RESEARCH
Vladislav Sevostianov, Harvard College '19
Effectiveness of Macroprudential Policies and Other Measures on Dedollarization in Portfolio Dollarized Economies: The Case of Armenia
THURJ Volume 10 | Issue 1
Abstract
The purpose of this study is to determine the most effective policy tool, with an emphasis on macroprudential policies, for decreasing dollarization in Armenia. Armenia exhibits portfolio dollarization, meaning inflation is not the root cause (inflation levels are very low in Armenia, and typically high inflation drives dollarization). The effects on the deposit dollarization rate by a number of policies was tested through six separate regressions using Matlab and through VAR modeling in Eviews. It was determined that to decrease the rate of dollarization, the policy of keeping dollar-denominated deposits in drams should be reversed, as this is a contributor to dollarization, and that loan loss provisions for dollar loans are particularly effective in decreasing dollarization. Making use of more loan loss provisions (and making them dynamic) for dollar loans and establishment of loan to value ratio caps for dollar loans would benefit the dedollarization process in Armenia.
Introduction
Dollarization
​
The dollarization rate of a country is calculated as being the ratio of foreign currency denominated deposits to total deposits. This is often also subdivided into a deposit dollarization rate and loan or credit dollarization rate. In this paper, dollars refer to all foreign currency (in addition to dollars, other currency like euros, yen, etc., are referred to as “dollars”). As a country becomes more dollarized, the central bank loses influence over the economy—since any actions that the central bank takes are severely diluted in impact. The greatest risk is that the central bank can no longer act as a lender of last resort—a problem that plagued Argentina and Uruguay. When a central bank no longer acts as a lender of last resort, it is impossible to guarantee bank stability, which may translate to jeopardizing an entire country’s financial stability. Dollarization also means that “residents are minimally benefited by the liquidity boost to domestic financial markets” (Naceur, et. al, 2015, and Levy-Yeyato, 2006). Central Banks find regulating the economy increasingly difficult with higher dollarization rates, and the actions the Federal Reserve or European Central Bank takes may often be counter to what is necessary in Armenia, or another heavily dollarized country.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​​
​
​​​
In calculating dollarization rates, a major criticism is that they are largely dependent on the exchange rate. For example, setting the Armenian dram 2:1 with the dollar, with 50% of the deposits in dollars and 50% in drams—gives a 50% dollarization rate (so say 50 dollars and 100 drams). Once the country suffers a strong depreciation, even though the economy is composed of 50 dollars and 100 drams, with a 4:1 exchange rate of dram to the dollar the dollarization rate increases to 67%, a dramatic increase. In reality, no change in the currency used has occurred in the country, though this exchange rate effect gives the illusion that dollarization has changed markedly. The more isolated a country is, the less such a change in dollarization will affect it, as domestic prices will not change.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​​​​
In Latin America, historically the prime culprit has been uncontrolled inflation driving dollarization, though it is not possible to say inflation itself is the root cause—rather: “The evidence is consistent with weak institutions driving inflation, which in turn leads to greater dollarization” (Rajan and Tokatlidis, 2005). Mwase and Kumah found that there are a number of determinants for dollarization:
-
Variables​
-
Returns/Price differential or relative returns/prices
-
​
-
-
Institutions
-
Bureaucracy quality (e.g., International Country Risk Guide (ICRG))
-
Entrepreneurial quality (e.g., Kauffman index of entrepreneurial activity (KIEA))
-
-
Prudential measures
-
Reserve requirements
-
Net open positions and other exchange-related measures
-
Other prudential requirements
-
-
Flight to quality:
-
Global risk aversion (VIX)
-
-
Controls:
-
Level of dollarization
-
Exchange rate regime (2015 (direct quote))
-
Neceur et. al., state that high dollarization rates in Caucus and Central Asia (CCA) countries are due to “asymmetric exchange rate policies, inadequate prudential regulation, financial stability concerns, and idiosyncratic economic factors.” Armenia follows a free-floating exchange rate policy (free market, not government determined) (CBA, 2010), and “has achieved relatively low inflation, [and] has implemented for a number of years a comprehensive strategy to reduce macroeconomic imbalances and vulnerabilities” (Rodriguez and Manookian, 2014). It would appear that Armenia should have low dollarization, though this is not the case—Armenia has one of the highest dollarization rates in the world.
​
High inflation and high dollarization typically cause dollarization of “asset substitution” form, meaning transactions occur in dollars (Levy-Yeyati, 2006). This may lead to a country officially adopting the dollar, as Ecuador and Panama have done (Alvarez-Plata and García-Herrero, 2007). In cases of low inflation, weak institutions in one form or another directly drive dollarization—people choose to hold money in another currency leading to “portfolio dollarization,” and the financial system becomes overly dollarized. Foreign asset demand also rises with exchange rate depreciations and uncertainty in general economic prospects for a country. This is in line with Armenia’s rise in dollar deposits during the financial crisis.
​
Neaceur et. al. suggested that CCA countries should deepen their financial systems to counter dollarization. Specific examples include “introduction of local-currency-denominated securities with credible indexation systems, development of markets for instruments to hedge currency risks, enhancement of non-banking institutions and capital markets, [and] improvement of credit information systems” (2015). Neaceaur et.al. also state that “a menu of policies aimed at macroeconomic stabilization, with a complement of prudential regulations is essential for the CCA countries” (2015). It is still unclear, however, which prudential regulations are most effective, and this may often be different from country to country.
​
Macroprudential Policies and Other Tools
The role of macroprudential policy is “to re-orient prudential regulation towards risk across the system as a whole—so-called systemic risk” (Bank of England, 2009). Properly implemented, prudential regulations lower risk to a financial system by spreading vulnerabilities across the entire system. Macroprudential measures include various reserve requirement manipulations, insurance premiums on dollarized deposits, limits on dollar lending, and strict risk management policies, such as caps on loan-to-value ratios and debt-to-income ratios for dollar loans, among others. By setting high reserve requirements for foreign exchange-denominated deposits and low reserve requirements for local currency, banks find it more expensive to hold foreign currency, and the cost is passed on to the depositor. This high reserve requirement can act as a sort of “tax” on dollar deposits. Throughout the present study, the “currency denomination of dollar deposits” variable refers to the decision made by the central bank in recent years to require reserves for dollar deposits to be held in varying percentages in drams, rather than in the original currency in which the deposit was made. For example, a value of 50% would mean that half the required reserve value would be held in drams and half in dollars.
​
The government may also implement liquidity ratios, such as a 4% floor for highly liquid assets to total assets and a 10% floor for highly liquid assets to demand liabilities in Armenia. However, liquidity ratios are not considered very effective in Armenia, as the ratios are not very strict and are not binding. Despite this, Armenia appears to still have high foreign exchange liquidity (Rodriguez and Manookian, 2014). The refinancing rate, the rate the central bank charges banks to borrow money when they are short on liquidity, influences the interest rates that banks charge when offering loans. Typically, the interest rate for dollar loans is lower than for dram loans, assuming that uncovered interest rate parity does not hold. Foreign-owned banks take advantage of this arbitrage opportunity, driving credit dollarization in developing economies (Basso et al., 2007). By raising the refinancing rate, interest rates for loans increase, decreasing the quantity of both dollar and dram loans. However, dollar loans are more affected due to their initially lower interest rate.
​
In Armenia, bank deposits denominated in drams are guaranteed up to 10 million drams, and those in foreign currency are guaranteed up to 5 million drams. This increases the risk of holding money in non-dram currencies, but the primary motivation is to increase financial credibility, which helps deepen the financial system.
​​
Loan-to-value caps and debt-to-income caps are both meant to manage risk. A high loan-to-value ratio is risky because there is little equity, and the losses for the creditor may be large in case of bankruptcy. This risk is magnified if the loan is given in dollars (for example, for a mortgage) and the borrowing entity’s income is in drams. By setting a cap on this ratio, risk may be decreased, helping counter dollarization. Armenia does not currently have such a ratio in place (CBA data). Having a cap lower for dollar loans than for dram loans helps discourage dollar loans and internalizes some of the risk. Likewise, debt-to-income caps help distinguish a risky loan from a safer one, and entities above the cap threshold will find it difficult or even impossible to borrow more money without first increasing revenue or decreasing outstanding debts. Since 2007, this has been 50% in Armenia (CBA, 2007). As with loan-to-value caps, setting a lower cap for dollar-denominated debts than for dram-denominated debts can help fight dollarization. Additionally, it may make sense for banks to take into account the currency denomination of the borrower’s income. Some countries sometimes implement regulations that make it difficult for non-exporters to receive foreign currency loans.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​​
​
​
​
​
​
​
​​
​
​
​
​
​
​
​
​
​​​​
​Loan loss provisioning works by creating a cushion for expected losses. During boom cycles, provisions are increased, which decreases profits due to the “expected” default of loans in the future. This provides a buffer that weakens the impact of bust cycles, during which defaults actually occur. During the bust cycles, only actual expenses remain, which helps mitigate the dwindling profits.​
​
​Provisions also apply to other sectors and work in much the same way, accounting for "expected expenses" for ordinary businesses during boom cycles to lessen the impact of bust cycles. While provisioning is in place in Armenia, its lack of movement makes it less dynamic and, therefore, less effective. Setting provision rates higher for dollar loans than for dram loans (as Armenia currently does) makes dollar loans more costly for banks, which helps discourage dollar loans.
​
Objectives
​
The primary objective of this work is to establish the effectiveness of macroprudential and other policies in combating portfolio dollarization, specifically in Armenia. This work may then be expanded as a cross-section to include other countries facing high portfolio dollarization, such as Georgia and Tajikistan.
Figure 1. Dollarization Rates in selected countries (from Naceur, et. al, 2015).
Figure 2. Rates of dollarization for loans and deposits in Armenia. During the
financial crisis of 2008 deposits dollarization significantly increased. Typically,
loan and deposit rates are well correlated (CBA data).
Figure 3. Time series data from January 2006 to May 2016 of the variables examined, 125 observations (CBA data and author’s calculations).
Methodology
Data
Data was retrieved from CBA sources and literature. A Hodrick–Prescott filter with a smoothing parameter lambda of 14,400 was applied to the time-series data, and the trend difference was taken. This step aimed to remove cyclical variation and establish long-term trends in the monthly data. Graphs showing the raw data before any detrending or smoothing can be observed below (Figure 3).
​
Regressions
Regressions using the Ordinary Least Squares (OLS) and Least Absolute Deviations (LAD) methods were performed in Matlab using code developed by James LeSage for the Matlab Econometrics Toolbox. Robust regressions using Huber’s t function, Ramsay’s E function, Andrew’s wave function, and Tukey’s biweight methods were also performed. Dollarization rates for deposits were used, as Levy-Yeyati (2003) suggests that credit dollarization changes typically follow deposit dollarization changes. For the case of Armenia, we can observe Levy-Yeyati’s claim below (Figure 4).
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​​
The effects of the refinancing rate, the foreign exchange rate, the dram-denominated deposits reserve requirement, the dollar-denominated deposits reserve requirement, the loan loss provisioning rate, and the currency denomination of dollar reserves were assessed on the deposit dollarization rate. This was analyzed by comparing the beta values as calculated in Matlab. These results are found in Tables 1 and 2. Certain macroprudential measures, such as capital requirements or debt-to-income ratios, were not included, as they have been constant since 2006, the start of this study.
Table 1. Least Absolute Deviations tests results as performed in Matlab. t-statistic in parentheses. *** Significant at 1%; ** Significant at 5%; * Significant at 10%.
Figure 4. Correlation of dollarization in deposits and loan in Armenia. (data from January 2006 to May 2016, CBA sources).
Results and Analysis
Regression
The refinancing rate and loan-loss provision rate both show negative relationships with dollarization, meaning that as either increases, dollarization decreases—indicating successful policy implementation. A higher refinancing rate results in higher interest rates on loans, which decreases the quantity of both dram and dollar loans. However, since dollar loans initially have lower interest rates, they increase more than dram loans, leading to a higher percentage of loans being denominated in drams. When the loan-loss provision rate (specifically for dollar loans) increases, banks must set aside more funds in designated provision accounts, which reduces the amount of dollars available for loans, also decreasing dollarization.
​
As the exchange rate variable has a positive relationship with dollarization—when the exchange rate increases, indicating a depreciation of the dram, consumers hold more dollars to maintain purchasing power, which drives up dollarization. The most counterintuitive result is that an increased reserve requirement for dollar-denominated deposits leads to an increase in dollarization. This seems contrary to the objective of a higher reserve requirement. This can be explained by the behavior of the “percentage of reserves denominated in the original currency for dollar-denominated deposits” variable. As dollar deposits are increasingly converted by banks into drams (as required by regulations for reserves), banks are left with an excess of dollars, which they begin to offer at more competitive rates for loans. This leads to an increase in credit dollarization, as more consumers receive dollar loans. Simultaneously, fewer loans are given in drams, as more drams are held as reserves instead of being available for loans. Rodriguez and Manookian (2014) reported a similar phenomenon. This effect is so strong that increased reserve requirements for dollar deposits lead to more dollarization, as even more dollars must be converted into drams, and fewer drams are available for loans.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
The positive correlation of increased reserve requirements for dram-denominated deposits is expected. If banks are required to keep a larger percentage of their drams as reserves, they lend out fewer drams in loans, which reduces the percentage of loans in drams, thus increasing dollarization. These results are consistent across all the regressions performed, and can be seen in Table 2.
​
​
Bayesian Vector Auto Regression
A Bayesian Vector Auto Regression (VAR) was performed to establish the shock effect of various policies on the rate of deposit dollarization. Using Eviews, the Bayesian VAR model was built with one lag and a Litterman/Minnesota prior. All data were processed with a Hodrick–Prescott filter, and the trend difference was taken before modeling to ensure stationarity. Impulse responses (Figure 5) and accumulated responses (Figure 6) were then generated from a one-unit decomposition over a period of 60 months.
Besides the exchange rate, the greatest shock on the dollarization rate comes from the currency denomination of reserve requirements. This policy increases dollarization, as do increases in dollar-denominated deposit reserve requirements and the exchange rate. All VAR results are consistent with those calculated in the earlier regressions.
Table 2. Different robust regression results as performed in Matlab. t-statistic in parentheses. *** Significant at the 1% level; ** Significant at the 5% level;
* Significant at the 10% level.
Conclusions and Policy Recommendations
This study examined the effects of macroprudential policy and other tools on dollarization in Armenia. Through 6 different Matlab regression tests and the construction of a Bayesian VAR model, the most effective policy was determined for Armenia, and it was also found that due to the currency denomination of reserve requirements, the tools did not always work as intended. Loan loss provisions were one of the most useful tools, while increasing the dollar-denominated deposit reserve requirement was counterproductive due to currency denomination requirements for foreign currency-denominated deposit reserves.
​
To best counter dollarization in Armenia, the following steps are recommended based on this study:
-
If Armenia is not facing the economic conditions that followed the 2008 financial crisis, the policy of keeping foreign currency-denominated deposits in drams should be terminated and gradually phased out. This is a significant driver of dollarization. Dollar-denominated deposit reserves should be kept in dollars. In fact, it may even make sense to keep dram-denominated deposit requirements in dollars, but further study is required.
-
Establish stricter loan-loss provisioning rates, and make them more dynamic. Specifically, it is crucial that the provisioning rate be higher for dollar-denominated loans.
​
Based on the literature, other steps could potentially include the establishment of stricter liquidity ratios, the establishment of loan-to-value ratio caps (with a lower cap for dollar loans), and perhaps forcing dollarization through the prohibition of dollar loans for major purchases such as mortgages and automobiles for borrowers whose earnings are not in dollars. While this particular study did not target these variables, international experience suggests Armenia would benefit from a wider dedollarization toolbox.
Possible Future Steps, Challenges, and Limitations
Future expansion of this study should include more testable variables, such as inflation, government debt, and measures of uncertainty and risk aversion, such as the VIX. A cross-sectional study examining these effects for many countries (especially those in a similar financial position like Georgia or Tajikistan) would be extremely useful.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
One of the greatest challenges for this study was data collection, as certain data that relates to the financial system is extremely difficult to gain access to. Additionally, as a non-Armenian speaker, much of the data was originally in Armenian, and the data search and translation process complicated the study. Since the study only concerned Armenia, the results should not be extrapolated to other countries. In addition, since there was a limited number of variables tested and dollarization depends on the interaction of all possible variables and the hard-to-quantify “animal spirits,” there may be important variables that have not been accounted for. Additionally, while different economic policy implementations have been correlated to the dollarization rate, a single variable itself changing may not be the cause of changes in dollarization but simply reflect other simultaneous economic developments. It may also be the case that while a policy is established at the government level, it is not actually implemented, or not as perfectly as we would expect from the government policy. However, this data is also classified as it would require microdata of banking entities.
Figure 5. Impulse response of the dedollarization rate.
Figure 6. Accumulated impulse response of the dedollarization rate.
References