless it is supported by other policies, such as an interest rate increase that signals the intended path of future market interest rates. Policymakers, seeking to avoid more unpopular policies, like devaluation or generalized trade protection, may thus sell reserves when they first face serious exchange market pressure. Given that the policy buys the government some time, we also expect reserve sales to be more likely before elections, so that more politically onerous policies can be delayed.

Capital Controls
Although there is continuing debate over the efficacy of capital controls, especially in times of crisis, 8 we abstract from this issue simply to assume that capital controls do have some impact on capital flows and domestic monetary conditions, specifically to reduce outflows and keep interest rates lower than they might otherwise be. However, this impact on interest rates is indirect and less likely to be observed by the electorate, hence less politically costly, than other policies. Like using reserves to prop up the currency, capital controls allow the government to retain some monetary policy breathing space, which is most politically valuable in the run-up to elections. So we expect governments to be more likely to impose capital controls before elections and thereby delay the imposition of more unpopular policies.

Raising Interest Rates
Tightening monetary policy to stanch capital outflows or encourage capital inflows has more direct and transparent effects on voters than reserve sales or capital controls, by way of its impact on aggregate demand and employment. When the central bank raises nominal interest rates for balance-of-payments reasons, it translates into increases in real interest rates as well due to sticky prices (Mishkin 1996). Firms, finding that their real cost of borrowing over all horizons has increased, cut back on their investment expenditures. Likewise, households facing higher real borrowing costs scale back their purchases of homes, automobiles, and other durable goods. These cutbacks in investment and spending cause aggregate output and employment to fall, which is unpopular with voters. 9 However, there can be a significant lag before interest rate changes influence spending and saving decisions, and thus have an impact on the overall economy. The time lag is variable, but research suggests that it can take six to eight quarters for a change in the central bank policy rate to have its full effect on the economy (Goodhart 2001). This extended lag suggests that a policymaker could mount a defense of the currency with an interest rate increase within, say, two or three quarters of an election without worrying too much about the negative impact on economic growth. Hence, if governments decide to mount an interest rate defense, they will do so aheadperhaps well aheadof impending elections.

Trade Barriers
The GATT/WTO has long accepted the right of member governments to use import restrictions to address balance-of-payments problems, and has actively regulated exceptions to GATT/WTO obligations that are justified on this basis (Eglin 1987;McCusker 2000;World Trade Organization 2009). The policy has been invoked dozens of times in GATT/WTO history by both developing and developed countries, including by the United States in 1971, which imposed a 10 % temporary import surcharge to address its payments problems (Stewart and Drake 2009). According to GATT/WTO rules, such import restrictions must not exceed those necessary to stop or forestall a serious decline in reserves, must be temporary, and must be broad-based, as in an Bimport surcharge,^which is a uniform tax applied to all or most imports. This last requirement is meant to prevent abuses, whereby a member government justifies protecting a particular domestic industry or sector on the basis of balance-of-payments considerations. Given that import surcharges raise the domestic price of all or most imports and import-competing goods, this policy option is likely to be unpopular. 10 Hence, we expect that policymakers will avoid using import surcharges. But when they do, we predict the policy will be adopted after less visible and less politically costly policies are implemented, and after elections, when voters cannot directly sanction the government for the pain that generalized trade protection brings.

Devaluation or Depreciation
The consequence of a depreciation for the electorate is an immediate and transparent reduction in purchasing power and real wages, as the prices of imports, and related goods, increase in domestic-currency terms. It is therefore not surprising that devaluations and large depreciations significantly lower governments' approval ratings and lead to large drops in their chances of being re-elected (Cooper 1971;Bernhard and Leblang 2006;Frankel 2005). Devaluations are politically costly to leaders. We therefore expect devaluations and large deprecations to be relatively rare. However, when devaluations do occur, governments will do all they can to delay them until after elections. When an election looms, governments will try to postpone devaluations, whether to get re-elected, 9 A key finding of the economic voting literature is that when the economy prospers voters reward the incumbent government, but when the economy falters they punish incumbents (Lewis-Beck and Stegmaier 2000; Hellwig 2010). 10 For a welfare analysis of import surcharges, see Congressional Budget Office (1985). or so that their successor will get the blame, or out of hope that something will happen to relieve pressure on the balance of payments (Blomberg et al. 2005).
From these general considerations we can derive some simple hypotheses for each of the measures, and some observable implications of these hypotheses. These expectations relate to both the timing and the sequencing of these five policy options: 1. Politicians are more likely to draw down reserves before other options, and to delay more transparent and unpopular measures until after elections. 2. Politicians are more likely to impose capital controls before other options, and to delay more transparent and unpopular measures until after elections. 3. If reserve use and/or capital controls do not succeed, politician will raise interest rates, perhaps even before elections, since the negative economic effects are realized with a substantial lag. 4. Politicians are unlikely to impose across-the-board trade barriers before they try other policy options. But if they do resort to generalized trade protection, it will be after elections. 5. Politicians are unlikely to devalue the currency before they try other policy options.
But if they do devalue, it will be delayed until after elections.
On the basis of these relatively simpleperhaps over-simplifiedexpectations, we move to a rudimentary and suggestive evaluation of the data.

Policy Response Data
This paper examines responses to balance-of-payments crises. We determine that a balance-of-payments crisis occurred if policymakers used at least one policy response to ease payments pressure, as in Forbes and Klein (2015). Our universe of observations is thus all instances (Bcases^) in which policymakers mounted a balance-of-payments defense, measured at the country-quarter unit of analysis.
We adopt Forbes and Klein's (2015) definitions of four Bpolicy responses^and add a fifth response to the menu. The four policy responses capture large changes involving major sales in foreign exchange reserves, sharp depreciations of the nominal exchange rate, substantial increases in interest rates, and adjustments in capital controls. We follow the definitions Forbes and Klein (2015) use because it eliminates the need for making potentially arbitrary, but influential, decisions about when policies are enacted. Many of our variables are continuous and would require discretion to determine when a Blarge^change occurs. We only depart from Forbes and Klein in our measure of capital controls because their measure is based on data that are only available between 1995 and 2011. As in Forbes and Klein (2015) we aim to restrict the cases to only Beconomically meaningful^cases (e.g., cases that are not caused by other unobserved factors, such as rising inflation) by imposing a number of conditions on the terms of the policy use. Table 3 in the Appendix provides detailed information on variable definitions, limitations, and sources.
Our additional and final measure of policy response is the use of trade restrictions. Trade restrictions include import surcharges, restrictions, and quotas as well as licensing, deposits, and other requirements. We identified these cases by reviewing all GATT and WTO Balance-of-Payment consultation documents between 1947 and 2011. 11 The Committee on Balance of Payments Restrictions consults with member states that maintain restrictions for balance-of-payments reasons to determine whether they conform to GATT/WTO provisions. Appendix Table 4 lists all cases of trade restrictions in our dataset from 1975 to 2010. We focus our analysis mainly on the cases of import surcharges. We define import surcharges as instances where the surcharge is an acrossthe-board duty on imports and not levies on specific products.
We focus our analysis on the years 1975 to 2010, for which we have better data coverage. For all five policies, we adopt Forbes and Klein Bexclusion window^that excludes any cases occurring in the three quarters before and after any major policy change to avoid using a country that has just made one of these policy changes.
In the 1975 to 2010 period, few quarters qualify as being major policy changes and hence, as appearing in our cases. Instances of substantial interest rate increasescomprise approximately five percent of country-quarters between 1975 and 2010. Fewer than two percent of country-quarters meet the cutoff for devaluations or reserve use. And, less than one percent of country-quarters meet our criteria for being a surcharge, or any trade protection, or imposing capital controls. Table 1 shows the distribution of cases that meet the criteria for our policy use definitions. Numbers on the diagonal are instances in which a policy was used in isolation whereas numbers on the off-diagonal, upper triangular part of the table present instances in which two policies were used simultaneously. For ease of presentation, the table excludes cases in which three policies are used simultaneously, which comprise less than one percent of all the cases of policy use. Appendix Table 5 shows the full distribution of all policy use, including these triples.
Use of a single policy in isolation is the most common way that countries use the five policies to address balance of payment pressure, although use of discount rates is more than 15 times more likely than the use of trade protection. Drawing down reserves and depreciating are also common policy choices with the imposition of capital controls being less common. The simultaneous use of two policies is fairly common among three types of policies: depreciations, discount rate increases, and reserve drawdowns. The table reveals that capital controls and trade protection are not often used in combination with other policies, although these two measures are also just less common policy choices.
We analyze how electoral timing affects politicians' choice of policy from a menu of policy options. We identify cases as being driven by political motivations if a policy is significantly more or less likely to be chosen in a quarter leading up to, or after, an election. We compute the quarter in which an election took place using the National Elections Across Democracy and Autocracy (NELDA) dataset version 4.0 (Hyde and Marinov 2012). The data set includes elections for a national executive figure, such as a president, or for a national legislative body, such as a parliament, legislature, constituent assembly, or other directly elected representative bodies. Elections must meet basic requirements to be included in the data, including: that voters must directly elect the person or persons appearing on the ballot to the national post in question and that voting must also be direct, or Bby the people^in the sense that mass voting takes place.
All else equal, we expect politicians to be more likely to sell reserves, impose capital controls, and increase discount rates prior to elections and to be more likely to devalue and impose trade protection following elections. Simple graphical presentations showing the number of times politicians select, or delay, using a policy due to an impending election largely support our hypotheses. Figure 1 below shows the number of times politicians imposed a major reserve sale or a major increase in interest rates in the quarter prior, or subsequent to, a national election. Politicians are almost twice as likely to sell reserves prior to an election than immediately after one. Contrary to our expectations, politicians are slightly more likely to impose a major interest rate increase following an election than before. However, this difference is relatively small. Furthermore, if voters are myopically retrospective, as in Bartels (2010), they can be implemented in the quarter after an election without doing much electoral damage because interest rate increases reduce aggregate economic activity with a lag of up to eight quarters.  Trade Protection 53 Table shows policy use for instances in which one or two policies are used and excludes instances in which three policies are used simultaneously, which comprise less than 1 % of cases. There are 188 countries in the sample Figure 2 similarly plots the number of times devaluations were imposed either before or after a national election. It appears that politicians do tend to delay devaluations until after an election.
The very small number of cases in which trade protection and capital controls were imposed before or after elections make drawing any inferences about the politically motivated timing of these policies unwise (there are seven total cases of use of trade protection around elections and 17 cases of capital controls immediately before or after elections). Despite the expressiveness of these graphical representations, the effects of election timing on policy choice shown above could be driven by a number of other factors. We attempt to isolate the effect of election timing by controlling for salient economic variables.

Political Economy Determinants of Policy Choice
In this section, we focus on the correlation between election timing and politicians' choice and sequencing of policies from the five policy options at their disposal. We estimate the likelihood of selecting a given policy from the menu using a logit model. We proxy for policymakers facing balance-of-payments pressure by conditioning on policymakers use of at least one of the policy options available to them. The unit of observation is countryquarter and the data span the first quarter of 1975 to the last quarter of 2010.
Our choice of controls includes a measure of change in the U.S. interest rate, as a measure of the global economic environment. To control for the domestic economic environment, we include a measure of GDP growth, changes in CPI inflation, and the current account balance as a percent of GDP. 12 To capture politicians' choice over the sequencing of policies we include a final category of variables that denote any changes over the previous year in four of the five policies which we are focusing on: changes in reserves (as a share of GDP), changes in  (2015), we also estimated models that included a dummy variable for being under an International Monetary Fund (IMF) Standby Arrangement. The variable was not significant and our core explanatory variables retained their significance and substantive effects.  * is significant at the 10 % level, ** at the 5 % level and *** at the 1 % level. Errors are clustered at country level Policy Responses to Balance-of-Payments Crises: The Role the country's policy interest rate relative to the U.S. rate, the percent change in the nominal exchange rate versus the U.S. dollar, and the addition of any new restrictions on capital outflows. All these covariates enter the regression lagged by one quarter, or by one year if only annual data are available. We do not include lagged changes in use of trade protection because, by construction, we only count trade protection as occurring if it is not within the Bexclusion window.^The exclusion window excludes any cases occurring in the three quarters before and after any major policy change. Table 2 presents the results from the logit regressions showing the probability of policymakers' selecting a given policy from the policy menu, when they are facing balance of payments pressure. Standard errors are clustered by country.
The logit models in Table 2 perform relatively well in predicting policymakers' policy choices. Despite the relatively low pseudo-R 2 , all the models correctly predict over 70 % of the cases, and over 90 % of the cases for three of the five policies (trade protection, devaluations, and capital controls). The results reveal that the estimates for the election timing and sequencing variables largely support our predictions. With respect to the election timing variables, we interpret evidence as supporting our hypothesis when we observe a positive and significant coefficient on the quarter before election dummy for less politically costly policies and a positive and significant coefficient on the quarter after an election dummy for more politically costly policies. With respect to sequencing, if we find that a previous policy positively predicts an equivalent or more politically costly subsequent policy, we interpret this as evidence for our hypothesis that when policymakers face continuous pressure that is not relieved by a single policy they will tend to choose among relatively less politically costly policies first and only move on to costlier options if pressure persists. We similarly interpret evidence for our hypothesis when we observe that a more politically costly policy negatively predicts a subsequent less politically costly policy.
Turning first to Model 1 predicting the use of reserves, as expected, the results show that, when faced with balance of payments pressure, policymakers prefer to draw down reserves, rather than use other policies, in the quarter leading up to a national election. Domestic vulnerabilities also affect use of reserves, as countries experiencing slower economic growth, increases in inflation, or deterioration in their current account balance are significantly more likely to sell reserves. With respect to sequencing, a decrease in reserves over the previous year is associated with a lower likelihood of a subsequent major reserve sale. Policymakers can usually avoid a major reserve sale by imposing a smaller sale first. If pressure persists, they will be more likely to do so before an election owing to its relatively lower political cost. Figure 3 below plots the predicted probability of selling reserves in the quarter before and after an election. The predicted probability of selling reserves decreases approximately 13 % in the quarter following an election compared to the quarter before an election, holding all other variables at their means. A test of the difference in these predicted probabilities reveals that the difference in these predicted probabilities is statistically significant at the 95 % confidence level (Prob > chi2 = 0.0198).
The election timing coefficients are not statistically significant in Model 2, predicting use of capital controls, likely due to the small number of cases in which capital controls were used, and the even smaller number of cases in which they were used prior or subsequent to an election. Only two factors in our model statistically significantly predict use of capital controls: indicators for the previous use of other policies to combat balance of payments pressure. The choice to increase the discount rate negatively predicts subsequent capital control increases, whereas reserve sales positively predict later capital control imposition. Confirming our hypothesis, policymakers will first try to sell reserves and then impose capital controls, if pressure persists. However, we also find that policymakers will try to increase discount rates first, a less politically costly option, and will avoid subsequent capital control increases.
A priori, we expected policymakers to increase discount rates before elections. However, we find no effect of election timing in Model 3 predicting policymakers' choice to increase discount rates. Instead, domestic vulnerabilities predict increases in discount rates, including: rapid economic growth, decreases in inflation, increases in the change of the U.S. interest rate, and increases in the current account balance. Depreciations are associated with a lower likelihood of a subsequent major discount rate increase. We interpret this as evidence for our hypothesis, if policymakers pursue the more politically costly route of devaluation they will be less likely to use a less costly option, such as a discount rate increase, after.
Turning next to Model 4 on devaluations, we find that policymakers are significantly likely to delay devaluation until after elections. The results indicate that policymakers will try smaller devaluations and increases in discount rates first. If pressure persists, they will wait and impose a large devaluation post-election. Policymakers will also be more likely to devalue if there is a decrease in U.S. interest rates or a slowdown in domestic economic activity.
We plot the predicted probabilities of devaluing in Fig. 4 below. Holding all other values at their means, the predicted probability of devaluing increases almost 12 % in the quarter following an election compared to the quarter before the election. This difference is statistically significant at the 99 % confidence level (Prob > chi2 = 0.0014).
Finally, although we predicted that policymakers would be more likely to delay use of trade protection until after an election, Model 5 does not show that election timing plays a significant role in policymakers' decision to impose protection. However, the small number of instances in which trade protection was used, and the notably small number of cases in which protection was used around election time, once again limit us. In fact, the only covariates that predict trade protection in the model are a slowdown in the domestic economy and a prior attempt to draw down reserves, indicating that policymakers will try to sell reserves first to avoid imposing trade protections.
Overall, we take these estimates as supporting our hypotheses that when faced with balance-of-payments pressure, policymakers will prefer to use relatively less politically costly policies (reserve sales, interest rate increases, and imposition of capital controls) leading up to an election, and will delay more politically costly policies, namely devaluations and trade protection, until after elections. We believe that with more data, or potentially with data at a more disaggregated unit of time, we would be able to estimate the effect of election timing on the choice to use capital controls, discount rates, and trade protection, rather than other policy options, as well.

Conclusion
Governments have a number of tools at their disposal to address balance-of-payments difficulties. They can draw down reserves, employ capital controls, raise interest rates, impose import surcharges, devalue the currency, or sequence these policies in many possible combinations with one another. While economists and political scientists have analyzed these policies in isolation from one another, we treat them here as a menu of Bbad options^available to governments and we try to explain the choice, timing, and sequencing of these policies. We assume that governments seek to minimize the political costs of resolving a payments crisis. We argue that they will adopt lowercost policies before moving on to higher-cost options, and that they will delay highercost options until after elections. BPolitical costs,^in turn, are a function of the visibility of the policy and the speed or intensity of its effect. Our expectations are that governments will draw down reserves and employ capital controls before moving on to more transparent options with delayed impacts, like interest rate hikes, and that they will postpone transparent options with immediate impacts, such as devaluations and import surcharges, until after elections.
We evaluated these expectations on an original quarterly dataset of all instances in which one or more of these five policy options was used by a government to address severe payments problems. To avoid making arbitrary coding decisions about when each of these policies were implemented, we closely follow Forbes and Klein's (2015) definitions. We also extend Forbes and Klein (2015) to include the use of a fifth policyimport surchargesand we expanded coverage of all policies to 188 countries from 1975 to 2010.
We found support for several of our hypotheses. With respect to reserve sales, we found that use of this policy over the previous year is associated with a lower likelihood of a subsequent major reserve sale, but that if pressure persisted, policymakers are more likely to draw down reserves before an election than after one. This suggests that policymakers can usually avoid a major reserve sale by imposing a smaller sale first. But if that does not work, they will be more likely to resort to a large reserve draw-down before an election.
We also found that a decrease in reserves over the previous year is associated with a higher likelihood of a subsequent major capital control increase, and that an interest rate increase over the previous year is associated with a lower likelihood of a subsequent major capital control increase. These results suggest that policymakers will first try to sell reserves, as we predicted, and then impose capital controls if pressure persists. Policymakers will also try to increase discount rates first and avoid subsequent capital control increases, which may be due to the fact that capital controls are unpopular with investors.
With respect to interest rates and devaluations, we found that a depreciation of the currency in the previous year is associated with a lower likelihood of a subsequent major interest rate increase. Furthermore, increases in discount rates and devaluations are associated with a higher likelihood of a subsequent major devaluation. These findings suggest that policymakers will try smaller devaluations and increases in discount rates first. If exchange-market pressures persist, they will wait to impose a large devaluation until after an election.
Our findings regarding trade protection are limited due to the small number of cases in which trade protection was used around election time. However, we did find that a decrease in reserves over the previous year is associated with a lower likelihood of imposing trade protection. This result suggests that policymakers will try to sell reserves first to avoid imposing more visible and political costly import surcharges.
Overall, we find these results to be quite promising. They suggest that policymakers confront balance-of-payments problems with awareness of the political ramifications of each policy option before them. There are no good policy options during a crisis. What governments face is a set of bad options that differ in terms of their transparency to voters and the speed and intensity of their economic impact. Intuitively, governments appear to sequence policies, starting with less politically costly policies and moving to more politically costly policies as needed. Highly transparent policies with immediate and intense impacts are delayed until after elections. We only count cases where the reserve to GDP ratio is at least 5 %. We impose an exclusion window that excludes any cases occurring in the three quarters before and after any major policy change. We only count those quarters in which the 25 % depreciation was not preceded by a quarter in which annual inflation was 20 % or higher so as to avoid episodes in which this depreciation just reflects a response to inflation. We impose an exclusion window that excludes any cases occurring in the three quarters before and after any major policy change. We restrict the cases to quarters in which the annual rate of inflation is less than 20 % to avoid changes in interest rates that are due to rising inflation. We impose an exclusion window that excludes any cases occurring in the three quarters before and after any major policy change.
IMF, IFS accessed on March 17, 2012 Use 1970-2010, quarterly due to missing data in prior years Capital Controls Dummy equal to one if there was a decrease of more than 1.5 standard deviations in its degree of capital account openness.
We impose an exclusion window that excludes any cases occurring in the three quarters before and after any major policy change.
Chinn and Ito (2006) We count the policy as occurring in the first quarter in which a country imposed the restriction. We impose an exclusion window that excludes any cases occurring in the three quarters before and after any major policy change.
GATT and WTO Balance-of-Payment consultation documents 1947-2011, quarterly