Dollarization of the economy in the Post-Soviet union countries
Table of contents
Introduction
. Theoretical Background
. Posing research question
. Methodology
. Discussion of the resultsand limitations12
Introduction
the mid 20th century to our days the international financial
system has undergone a plenty of changes. For example, there was the demise of
Bretton-Woods system in 1973, which was used to recognize the gold standard as
a fundamental framework for the financial system during previous thirty years.
It was implemented to oblige each country to conduct monetary policy directed
on maintenance of the exchange rate by linking its currency to gold as well as
the ability of the International Monetary Fund (IMF) to fix temporary
imbalances of payments. This in turn implied the system of fixed exchange
rates. Abandonment from the gold standard made the U.S. dollar as the reserve
currency for many states. By being a reserve currency, dollar was also a kind
of an asset that manage to keep its value while many other currencies
experienced inflation and devaluation. This happened during high inflation
periods in countries other than the USA. Thus, for instance, phenomenon of
hyperinflation occurred on emerging markets of Russia (1992-1994, 1998),
Brazil, Peru etc. At that time domestic currencies depreciated several times.
Obviously, it lowered the real value and purchasing power of the currencies.
For that reason companies and households attempted to save the value of their
savings by using dollar-denominated deposits. Dollar deposits did not lose
their value and, for instance, served as a resort for savings. Thus, the
phenomenon of dollarization took place.has certain consequences onto the
country’s economy. These aftermaths are both positive and negative. Thus, on
the one hand, for example, dollar assets allow hedging against inflation. But,
on the other - reduces effectiveness of monetary mechanisms, reduces
seigniorage and so forth. These aspects will be thoroughly addressed later.,
while being affected by dollarization, governments lose the right to influence
its own monetary policy by adjusting the money supply. Loss of monetary policy
control leads to uncertainty in what is likely to happen with the country’s
monetary system.addition, highly dollarized economies suffer from high
pass-through of exchange rate changes into import prices, which leads to
increase in the general price level. What is more, pass-through literature
gives evidence that exchange rate changes are reflected in imports in a more
rapid way if market is a developing country. (Frankel, 2010) Countries which
will be considered in the research are developing economies. Hence, they are
exposed to high pass-through as well.this sense governments would benefit if
they know what the main determinants of dollarization process are, which
factors favor its decrease and which do not.we still can find some countries
and groups of countries with high dollarization degree of economy. Some of the
states have managed to overcome dollarization (Angola, Peru, Turkey etc.) to a
certain extent, while others could not (Belarus, Moldova, Serbia etc.). Each
country had specific economic conditions while passing through dollarization
period, which could foster or otherwise slower de-dollarization processes.
However, this research will be focused on a range of countries which are
geographically placed closely to each other and have been historically linked
to one another by means of economy, politics etc. - Post-Soviet union countries.before
we dive into their description, it is worth to mention, that there are some
close countries with similar dollarization trouble. For instance, Caucasus and
Central Asia countries.dollarization in these countries is rather high in
comparison with other developing and emerging economies. Despite reliable and
strong growth of economy during past two decades after the dissolution of the
Soviet Union, achievements in macroeconomic stabilization, dollarization still
remains persistent in Caucasus and Central Asia economies. Although degree of
dollarization in this economic region has dropped since 2000, the trend
reversed for a certain time after the global financial crisis. Deposit
dollarization grew sharply, but loan dollarization increased moderately.
Recently, foreign currency denominated deposits and loans have risen because of
the valuation effects caused by currency depreciation in several countries.
Deposit dollarization remained at 46 percent on average in the last quarter of
2013, whereas loan dollarization was a bit lower - 40 percent. (Naceur et al.,
2015). Some of these countries considered in the previous study are taken into
account in the present research as well.more example is Sub-Saharan Africa
(SSA). In SSA countries dollarization is present, prominently, it also remains
persistent and significant at more than 30 percent for bank loans and deposits.
(Mecagni et al., 2015). The aforementioned facts confirm that problem persists
and studying of de-dollarization drivers in Post-Soviet union countries is
relevant.we will focus attention on Post-Soviet union countries. Nowadays many
Post-Soviet union countries can be characterized by rather high dollarization
level - approximately from 15% to nearly 70% (Data from Central Bank of each
country).example, financial dollarization in Armenian economy has been high
throughout a long period of time, but it has experienced high fluctuations.
Thus, in the third quarter of 2014, nearly 60 percent of all deposits and loans
were denominated in dollars (Picture 1), which was slightly below the average
of 65 percent. Fluctuations in dollarization rates were rather wide: it peaked
at more than 80 percent in the 2000s, before falling below 40 percent in 2007
and 2008, until the global financial crisis stimulated dollarization increase
again. (Rodriguez et al., 2014).all the aforementioned facts we can conclude
that foreign currency was and still remains popular in transaction economies of
many countries. From the historical perspective it can be fairly noticed that
current international economic conditions can be characterized by high degree
of co-integration because of globalization processes, which means that economic
cooperation within countries under consideration has expanded as well: volumes
of energy transit have increased, countries are collaborating in the sphere of
agriculture, railroad transit and so forth.
Picture 1 Dollarization in Armenia, 1995 - 2014
dollarization
economy deposit
By analyzing historical data on dollarization degree of the
aforementioned economies, present project is aimed on finding the answer to the
following research question: “What are the key de-dollarization factors in
Post-Soviet union countries?”order to achieve the aforementioned aim we had a
plenty of problems to solve. Thus, in chapter “Introduction” we discussed the
roots of dollarization phenomenon and pointed out the importance of the topic
under consideration. Next in “Theoretical Background” chapter we described
theoretical aspects of dollarization, its causes and mention positive and
negative consequences. In the same part we made a description of the most
relevant studies in this area and depicted the main results. Chapter “Posing
research question” includes explanation of methods and factors used in the
research, main data sources and stated hypotheses. In Chapter “Methodology” we
explain in details the principles of research methods and data collection
process. Chapter “Discussion of the results” encompasses data description, building
and running of regressions with latter interpretation of the effects. In the
final chapter “Conclusion and limitations” we cover the main findings of the
whole research, limitations and give recommendations for future research.of the
research can provide information concerning the crucial determinants of
dollarization in Post-Soviet union countries, which could be useful for
policymakers as they would better understand main factors of this phenomenon,
plan monetary adjustments in a more correct way, avoid policy shortcomings of
the past as well as for the future researchers of the adjacent area.authors
have already analyzed factors and measures that promoted lowering of
dollarization level in particular countries. Thus, previous studies by Galindo
and Leiderman (2005), Herrera and Valdes (2004), Goujon (2006), Leiderman et
al. (2006) pointed out several measures based on de-dollarization factors for
solving the problem used by governments in Israel, Chile, Vietnam and Peru.
Among them: deepening the local market for government bonds denominated in
local currency, introduction of indexed instruments, floating exchange regime
and restrictive monetary policy. In addition, research by Ponomarenko and
Solovyeva (2011) found out that weak home currency exchange rate was the most
important factor that promoted dollarization in Russia before 2011., for the
purpose of the research it is also necessary to pay attention at papers which
were focused on groups of countries. Thus, Rennhack and Nozaki (2006) studied
deposit dollarization factors in Latin American countries and determined that
higher flexibility of exchange rate can foster de-dollarization.
García-Escribano (2010) presents evidence that number of dollarized deposits and
credits in Latin America countries have been reduced by means of macroeconomic
stability, flexibility of exchange rate, the implementation of prudential
regulations that can better represent currency risks.implemented in the paper
include Ordinary Least Squares, normalization. Special attention is given to
fixed and random effect models.
The paper is organised as follows: the introduction provides
the importance
of the problem. Section 2 gives theoretical background about the phenomenon of dollarization,
its pros and cons as well as literature review. Then, Section 3 introduces the
research design which describes the framework of the study. Section 4 describes the
methodology of the research. In Section 5 I present the discussion of results gained after
the investigation. Section 6 encompasses comments
on the research results and limitations. Finally the volume of the research
makes up thirty six pages without appendix and literature sources.
1. Theoretical Background
, the crucial concept in the research is dollarization.
Accordingly, it is worth giving a notion. Authors have similar approaches to
dollarization definition. Thus, they define dollarization as a situation when
residents hold a significant portion of their assets in the form of foreign
currency-denominated assets (Balino, 1999). Or, more generally - as a situation
when residents officially or unofficially prefer to use foreign currency as a
form of legal tender for carrying out transactions. In case of unofficial
dollarization agents tend to use foreign currency for transactions. However, it
might not be the legal tender. Official dollarization in turn considers foreign
currency to become the tender within the country, but the national currency is
also accepted.primary reason for dollarization is to replace a less stable
currency by a more stable one. Dollarization mainly involves the US dollar
(however, other currencies can also be taken into account, e.g. euro). This
phenomenon is peculiar to developing countries which represent high inflation
levels retrospectively - Bolivia, Bulgaria, Cambodia, Israel, Peru, Poland,
Russia etc. (Alvarez-Plata and García-Herrero, 2008).
Generally, process
of dollarization can be characterized from several dimensions. One of them -
currency substitution - has been mentioned earlier. In this case foreign
currency serves as a medium of exchange instead of domestic currency. Next
aspect refers to such notion as unit of account. By being a unit of account
foreign currency is used in the pricing and accounting processes. Another
feature is asset substitution, which implies that foreign currency is utilized
as a store of value. This in turn means deposit and loan dollarization (also
called capital flight). It is common to single out several factors which favor
capital flight:) Hedging against volatility driven by risk of return) Hedging
against inflation and national currency-denominated assets depreciation) Market
imperfections and poor financial intermediation (e.g. underdeveloped debt
markets)) Institutional aspects: lack of credibility to foreign exchange
rate peg, limited foreign exchange availability, de facto dollarization has
certain advantages and disadvantages. Thus, theory points out three aspects of
positive effects: hedging, policy anchor and financial deepening. The former
allows hedging against inflation and supports portfolio diversification. The
second one promotes macro discipline by using foreign exchange rate as an
anchor for monetary policy. The later implies using instrument for domestic
investment to form an alternative to capital flight, fostering financial
deepening.addition, Lin and Ye (2013) point out another feature of dollarized
economies which can be referred to positive ones. In the study they evaluate
the average treatment effect of dollarization on bilateral US trade with six dollarized
countries and on bilateral trade of the dollarized countries while carefully
controlling for on-random selection of policy adoption. They found strong and
robust evidence that dollarization not only significantly increases bilateral
US trade with dollarized countries, but promotes trade among dollar-zone
countries as well. Their results also suggest that the trade-enhancing effects
of dollarization are substantial.sides are represented by next aspects:
monetary policy, fiscal, balance sheet risks and lender-of-last-resort
limitations. The first one causes reduction in effectiveness of monetary
transmission mechanism. Fiscal aspect means seigniorage decrease. Balance sheet
risks represent liquidity and solvency risks caused by exposure of public and private
sectors to foreign exchange rate volatility when assets and liabilities are
mismatched. The last aspect implies the reduction of lender-of-last-resort
ability to stabilize bank system.is also worth to mention that all considered
economies are transition. They differ from developed countries in a range of
ways. Thus, they have high trade volatility, low credibility in terms of risk
default and price stability and other imperfections. In additions, they are
considered as price-takers of the world prices, which mean they perceive price
as given with little power to influence it. Many of the consumed goods are
imported from outside. That means goods are nominated in foreign currency. But,
since residents pay in local currency exchange rate can significantly change
price in national currency. This phenomenon is called pass-through effect.
Highly dollarized economies suffer from rapid and high pass-through of exchange
rate changes into import prices, which leads to increase in the general price
level and can cause rise in inflation level. Pass-through literature gives
evidence that exchange rate changes are reflected in imports in a more rapid
way if market is a developing country. (Frankel, 2010)more point peculiar to
transition economy is balance sheet effect. Balance sheet effect has become the
most important effect among the various contractionary effects of devaluation
process. Banks and firms in emerging markets often borrow funds denominated in
foreign currency, even in spite of the fact that primary part of their revenues
is in local currency. The situation is named currency mismatch. In case when
currency mismatch happens accompanied a major devaluation, solvent firms meet
trouble while servicing their debts. Sometimes they may have to close plants and
lay off workers, or even go bankrupt.dollarization consequences overweight
positive ones as high degree of dollarization is a signal of a weak economy
(Alvarez-Plata and García-Herrero, 2008). That means finding a
way to lower the dollarization level is an important problem. However, in order
to find the solution for successful de-dollarization it is necessary to
understand what determines its decrease.point out the fact that though topic of
dollarization consequences has been given a plenty of attention recently, there
is yet a lack of attention towards the empirical aspect of the de-dollarization
process (García-Escribano, 2010). Actually this is a part of the void the project is aimed to fill.we need
to review literature devoted to the topic under consideration. In accordance
with the stated aim it is necessary to consider papers which have already
figured out key determinants of dollarization for different countries or groups
of states., Neanidis and Savva (2009) in their paper study the factors of financial
dollarization in economies in transition from a short-term perspective. With
the use of aggregate data of both deposit and loan dollarization with
periodicity of one month they study the determinants of short-run fluctuations
in dollarization degree. The results give evidence that:
a) Positive (negative) short-term depreciation
(monetary expansion) effects on dollarization of deposit are exacerbated in
countries with high dollarization;
b) Short-term loan dollarization is basically
driven by banks which are trying to match domestic loans and deposits, match
currency of liabilities and assets, international financial integration and
quality of institutions
c) Both types of short-term dollarization are
influenced by differentials of interest rate as well as deviations from the
desired dollarization level.researchers have addressed the problem of
dollarization in particular countries in transition. Thus, for example,
Ponomarenko and Solovyeva (2011) measured the effects of various factors and
provide the analysis of the short-term dollarization dynamics. They pointed out
next key determinants: exchange rate factor, foreign liabilities to total
liabilities ratio, net of deposits, changes in loan and deposit dollarization
level, the differential between interest rate in rubles and the weighted
average interest rates in euro and USD on loans and deposits. They conclude
that exchange rate factor has the largest effect on de-dollarization process.
In his paper Honig (2009) analyses the impact of exchange
rate regime on dollarization. He supposes that the dollarization of the
domestic banking system represents a source of vulnerability for emerging
market countries. He argues that the regime is far less important than the
literature has previously claimed. Using annual data on deposit and credit
dollarization for 1988-2000, author estimates the regression which includes
exchange rate regime, government quality, macro and regulatory controls.
Unofficial dollarization stems from a lack of belief in the national currency,
which finally results from the faith that the local government will not conduct
economic policy that would foster long-term stability of currency. Empirical
findings indicate that improved quality of government decreases unofficial
dollarization, whereas the exchange rate regime has no significant effect on
promoting dollarization.and Panizza (2003) found out that the dollarization of
foreign debt is another important aspect of financial dollarization. The
reasons for the country being unable to borrow abroad in its own currency
include the low level of institutional development, low credibility of monetary
policy and questionable fiscal solvency.(2002) and Barajas and Morales (2003)
evaluate the effects of exchange rate policy on financial dollarization. Using
a large sample of emerging market and transition economies, Arteta (2002)
provides evidence that a flexible exchange rate regime may amplify bank
currency mismatches by decreasing credit dollarization and increasing deposit
dollarization. Author analyzes the practice of dollar lending by constructing
an optimal portfolio allocation model and testing it using aggregate data for
transition economies. Arteta assumes that banks’ credit supply determines
credit dollarization; thus, he does not include firms’ hedging incentives as a
factor in his model. Barajas and Morales (2003) provide evidence that, at least
in the short run, greater exchange rate volatility reduces credit dollarization
in a sample of Latin American economies. Their results also indicate that both
bank asset and firm liability allocation decisions are important determinants
of dollarization.Nicolo and Honohan (2005) analyzed the dollarization of bank
deposits. Authors provide empirical evidence on the factors of deposit
dollarization, the role they play in fostering financial development, and on if
dollarization is related to financial instability. They found that:) Macroeconomic
policy credibility and the institutional quality are both key drivers of
cross-country variations in dollarization level;) Dollarization can possibly
promote financial deepening only in countries with high inflation levels;) Financial
instability is likely to be higher in economies with high dollarization
degree.Latin America countries Vetlov (2001) found that specific factors may
lead to dollarization-high devaluation expectations, inflation rate,
significant interest spread between domestic and foreign currency deposits,
current account deficits, and inadequate levels of international reserves.and
Levy Yeyati (1998) provided analysis of exchange rate which showed that the
share of dollars in the variance-minimizing portfolio depends on the stability
of the real exchange rate and of the domestic price level and their
correlation.authors have the same conclusions as Barajas and Morales (2003) on
exchange rate volatility. Thus, next several studies argued that higher
exchange rate volatility, by itself, encourages de-dollarization.and others
(2010) and Garcia-Escribano (2010) show that this happens if two-way movements
in the exchange rate are allowed. For example, after observing events in the
real world we can conclude that a move toward higher exchange rate flexibility
has further contributed to de-dollarization efforts. Such phenomenon happened
in Lao P.D.R. (1995), Poland (1995-2000), and Turkey (2001). The evidence also
shows that a trend toward local currency appreciation has significantly
contributed to deposit de-dollarization in Bolivia, Peru, Paraguay, and Uruguay
(2001-10), and that an increase in exchange rate volatility also encourages
de-dollarization. The rationale is that the possibility that the local currency
may appreciate increases the risk of holding balances in foreign currencies,
which may lose value in local currency terms. Other studies, however, purport
that the causal relationship between exchange rate volatility and
de-dollarization is generally not strong (Berkmen and Cavallo 2010).general,
most the studies made in the literature have had a primary goal to examine the
determinants of long-term dollarization by basically focusing on deposit
denominated in foreign currency. In a late survey of the literature, De Nicoló
et al. (2005) and Levy-Yeyati
(2006) sum up the main determinants of deposit dollarization. They included the
past rate of inflation in accordance with the currency substitution view
(Savastano, 1996 and Sahay and Vegh, 1996), the minimum variance portfolio
(mvp) of dollarization share in accordance with the portfolio view (Ize and
Levy-Yeyati, 1998), the institutional quality and the exchange rate pegs in
accordance with the institutional view (De Nicoló
et al., 2005 and Rennhack and
Nozaki, 2006).we will consider the factors of deposit and loan dollarization
discovered earlier for economic regions. Prior analysis by Kokenyne et al.
(2010) point out the significant role of macroeconomic stabilization and
exchange rate volatility in explaining foreign currency loans for twenty one
countries and foreign currency deposits for thirty two countries from Emerging
Europe, Latin America and Africa., paper by Naceur et al. (2015) represents the
first comprehensive paper which managed to explain the determinants of both
foreign currency loans and deposits with a focus on economies of Caucasus and
Central Asia. Addressing to this article is relevant since it includes some
countries which will be considered in the present research.authors show problem
of the dollarization in Latin American economies and Emerging European markets
is much more thoroughly studied than that in the Caucasus and Central Asia
(CCA) countries. In the literature among some relevant studies devoted to the
CCA economies, De Nicolo et. al (2005) cover a plenty of countries in addition
to the CCA economies and pay attention only at the consequences and causes of
dollarized deposits. Honohan (2007) studies short-run deposit dollarization
variations as well as the effects of changes in exchange rate using a sample
that covers several CCA economies.and Petrova (2008) concentrate on the factors
of loan dollarization only in the sample of twenty one transition economy,
which covers five CCA countries. Neanidis and Savva (2009) examined short-run
variations in both deposit and credit dollarization taking into account a
number of countries with transition economy: Georgia, Armenia, Kyrgyz Republic.
García-Escriban and Sosa (2011) studied the loan
and deposit de-dollarization experience. They focused on a group of Latin
American economies: Bolivia, Peru, Paraguay, Uruguay. Authors found that
appreciation of exchange rate was a key driver explaining de-dollarization of
deposits, whilst the prudential measures implementation which create incentives
for internalization of dollarization risks, the growth of a local currency
capital market as well as successful deposit de-dollarization have all made
contribution to a decrease in loan dollarization in these economies.economic
region that has been analyzed earlier is Sub-Saharan Africa. Mecagni et al
(2015) show that efforts devoted to dollarization reduction in SSA economies
during the previous ten years had mixed results., dollarization degree has been
much higher and more persistent in SSA countries than in the other countries.
Secondly, there have been not so many successful episodes of dollarization
reduction. On the one hand, a downward trend could be seen in Angola. However,
on the other hand, countries like the Democratic Republic of the Congo, São
Tomé, Liberia and Príncipe still have the same dollarization levels as in 2000s. In
addition, the study supports the point of view that depreciation of nominal exchange
rate and inflation are main factors of dollarization, favoring the currency
substitution aspect that foreign currency is used for hedging against risk of
inflation. Instability in politics, dependence on export of primary
commodities, limited development of financial market also play a significant
role in explanation the SSA dollarization levels.also give a kind of summary
for dollarization problem and ways to solve it. Thus, they conclude that the
SSA experience proves the point that successful de-dollarization needs time,
coordinated and persistent efforts to introduce an appropriate range of sound
macroeconomic policies, microprudential measures, market-based incentives.
Direct control and mandatory measures seem to be effective in case when they are
used as a supplement to a market-based strategy. Countries in SSA and all
around the world that finally managed to cause a significant decrease in the
use of foreign currency were successful in introducing sustained processes of
stabilization and disinflation, which, in the end, helped to increase the
attractiveness of local currency usage.
2.
Posing research question
main goal of the present research is to find out what are the
key determinants of dollarization in Post-Soviet union countries. Achieving of
the aim requires several measures to be made.of all, it is necessary to decide
what factors and variables to chose. In order to answer this question I
searched for indicators which have already been taken in previous researches.
Thus, if to sum up the factors different authors used in their papers, we will
find out most commonly used variables which would reflect dollarization
process., we need to mention that dollarization is a phenomenon. We are to find
a way to measure it. Several approaches to measuring dollarization level exist.
As it has already been said, some researchers point out two types of
dollarization: deposit and loan dollarization. In this case scholars focus on
analysis of financial dollarization. However, it considers only deposits and loans
as the approach shows. It does not take into account cash held by residents. It
can make up a significant value as it is presented on example of Russia by
Timofeev (2015). Unfortunately, inclusion of cash may distort the results
because this value is very hard to measure accurately and existent measurements
are not solid and quite imprecise.will model dollarization degree as difference
between M2X to GDP ratio and M2 to GDP ratio in national definition. It is
obvious that by subtracting the latter figure from the former we will get the
figure of dollarization level in percentage value. Indicators which are
considered to explain dollarization degree are next: past inflation rate,
exchange rate, differential between deposit interest rate in national currency
and deposit interest rate in foreign currency, corruption and banks’ net
foreign asset.it is worth to explain in brief why these factors may have impact
on dollarization degree. Firstly, we will address to inflation indicator.
Inflation represents increase or decrease in the country price level. If it
goes up the whole price level rises as well. In this case, having the same
amount of money as before increase, residents are able to buy less economic
goods. Inflation fosters depreciation of savings and cash in national currency.
That means if inflation rises economic agents may tend to switch from keeping
their cash and deposits in national currency to holding them in foreign
currency. Thus, they can make dollarization level grow.it is necessary to focus
attention on exchange rate. In the present research this figure depicts how
much foreign currency costs in terms of national currency. Thus, economic agent
may possess certain assets in foreign currency (for example, dollar). If dollar
exchange rate increases then value of the assets will increase in terms of
national currency and vice versa. Due to temporary fluctuations of exchange
rate some economic agents may desire to increase their wealth by speculation,
which can also cause the increase of dollarization degree.we will pay attention
to deposit interest rates. Economic agents who wish to place their savings in
banks or elsewhere may do it in national or foreign currency. Primary reason of
placing money in bank is to get premium in form of deposit percent for allowing
bank to use the funds. However, it also serves as a way of saving money
purchasing power by preventing its depreciation due to inflation increase.
Thus, economic agent compares deposit interest rates in national and foreign
currency and chooses the one with the best outcome. By observing difference
between deposit interests rate in local and foreign currency we assume that
greater difference causes decrease of dollarization, since local currency
deposits become more attractive. In addition, if agent supposes exchange rate
(for example, dollar) to rise in future (that is national currency will
depreciate to dollar) he may prefer to make deposit in foreign currency.factor
is represented by corruption index. The main assumption here is that a higher
figure of the corruption index represents a lower corruption degree. Thus, as
the index rises, the institutional dollarization view should prescribe a lower
dollarization degree. However, as Neandis and Savva (2009) argue the
statistically corruption effect may be diverse and this can happen due to the
restricted dataset.variable under consideration is banks’ net foreign assets.
It is referred to the currency mismatch situation. We include the banks’ net
foreign assets because banks can match the level of overall liabilities and
assets by currency. This fact implies that banks can substitute loans in
foreign currency with foreign assets to borrowers in domestic country. That
means for a certain level of deposits in foreign currency a short term increase
of net foreign assets is likely to decline dollarization degree.the task is to
make up database for quantitative analysis. Data will be taken from the
official web sites of central banks in each country in “Statistics” section,
World Bank Report «World Development Index» (WDI), International Monetary Fund
statistics and Corruption Perception Index. Thus, we will gain a dataset of 13
countries: Azerbaijan, Armenia, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyz
Republic, Latvia, Lithuania, Moldova, Russia, Tajikistan and Ukraine.
Unfortunately Central Banks of Uzbekistan and Turkmenistan didn’t provide free
data for deposit and loan structure by national and foreign currency as well as
for the aforementioned ratios of M2. Time interval was defined by the availability
of data and covers period from 1995 to 2014.we should also state the hypotheses
of the research. Thus, in accordance with the theoretical aspect of each factor
effect and with what has been summarized by De
Nicoló et al. (2005), Levy-Yeyati (2006) and Neanidis and Savva (2009) we can point
out several hypotheses related to the impact of each factor. Briefly we can
describe them as follows:) The first hypothesis is H0: increase in deposit
interest rate differential causes decrease of dollarization degree. Against H1:
increase in deposit interest rate causes increase of dollarization degree) The
second hypothesis is H0: exchange rate decrease (appreciation of national
currency) fosters decline in dollarization level. Against H1: exchange rate
decrease (appreciation of national currency) causes increase in dollarization
level.) And finally the third one is H0: increase of the net foreign
assets decreases dollarization level. Against H1: increase of the net foreign
assets causes increase in dollarization level.that I plan to use encompass
Ordinary Least Squares, as proposed by De
Nicolò et
al. (2005) and Levy-Yeyati (2006). In addition, these authors utilized average
values over determined period of sample or a concrete year as well as lagged
figures as explanatory variables. However, I will use the data on the end of
the period chosen to take into account the value of flow, not stock value,
since economic agents behavior is determined mainly by end-year figures in some
economic variables, not by average values. But I will also use method of lagged
variables as mentioned not only by these authors but also by Ponomarenko and
Solovyeva (2011) and Neanidis and Savva (2009).also recommend running pooled
regression as well as fixed and random effect models (Neanidis and Savva,
2009). Estimation by pooling all observations together and running the
regression model neglects the cross section and time series nature of data. The
major trouble with such model is that is does not distinguish between the
various countries in the sample. In other words, combining countries by pooling
the heterogeneity or individuality that may exist among companies is
denied.fixed effect model allows for heterogeneity or individuality among
countries by allowing having its own intercept value. The term fixed effect is
used due to the fact that although the intercept may differ across countries,
intercept does not vary across time - it is time invariant.case of random
effect model countries have a common mean value for the intercept. The model
presumes that individual effects have occasional nature. (Wooldridge, 2013),
fixed and random effect models are suitable for usage in certain circumstances
due to difference in their nature. Thus, before running fixed or random
regression we need to choose an appropriate one. Hausman test will be used in
order to decide on which one to choose.running regressions we will get
estimated influence of each chosen factor. Once we get it we will be able to
make a conclusion about positive or negative influence and compare the findings
of this paper with previous ones., in accordance with the theory and previous
studies, we assume exchange rate to have positive sign, whereas deposit
interest rate differential and net foreign assets - negative.
3.
Methodology
part of the proposal explains the methods used in carrying
out the study. Various authors addressed the estimation of dollarization in
different ways. Earlier researches by De Nicolò
et al.
(2005), Levy-Yeyati (2006) are based on cross-sectional ordinary least squares
(OLS) regression. They utilized average values over determined period of sample
or a concrete year as well as lagged figures as explanatory variables. Such
approaches describe the long-term determinants rather than reflecting
short-term variations and outwit the problem of some regressors endogeneity to
the dollarization level by using lags.recently, Neanidis and Savva (2009) have
used pooled OLS with robust standard errors adjusted for heteroskedasticity.
They next consider the unobserved country-specific effects with the use of the
fixed effects estimator on data with period of one year.estimating
dollarization in Russia Ponomarenko and Solovyeva (2011) based the estimation
approach and variables choice on the aforementioned research of Neanidis and
Savva (2009) that is considered to be a common draft of a comprehensive review
of financial dollarization modeling in emerging markets.defining econometric
approach it is worth to address the data aspect. This research utilizes yearly
data on the supposed key dollarization factors (as exchange rate, inflation
etc.) collected from 1995 to 2014 inclusively. Information will be taken from
official sources like the Central Banks, World Bank Report «World Development
Index», International Monetary Fund statistics and Transparency International:
Corruption Perception Index. Necessary data is represented by historical
observations and has panel data structure: information on several indicators
during last several years for a range of countries.the description of the
variables will be presented. As has been mentioned earlier the analysis
includes six variables: dollarization degree of deposits, inflation, corruption
index, exchange rate to dollar, deposit interest rate differential and net
foreign assets. After computing all the descriptive statistics the results were
formed in two tables and now they are placed in appendix (Table 1, Table 6).
Then it is necessary to explain the results.
1statistics
|
Dollarization
degree
|
Corruption
|
Deposit interest
rate differential
|
Exchange rate
|
Net Foreign
Assets
|
Inflation
|
Mean
|
0,098
|
3,106
|
3,174
|
335,182
|
1,14*1012
|
14,451
|
Median
|
0,063
|
2,600
|
2,225
|
11,257
|
2,60*109
|
7,610
|
Maximum
|
0,624
|
6,700
|
22,27
|
10224,10
|
3,48*1013
|
411,750
|
Minimum
|
0,001
|
1,500
|
-6,326
|
0,480
|
-3,94*1013
|
-8,525
|
Std. Dev.
|
0,115
|
1,332
|
4,692
|
1268,467
|
5,57*1012
|
38,396
|
Variance
Coefficient
|
116,62 %
|
42,89 %
|
147,83 %
|
378,44 %
|
265,69 %
|
Observations
|
193
|
148
|
93
|
193
|
203
|
193
|
First of all, after having a glance on mean and median in
case of deposit dollarization determinants we can see that only dollarization
degree variable has distribution close to normal, since these two figures are
relatively close to be equal.continuation of the table analysis it is worth to
consider skewness and kurtosis values. Skewness is a measure of asymmetry of
the series distribution around their mean. Kurtosis measures peakedness or
flatness of the distribution of the series.can observe that all variables are
positively skewed. Thus, since figure of skewness is positive for all variables
we might imply that distributions have some outliers. In addition, it means
that on average figures of most observations exceed the value of the mode and
that sample includes observations with relatively high values - countries with
relatively high dollarization, weak local currency, high inflation etc.the
considered sample kurtosis value is greater than three in all cases, which
means all variables are leptokurtic and many values are far away from the mean.
More variance is caused by the infrequent extreme deviations. We can make
similar conclusion about the variables’ values: some countries have relatively
high dollarization, weak local currency, high inflation etc.Bera criterion
represents a test statistic for testing whether the series is normally
distributed. Its p-value for the sample is 0,00000 in all cases. That is
another confirmation of no close to normal distribution of these variables. One
more argument in favor of aforementioned results is Q-Q plot (Picture 3). We
can hardly state that any of the presented quantiles are close to the line of
normal distribution. And, in conclusion, the distributions itself is situated
in appendix (Picture 4).of variation also reports that none of the samples can
be called homogeneous, since none of the values is 33% or less (Table 1). That
means data for each indicator varies significantly - from low to high values.is
also worth to include correlation matrix in order to find out what presumable
connection is there between independent and dependent variables. However, we
will take into consideration two correlation matrixes. The reason is that
deposit interest rate differential covers almost two times fewer observations,
than other variables, which can be noticed from Table 1. Thus, we will run two
separate regression models with and without this variable as well., we will
consider correlation matrix without deposit interest rate differential (Table
2). Matrix shows that dollarization degree and corruption index, exchange rate,
net foreign assets are positively and significantly correlated; whereas
connection between inflation and dollarization is positive, but not
significant.can also observe that net foreign assets have positive significant
dependence with corruption index and exchange rate. Inflation and exchange rate
are positively connected as well. Though the highest correlation value between
regressors is close to 0.5, which is moderate, it may cause multicollinearity
further. However, to say it for sure, it is necessary to run the regression and
then to arrange Variance Inflation Factor test.
2Table without Deposit Interest Rate Differentials
Observations:
148
|
Dollarization
degree
|
Corruption Index
|
Exchange rate
|
Inflation
|
Net Foreign
Assets
|
Dollarization
degree
|
1,000000 x
|
x
|
x
|
x
|
x
|
Corruption Index
|
0,502242 0,0000
|
1,000000 x
|
x
|
x
|
x
|
Exchange rate
|
0,351323 0,0000
|
-0,078473 0,3431
|
1,000000 x
|
x
|
x
|
Inflation
|
0,062620 0,4496
|
-0,010210 0,9020
|
0,142266 0,0846
|
1,000000 x
|
x
|
Net Foreign
Assets
|
0,157003 0,0567
|
-0,157939 0,0552
|
0,493604 0,0000
|
0,067187 0,4172
|
1,000000 x
|
correlation table (Table 3) includes deposit interest rate
differential. Here we can notice that deposit interest rate differential and
dollarization degree are negatively correlated, however, the connection is not
significant. In addition, all other indicators but net foreign assets now have
the same signs but connection is insignificant in each case.order to identify
outliers Box-plot graph will be used. Plotted graphs are presented in Appendix
(Picture 5). Graphs show that there are some outliers in each case but
corruption index. But taking into consideration economic sense and box plot
results we will exclude only some observations. Thus, we have 147 observations
left in case without deposit interest rate differential factor and 74 - in case
of inclusion.final point will be determination of the key drivers. After
running the regression we will find out the impact of each considered
factor.carry out the research it is necessary to define the suitable method
that can be used to run the regression model. In this paper method of Ordinary
Least Squares will be implemented. Using of OLS is adequate since it has
already been proposed by earlier researchers.
3Table with Deposit Interest Rate Differentials
Observations: 74
|
Dollarization
degree
|
Corruption Index
|
Deposit interest
rate differential
|
Exchange rate
|
Inflation
|
Net Foreign
Assets
|
Dollarization
degree
|
1,000000 x
|
x
|
x
|
x
|
x
|
x
|
Corruption Index
|
0,021621 0,8549
|
1,000000 x
|
x
|
x
|
x
|
x
|
Deposit interest
rate differential
|
0,061243 0,6042
|
1,000000 x
|
x
|
x
|
x
|
Exchange rate
|
0,051989 0,6600
|
-0,189914 0,1051
|
0,298511 0,0098
|
1,000000 x
|
x
|
x
|
Inflation
|
-0,190782 0,1035
|
-0,191870 0,1015
|
0,517061 0,0000
|
0,140409 0,2328
|
1,000000 x
|
x
|
Net Foreign
Assets
|
-0,129541 0,2713
|
-0,148120 0,2079
|
-0,007435 0,9499
|
0,495716 0,0000
|
0,004139 0,9721
|
1,000000 x
|
Also such method as normalization will be used. The reason
for its implementation is that different countries due to different size have
very distinctive economic figures, which distort the model. In order to provide
an accurate computation we need to take into consideration the size of certain
country’s economy.to the presence of multicollinearity in the model, functional
form of some variables might be changed to log form. But all models run will be
linear in both variables and parameters. Models will be run using White
standard errors in order to avoid heteroscedasticity and get robust
estimates.in order to take into consideration the heterogeneity and some
individual features of companies we will implement fixed/random effect models.
To decide on what model to choose Hausman test will be applied. All
calculations will be performed in the program “Econometric views” ver. 8.0.we
need to address the methods of computation for each variable. As it has already
been mentioned in previous section we use variables that reflect aspects of
dollarization process such as currency substitution, portfolio modeling and
institutional. We should also point out one more time the variables which are
necessary for the research. Thus, in theory maximum figure of variables that we
can consider is constrained by the square root of observation number: = 16,12 (13 countries and 20 years)
which does not exceed the number we have chosen.into account the limitations of
data availability, we will use next variables that have been proposed earlier
in two basic articles (Neanidis and Savva(2009), Mecagni et. al (2015)):
exchange rate to dollar, deposit interest rate differential, net foreign
assets, inflation rate and corruption index. As authors have proposed earlier,
we will use inflation rate and corruption index as control variables as well.,
it is also obligatory to give information about the dependent variable. Authors
propose different approaches to measuring the phenomenon of dollarization.
Thus, some (Neanidis and Savva (2009) estimate separately dollarization degree
of deposits and loans on the end of considered period. It is computed and
determined as ratio of deposits (loans) denominated in foreign currency to the
total value of deposits (loans) of residents (individuals and legal entities)
of the country under consideration. Authors implemented first differences of
dependent variable and take into consideration its variation across time. This
method accurately considers both sides of financial dollarization. However, it
has certain flaws. Thus, it takes into account only deposits and loans and
neglects cash held by residence which can make up a significant value and may
influence estimation. However, inclusion of cash may distort the results
because this value is very hard to measure accurately and existent measurements
are not solid and quite imprecise.will utilize other approach, which, however,
considers only deposit side of dollarization. Another reason for using this
approach is raising number of observations. It allows to expand database more
than two times and include countries that otherwise would not be considered at
all. Nature of the approach is easy to demonstrate on an example. Thus, we will
address to Russia.the 1st of march 2016 money aggregate M2 in national (e.g.
ruble) definition made up 31 trln rubbles, whereas indicator of broad money -
M2X on the same date equaled 51,4 trln rubbles. The indicator includes M2 and
value of deposits in Russian banks in foreign currency. That means a
significant portion (17 trln exactly in accordance with the exchange rate) held
by Russian individuals and entities was kept in national banks but in form of
foreign currency deposits. (Timofeev, 2015)it is also necessary to add cash in
foreign currency. Its volume reached 47,5 mlrd dollars by the end of 2015 (7,1
mlrd in banks’ cash departments and 40,3 mlrd dollars on hands in private
sector) that gives another 3,1 trln rubbles with the exchange rate correction.
Picture 2 Money aggregate M2X
conclusion, total value of money assets in Russia reaches
around 55 trln rubbles. The example clearly illustrates the estimated
components of money currently circulating in the economy and makes it simple to
understand the idea. Next steps in computation of dollarization degree are
focused on extracting foreign money portion.took data on M2 and Gross Domestic
Product (GDP) from International Monetary Fund statistics in section “Monetary
data based on standardized report forms”, “Monetary data based on
non-standardized report forms” and “National accounts”. “Standardized” and
“non-standardized” section differs in the countries included. Some report
statistics based on certain forms while others do not. Both values were
presented in millions that is why in order to get M2 to GDP ratio in national
definition we merely divide M2 by GDP.on M2X was taken from World Development
Indicators which is The World Bank’s annual report. It was presented as a ratio
and no further steps for its computation were necessary.X includes aggregate M2
in national definition (which encompasses physical cash and coin, demand
deposit and traveler’s checks in national currency) and demand deposits in
foreign currency placed in national banks. Thus, deposit dollarization degree
is computed as difference between the aforementioned ratios. M2 to GDP and M2X
to GDP ratios include GDP value which allows us to take into consideration the
size of each economy.inflation rate represents annual value of increase (or
decrease) of price level in the country. It was taken from the World
Development Index as a figure of Consumer Price index. The consumer price index
(or CPI) represents a measurement that analyzes the weighted average of prices
for consumer goods and services basket, which includes such expenses as transportation,
medical care, food and etc. The CPI is computed by considering price changes of
each item mentioned in the predetermined basket of goods and services and
taking the average value; all the goods are weighted in accordance with their
importance.between deposit interest rate in national and foreign currency was
computed by subtracting latter from the former. However, the figures have
certain methodology of computation. To begin with, both indicators were taken
from the IMF statistics in “Monetary” section. IMF “Metadata” gives explanation
of computational methods for each indicator. As methodology shows, IMF
differently approaches the consideration of deposit interest rates for
different countries. Thus, for instance, deposit interest rates in both currencies
for Azerbaijan and Georgia are computed as weighted average for one-month and
twelve-months deposits in accordingly. However, we can only take them as
given.rate represents the value of foreign currency denominated in national
currency. In this research we implement annual official dollar exchange rate
expressed in local currency units. Official exchange rate is referred to the
exchange rate which is determined by the local authorities or to the exchange
rate determined in the legally operating exchange market. Its value is computed
by taking an annual average which is based on monthly average figures (local
currency units in relation to the U.S. dollar). The data on exchange rate was
taken from IMF “International Financial Statistics”.index is defined as The
Corruption Perceptions Index (or CPI) by Transparency International. The index
was introduced in 1995 as an indicator which was used to measure corruption
perceptions in the public sector and included different states around the
world. Its methodology covers 4 basic points: data source selection, rescaling
of source data, rescaled data aggregation and reporting an uncertainty
measure.) Data source selectionCPI is based on a plenty of sources which
register corruption perceptions. Each of the sources is evaluated due to the
certain criteria. Then Transparency International contacts with each
institution which provides data to verify the methodology used to build up
scores and to get a permission to publish the newly scaled scores from each of
the sources, alongside the mixed index score.) Rescaling of source data
Sources are then standardized to be comparable with other
sources, for compilation to the CPI scale. The standardization process converts
all the sources to a scale ranging from 0 to 10 where 0 represents the highest
level of corruption, and 10 the lowest degree of perceived corruption.) Rescaled
data aggregation
CPI score for each
country is calculated as an average of all the rescaled scores for the country
(it is necessary to mention that none of the imputed values as a value of score
for the aggregated CPI is used). Any country will be given a score if and only
if there are at least three sources of available data from which to calculate
the average.) Reporting an uncertainty measureCPI scores will be reported
alongside a confidence interval and standard error which depicts the variance
of the value of the data source that contains the CPI score. Our research will
utilize only the value of the index itself’ net foreign assets represent the
total sum of all foreign assets held by monetary government and deposit money
banks, minus their foreign liabilities. This figure is adjusted for the changes
in exchange rates and valuation as well. The net foreign assets position shows
whether nation is a net creditor or a net debtor to the rest of the world.
Thus, a positive net foreign assets balance indicates that it is a net lender,
whereas a negative one shows that the country is a net borrower. Data under
consideration is in current local currency units (LCU).method that worth
considering is variance inflation factor. In a multiple regression, variance
inflation factor (VIF) is used to indicate the multicollinearity.
Computationally, it can be calculated as reciprocal of tolerance: 1 / (1 - R2).
Holding all other things equal, scientists desire lower VIF level, as higher
VIF magnitudes are considered to affect the results associated with a certain
multiple regression analysis adversely. In fact, the practical utility of VIF,
as distinct from definition of tolerance, is that VIF indicates the value of
the standard errors inflation associated with a certain beta weight that is
because of multicollinearity. (Wooldridge, 2013)
.
Discussion of the results
all the necessary explanations have been given it is worth
considering the empirical part of the paper. All auxiliary graphs and tables
will be placed in appendix. To begin with, description of empirical task-plan
is next:) Building an econometric model
) Parameterization and specification
) Conducting a set of tests) Interpretation of
the results and coefficients) Explanation of the limitationsstep of the
research includes building a regression model. Firstly, linear model will be
run, since in articles researchers implemented exactly linear model. It also
will be run in accordance with White heteroskedasticity-consistent estimates in
order to get robust results. Thus, we have the next model (1):
(1)
aforementioned equation is a model which further will be
referred as a model with standard specification. Below we present the logarithm
specification (2) :
(2)
we will describe the results of estimation. Firstly, we
estimated linear pooled regression without inclusion of deposit interest rate
differential. The results are presented below in the Table 4 (Pooled regression
(1)).
4specification
Variable
|
Pooled
regression(1)
|
Pooled
regression(2)
|
Random
Effects(3)
|
Pooled
regression(4)
|
Const
|
-0,062 **
(0,028)
|
0,085 *** (0,01)
|
0,084 ***
(0,027)
|
0.0004***
(0,008)
|
Corruption
|
0,045 *** (0,01)
|
x
|
x
|
0,0004
(0,008)
|
Inflation
|
6*10-5 (0,0001)
|
-1,58*10-5
(0,0002)
|
-9,01*10-5
(0,0003)
|
-0,0009 (0,0007)
|
Exchange rate
|
6,23*10-5 ***
(9,81*10-6)
|
5,73*10-5 ***
(6,49*10-6)
|
4,19*10-5 ***
(2,31*10-5)
|
0,0009 ***
(0,0002)
|
Net Foreign
Assets
|
2,95*10-15
(8,78*10-16)
|
1,46*10-16
(1,00*10-15)
|
-4,31*10-13 ***
(9,98*10-14)
|
Deposit interest
rate differential
|
x
|
x
|
x
|
-0,001 (0,001)
|
Adjusted
R-squared
|
36,9 %
|
5 %
|
0,79 %
|
2,42 %
|
F-statistic
|
22,369
|
4,31
|
1,5
|
1,36
|
Prob
(F-statistic)
|
0,000
|
0,006
|
0,216
|
0,249
|
Observations
|
147
|
188
|
188
|
74
|
Notes: * - 10%significance level,**-5% significance
level,***-1%significance level
we see that increase of exchange rate, net foreign assets and
corruption index are positively correlated with dollarization. However,
VIF-test shows that model is exposed to multicollinearity (Table 7). Exclusion
of corruption index helps to solve the problem, but now only exchange rate
effect remains positive and significant (Table 4, Pooled regression (2)).we
will implement fixed/random effect model. Hausman test showed that random
effect model is suitable to use (Table 8). We exclude corruption as well
because of the multicollinearity. The random model estimation results are
presented in the Table 4 (Random Effects (3)).we can see that model is not
significant: p-value >0,1. That means we can not trust the results.we will
include deposit interest rate differential variable. And here again we have
insignificant model (Table 4, Pooled regression (4)). This problem will be
solved with the use of log form of dependent variable. Thus, after estimation
of log pooled regression we placed results in Table 5 (Pooled regression
(5)).using log specification allowed us to get significant model. Here we can
mention that all variables but inflation and corruption are significant.
Increase of all significant factors, but exchange rate, have negative effect on
dollarization level. But VIF test indicates multicollinearity. Thus, again we
have to exclude corruption index variable (Table 5, Pooled regression (6)).
After exclusion of corruption index all factors still have the same sign and
significance.
Table 5specification
Variable
|
Pooled
regression(5)
|
Pooled
regression(6)
|
Fixed Effects(7)
|
Fixed Effects(8)
|
C
|
-2,224 ***
(0,612)
|
-3,098 ***
(0,264)
|
-0,449 (0,655)
|
-3,06 ***
(0,306)
|
Corruption
|
-0,377 (0,263)
|
-
|
-0,904*** (0,239)
|
-
|
Inflation
|
-0,019 (0,026)
|
-0,023 (0,024)
|
-0,066* (0,027)
|
-0,028 (0,029)
|
Exchange rate
|
0,046 ***
(0,011)
|
0,054 ***
(0,012)
|
0,018 ***
(0,015)
|
0,047 ***
(0,013)
|
Net Foreign
Assets
|
-1,95*10-11 **
(7,4*10-12)
|
-1,81*10-11 ***
(6,69*10-12)
|
-2,62*10-11***
(8,73*10-12)
|
-2,20*10-11 ***
(8,03*10-12)
|
Deposit interest
rate differential
|
-0,099 **
(0,04))
|
-0,111 ***
(0,032)
|
0,017 (0,049)
|
-0,072 * (0,037)
|
Adjusted
R-squared
|
18,2 %
|
17,6 %
|
38 %
|
16,6 %
|
F-statistic
|
5,89
|
3,48
|
1,91
|
Prob
(F-statistic)
|
0,002
|
0,0003
|
0,0001
|
0,024
|
Observations
|
74
|
93
|
74
|
93
|
Notes: * - 10%significance level,**-5% significance
level,***-1%significance level
we will estimate fixed/random effect model for log
specification. With the use of Hausman test we determine that we should use
fixed effect model (Table 9). The results of estimation are mentioned in Table
5 (Pooled regression (7)). Here as well we have to exclude corruption index so
as to get rid of multicollinearity (Table 10).the final model with log
specification all variables under consideration are significant (Table 5,
Pooled regression (8)). Exchange rate increase positively influences
dollarization, whereas rise in net foreign assets and deposit interest rate
differential have negative impact. The whole model is significant as well and
as R2-adjusted reports it explains almost 17% of the total variation.we will
give detailed interpretation of the effects.) If exchange rate rises by
one unit, then deposit dollarization rises by 4,692 percent holding all other
variables constant.) If net foreign assets rise by one unit, deposit
dollarization falls by 2,2*10-9 percent holding all other variables constant.) If
difference between deposit interest rates in national and foreign currency
increases by one unit deposit dollarization declines by 7,166 percent holding
all other variables constant.is more, though coefficient of corruption index is
insignificant in fixed effect regression it has the right side as has been
earlier found by Neanidis and Savva (2009). Also positive effect of exchange
rate is persistent in all models. Inflation coefficient shows inadequate sign
almost in all cases, but it is insignificant.accordance with the results
mentioned earlier we can conclude that all three stated hypotheses have been
confirmed. Thus, for Post-Soviet union countries we can say that rise of
exchange rate has positive effect on deposit dollarization degree, whereas increase
in both deposit interest rate differential and net foreign assets - negative.
Conclusions
and limitations
we need to sum up the conclusions and discuss the results.
Thus, research signifies that stronger currency, more attractive deposit
interest rates in national currency and rising net foreign assets are important
determinants of deposit de-dollarization for Post-Soviet Union countries. Such
conclusions fully correspond to what authors have previously gained in their
articles (see Theoretical Background). What is more, all of the hypotheses have
been accepted. However, we need to take into account that some factors were not
significant and we had to exclude several variables from regressions.present
study approaches dollarization from deposit point of view. In future research
it would be relevant to consider loan aspect of dollarization as well and use
data of a higher frequency to increase number of observations.is also worth to
expand the research by replacing some factors or including other factors which
would allow making more accurate estimation. For example, in future it would be
interesting to replace CPI-index with GDP deflator as an inflation factor and
compare the results.we need to mention that any research has certain
limitations. This one does as well. First of all, since I take data only for a
certain economic zone - Post-Soviet states - findings of the project can not be
used for other economic zones. It is obvious that each country is placed into
unique economic and political situation at any point in time and using the
results for others is unacceptable., it is also necessary to focus attention on
time constraints: time period used in the research considers 20 years. We do
not consider the earlier data, due to constraints of data availability.is more,
by the time this study is finished, new data for year 2015 (which can possibly
have influence on the results) will become available. However, it will not be
taken into account, since all the implications will be made by then.addition, not
all countries were taken into consideration. Some of them did not provide
available data of M2 and M2X aggregates, which mean that deposit dollarization
figures could not be calculated via the utilized method.limitation implies: due
to the fact that data on dollarization degree was missing for some periods for
different countries, the gained effects might be imprecise or may significantly
differ from the genuine ones.used for computation of deposit interest rates in
different countries are different as well. This fact limits their
comparability. Thus, for instance, in Russia rates were calculated as average
of deposits with period of one month, whereas in Azerbaijan International
Monetary Fund computed average of twelve months deposit interest rates (IMF:
Country Notes).more limitation concerns that exchange rate. Explanation is
straightforward: if dollar appreciates against local currency then the whole
value of dollars expressed in national currency would rise, which would result
it increase of M2X and consequently in dollarization degree. In this case,
dollarization degree changes without any economic transactions: if nobody sells
or buys dollars it still changes. However, reality differs and a large number
of economic transactions happens each day. That means the estimated exchange
rate coefficient encompasses two effects: dollar appreciation (depreciation)
and foreign exchange operations. However, it is very difficult to separate such
effects, since it raises doubt whether we can precisely observe foreign
exchange operations. One more trouble causes presence of operations with other
currency (e.g. Euro, Yen, GBP etc.). Nevertheless, we can recommend using first
differences of exchange rate in order to avoid foreign currency appreciation
(depreciation) effect., the research is exposed to the problem of endogeneity,
which means that we can not include all the factors which affect deposit
dollarization. There will always be some of them left without attention.
References
) Alvarez-Plata,
P. Garcia-Herrero, A. (2008), "To Dollarize or De-dollarize: Consequences
for Monetary Policy", Working Papers, BBVA Bank, Economic Research
Department, No. 0808
) Arteta,
C. (2002), “Exchange rate regimes and financial dollarization: does flexibility
reduce bank currency mismatches?” Board of Governors of the Federal Reserve
System, International Finance Discussion, No. 738.
) Balino,
T., Bennett, A. and Borensztein, E. (1999), "Monetary Policy in Dollarized
Economies." International Monetary Fund. pp. 1-36
) Barajas,
A., Morales, R.A., (2003), “Dollarization of liabilities: Beyond the usual
suspects” IMF Working Paper, No. 03/11.
) Berkmen,
P. and Cavallo, E. (2010), “Exchange Rate Policy and Liability Dollarization:
What does the Data Reveal about ausality?”, Review of International Economics,
No. 18(5), pp. 781 - 795.
6) De Nicolò, G., Honohan,
P. and Ize, A. (2005), “Dollarization of the Banking System: Good or Bad?” IMF Working
Paper 03/146, International Monetary Fund.
) Eichengreen,
B., Hausmann, R., and Panizza, U. (2003), "The Mystery of Original
Sin", Harvard University Press
) Frankel,
J. (2010), “Monetary Policy In emerging Markets”, No. 16125. pp. 1-88.
) Galindo,
A. and Leiderman, L. (2005), "Living with Dollarization and the Route to
Dedollarization", Inter-American Development Bank, Working Paper No. 526.
) Garcia-Escribano,
M. (2010), “Peru: Drivers of De-Dollarization”, International Monetary Fund,
No. 10/169
11) García-Escribano, M. and
Sosa, S. (2011), “What is Driving Financial De-dollarization in Latin
America?”, IMF Working Paper, No. 11/10.
12) Goujon,
M. (2006), "Fighting Inflation in a dollarized economy: The case of
Vietnam", Journal of Comparative Economics, Vol. 34, pp. 564-581.
) Herrera,
L. and Valdes, R. (2004), “De-dollarization and Nominalization: The Chilean
Experience", Inter-American Development Bank: Economic and Social Studies
Series.
) Honig,
A. “Is there a link between dollarization and banking crises?”, Journal of
International Development, Vol. 18, No. 8, pp. 1123-1135
) Honohan,
P., (2007), "Dollarization and exchange rate fluctuations", The World
Bank, Policy Research Working Paper, No. 4172.
) Ize,
A., and E. Levy Yeyati. (1998), “Dollarization of Financial Intermediation:
Causes and Policy Implications”, IMF Working Paper, No. 98/28
) Krupkina,
A. and Ponomarenko, A. (2015), “Deposit dollarization in emerging markets:
modeling the hysteresis effect”, BOFIT Discussion Papers, No. 32/2015.
) Kokenyne,
A., Ley, J. and Veyrune R.(2010), “De-dollarization”, IMF Working Paper, No
10/188.
) Leiderman,
L., Maino, R. and Parrado, E. (2006), "Inflation Targeting in Dollarized
Economies", IMF Working Papers, No. 06/157
) Levy,
E. and Yeyati, E. (2006), "Financial Dollarization", Economic Policy,
pp. 61-118.
) Lin,
S. and Ye, H. (2013), “Does Inflation Targeting Help Reduce Financial
Dollarization?”, Journal of Money, Credit and Banking, No. 45 (7), pp.
1253-1274.
) Luca,
A. and Petrova I. (2008) “What drives credit dollarization in transition
economies?” Journal of Banking & Finance, Vol. 32, pp. 858-869
) Mecagni,
M. et al. (2015), “Dollarization in Sub-Saharan Africa: Experiences and
Lessons”, IMF Working Papers.
) Naceur
et al. (2015), “How to De-Dollarize Financial Systems in the Caucasus and
Central Asia”, IMF Working Paper, pp. 3-21.
) Neanidis,
K., and Savva, C. S. (2009), “Financial Dollarization: Short-Run Determinants
in Transition Economies”, Journal of Banking and Finance, Vol. 33, pp.
1860-1873.
) Ponomarenko,
A., Solovyeva, A. and Vasilieva, E (2011), “Financial Dollarization in Russia:
Causes and Consequences”, BOFIT Discussion Paper, No. 36/2011.
) Rennhack,
R., Nozaki, M. (2006), “Financial dollarization in Latin America”, IMF Working
Paper.
) Rodriguez
et al. (2014), “Republic of Armenia: Selected Issues”, IMF Working Paper,
No.15/66. pp. 2-26.
) Staines,
N. (2014), “De-Dollarization: A Cross-Country Perspective”, IMF Working Paper,
No. 14/169.
) Timofeev,
D. (2015) “Non-Keynesian savings of Russians”, Basic research program. WP BRP.
National research university Higher School of economics, No. 20.
) Vetlov,
I. (2001), “Dollarization in Lithuania: An Econometric Approach,” Discussion
Paper, Bank of Finland
) Wooldridge
, J. (2013), Introductory Econometrics: A Modern Approach. Cengage Learning
Appendix 1
6statistics
|
Dollarization
level
|
Corruption
|
Deposit interest
rate differential
|
Exchange rate
|
Net Foreign
Assets
|
Inflation
|
Skewness
|
2.467084
|
1.290590
|
1.537720
|
5.809132
|
0.847146
|
8.151202
|
Kurtosis
|
9.922089
|
3.546368
|
6.606207
|
39.66183
|
27.98506
|
Jarque-Bera
|
581.1016***
|
42.92622***
|
87.04437***
|
11894.22***
|
5304.422***
|
45055.57***
|
Observations
|
193
|
148
|
93
|
193
|
203
|
193
|
Picture 3 Q-Q plot graphs
Picture 4 Distribution
Graphs
Picture 5 Box-Plot graphs
Appendix 2
results
7pooled regression
Variable
|
Uncentered VIF
|
Exchange rate
|
1,264
|
Inflation
|
1,537
|
Net Foreign
Assets
|
1,548
|
Corruption
|
59,723
|
C
|
58,722
|
Table 8test
Test Summary
|
Probability
|
Cross-section random
|
0,8157
|
Table 9test
Test Summary
|
Probability
|
Cross-section random
|
0,0618
|
Table 10test
Variable
|
Uncentered VIF
|
C
|
31,001
|
Inflation
|
5,541
|
Exchange rate
|
4,316
|
Net Foreign
Assets
|
2,846
|
Deposit
dollarization differential
|
3,251
|
Corruption
|
23,621
|