The determinants of reform. The case of transition countries

Table of Contents

1. Introduction

2. Literature review

3. Hypotheses

4. Data and Methodology
4.1. Data sources and description
4.2 Methodology and Model Forming

5. Model Estimation and Results
5.1 Determinants of Reform
5.2 Determinants of Separate Reform areas
5.3 Reform determinants: EU and Caucasus region sample

6. Robust and Sensitivity Checking
6.1 Reform determinants: Comparing EBRD and Campos-Horvath indexes
6.2 Sensitivity Checking: Estimation of Several Models

7. Concluding Remarks

8. Conclusion

Bibliography

Appendixes

Appendix A: Descriptive specifications

Appendix B: Regression results

Abstract

The work estimates the reform determinants for 24 transition countries using spatial econometrics by maximum likelihood estimation. In the paper is included determinants already used by other authors, as well as, two new variables – export and foreign direct investments measures. Another distinctive characteristic is inclusion of spatial endogenous and exogenous variables as explanatory variables through the use of weights matrix - W. Spatial interaction is reflected by scoping on the importance of distance and informational spillovers. Obtained spatial interaction is positive and high in value. Coefficient rho varying in the range 0.22 to 0.71 indicates that in open economy countries adopt success and failure of reforming in neighboring countries. From spatial exogenous coefficients I obtained significant democracy, inflation, export and FDI coefficients. It gives idea that processes occurring in one country: level of investments, quality of democracy and level of domestic production is affecting and stimulating reforming in countries under similar economic or geopolitical status. I also found that the initial effect of GDP growth, FDI and democracy are important determiners of reforming process.

Keywords: Reform Determinants, Spatial weights matrix, Spillover effect

Affiliation: The author grants to Charles University permission to reproduce and to dis-tribute copies of this work document in whole or in part.

Acronyms

illustration not visible in this excerpt

1. Introduction

Reforms are important indicators for progress and growth in any country, while, at the same time, progress and growth are inevitable for reforming. Different types of reforms are implemented in the economic, social or political fields considering the country’s characteristics and place in the world development path. But reforms in transition countries have vital importance. The beginning of transition process is considered as the period of fall of Berlin Wall and collapse of the Soviet Union, when the geopolitical map of the world significantly changed. Formed independent countries were faced to necessary transformations in economy. From early transitioning till nowadays the process is aimed to transform centrally planned economies into market economies. In this regard, the implementation of different reforms is addressed to reach particular point of development. How fast and effective will be transition depends on the characteristics of reforms. The empirical literature on reforms and growth in transition began from around 1991. From early transition economists started to test formal hypotheses and obtained some substantive results about the topic as they already had enough empirical data. It was immediately recognized that a range of reforms would be needed for stable and sustainable growth, from early reforms such as price liberalization and small-scale privatization, to deeper institutional reforms such as corporate restructuring, competition policy and financial sector development (Falcetti at al. 2005).

During very first years of transition the process was mostly studied in the contest of growth in transition countries – mainly the determinants of growth, and reforms were considered as one of them. Time brought the researchers to the point to study reform determinants. The fundamental problem while implementing a political or economic reform is that their outcome is substantially uncertain (Fidrmuc & Karaja, (2013)). The existence of uncertainty causes lack of support from the society towards reform and lack of belief in their success. Trust towards reforming is related to the political system in the country and among other determinants democracy is one of the most interesting to study. When the country is governed by rule of law reforms are highly expected to be successful (Kim & Pirttilä (2006)), though the growth and high unemployment rate lead also to intense reforming.

Studying reform determinants results in substantive empirical results, though authors use different methodologies to test hypotheses (Fidrmuc & Karaja(2013), Elhorst at al.(2013), Campos & Horvath(2012), Fidrmuc (2003)). The considerable amount of works is addressed to informational spillover effects that are inevitable in modern globalized era (Drukker David M., Hua Peng, Ingmar R Prucha. & Rafal Raciborski(2013), Fidrmuc Jan & Elira Karaja(2013)). The concept of adapting the successful policies and reforms by neighboring countries is tested considering different methods of measuring spillover effects. In this regard, in the literature arises spatial interaction concept among countries and use of spatial econometric in studying different economic aspects becomes more and more a reliable practice. Spatial econometric is based on the distance. In the econometric model distance measure is included and it captures spatial interaction. The work aims to show that together with economic indicators that directly affect reforming in one country, the same economic measures have impact on reforming in other countries – in neighboring countries that are located on some distance from particular country. What is more important is that in this paper addresses spatial interaction among reforms, testing the spillover of reforms in neighboring countries. It brings new dimension of reform determinants measuring.

This work is structured as follows: chapter 2 provides comprehensive analysis of the existing works in the field of reforming processes in transition countries. Chapter 3 presents the hypotheses that are tested in the paper using the data and methodology described in the chapter 4. The 4th chapter provides description of dependent and explanatory variables and the methodology that is applied during empirical analysis. Following chapter 5 contains empirical results obtained by application of methodology presented in Chapter 5, as well as discussion and interpretation of the results. Chapter 6 provides the results from model robustness and sensitivity checking. Chapter 7 presents concluding remarks and policy implications. Chapter 8 provides overall conclusion. The references and appendixes containing additional tables and graphs used for discussing specific concepts can be found at the end of the paper.

2. Literature review

The aim of the work is to study the determinants of reform in transition economies. Before focusing on the main study questions of the paper, in this part will be presented the theoretical and empirical findings on transition and reforms. Why in the world we have transition economies, why are the reforms introduced in such economies interesting and important for studying, what are the forces determining the success of reforms – these issues will be covered by reviewing the existing literature on the theme.

Paul Hare & Gerard Turley(2013) in „The Handbook of the Economics and Political Economy of Transition“ write that transition economics, as a specific part of economics appeared in the science only after the 1990. The period was full of systematic changes in European countries, mainly in Poland, Hungary and Germany. In the first two countries the free election system was implemented, and in Germany the Berlin wall was dismantled. These rapid changes happened in eastern part of Europe. The huge impact on the world’s economy had the collapse of the Soviet Union in 1991. As the result the fifteen, more or less, independent states were formed. Post communist and post Soviet countries with their changing systems from centrally planned to democratic one, leaded to term “transition economic” appear in the economic science.

Transition countries(TC) were considered mainly as economies moving from planned system to free and democratic one. In fact, it did not turn out to be true for all countries. The discussions lead to seven elements to be realized during transition: dismantle the old system; macroeconomic stabilization; domestic price liberalization; trade liberalization; privatization and restructuring of inherited state owned enterprises; social safety net; diverse institutional reforms. Our focus areas - reforms – and their importance were underestimated, mainly because the reform implementations were expected at later phases.

Svejnar (2002) in his paper evaluates progress of transition on the example of five European countries, The Czech Republic, Hungary, Poland, Slovakia and Slovenia—in comparison of Russia. The author discusses the strategies implemented by the countries, and the corresponding results in terms of growth, employment and other economic indicators. The above mentioned seven elements of transition, in Svejnar’s work are divided into reform types: 1 and 2. First type covers macro stabilization, price liberalization and dismantling the communist institutions. Second type of reforms includes the change and development of regulation system and institutions that are fundaments of forming free economy. The type 1 and 2 reforms were implemented in all transition countries with the same or different sequencing. The author presents results for the chosen countries, but as the process characteristics were same in all transition countries, we can derive the overall evaluation in terms of employment – in all transition economies, at the beginning, labor productivity declined, and after, the stop of declining in employment caused growth in productivity and started the recovering second decade of transition.

Roland (2002) also presents the assessment of the process of transition in the first decade and the overall results in the countries. He uses different approach of explaining the changes. The arguments in the paper come from political economy. Precisely the ideas are divided into arguments from normative and positive political economy. The great importance of this work is that it gives recommendations of how make the reforms to be enacted. The normative political economy focuses on the uncertainty that is in the society whether the reform is proper or will be successful. Trust and support is lacking in majority cases during transition.

Ex ante and ex post political constraints determine the reforms to be enacted, while the positive political economy arguments look at institutions and regulations in the country as exogenous variables in transition analyzes. Though, we should consider that transition process involves the changes in institutions and in governing systems of the country. Considering the fact that reforms are part of transition process, institutions and system of the country can be considered as one of the determinants of the reform. Roland(2002) underlines that democratic reforms were always preceding economic ones. That strengthens the idea, that if the system is democratic in the country it should have positive influence on the success of economic reforms. Systems, regulations and newly built free societies are the main measures of well doing transition together with the economic measures.

While discussing the elements of transition from different angles and with different ideas of different authors, considering the speed of transition would be good instrument to predict the possible end of transition. But the thing is almost unmeasured. Castanheira & Roland(2000) analyze the optimal speed of transition. Authors mainly focus on the starting point of transition- state owned economy- and the expected end of transition process- the private capital in the economy. They analyze how fast this transformation should happen to obtain the optimal speed of transition. Dismantling the centrally planned system involves privatization of already inefficient state owned enterprises. But the mistake is rapid process. It generally leads to lower transition speed. The same logic should work for other elements of transition: the fast liberalization, internal and external, can bring the economy to disorder.

The importance of sequencing and speed of reforms in transition is shown in the paper of Staehr (2005). Where the link between the two mentioned is examined by using principal component approach for forming reform clusters. Author uses 25 transition economies panel data for 1989-2001 periods, excluding some countries (Bosnia-Herzegovina, Yugoslavia, Mongolia, Vietnam, and China). The motivation for studying impact of speed and sequence of the reforms for economic growth was the debates over strategies of reforms when transition economies suffered from production fall. The results present the positive impact of overall reform indicator’s on growth. The same pattern of influence is obtained separately for small-scale privatization and for liberalization and their early implementation also is beneficial, while large-scale privatization and bank liberalization have negative impact on economic growth. The speed of reforms can cause higher growth for long future, but not for short time. Still the arguments supporting or neglecting the positive influence of rapid reforms on growth are not obtained from the study.

In „The Handbook of the Economics and Political Economy of Transition“, authors Campos & Coricelli (2013, “Economic growth in the transition from communism”) summarize the two decades of transition in terms of growth in corresponding countries and growth in connection with reforms. Authors stress out that during this period, mainly, was studied the macroeconomic influence of structural reforms on economic growth. From 1980s the implementation of reforms were ending in different results – in success or failure. This experience leaded to different studies. But the relationship examined, was, mostly, impact of reforms on growth. In major cases, the papers about transition economies focus on this type of relationship. The growth is considered as the main indicator showing the country’s progress in transition.

Continuing the former idea, Falcetti, Lysenko & Sanfey(2005) in their work examine the link between reform and growth. Authors, considering the studies of many other economists, follow the idea that this linkage is not plain one, it is complex. To measure the influence of market oriented reforms on growth is almost impossible, as in the economy we find many determinants determining the growth. Authors also emphasize the existed EBRD(European Bank for reconstruction and Development) transition scores cannot fully reflect the implemented reform results and accordingly, it is impact on growth. But the results from the analysis still are robust, showing the sedative influence of reform on growth in corresponding country and period. The main expectation from the authors was the continuing study of the topic until the time when the economists do not understand the effects of different reforms and the path of the growth of economies. Their paper was published in 2005 and the fact is that nowadays there is more information and more researches on the topic. Especially there is growing number of studies on reform determinants.

During transition the link among reform, growth and democracy was also studied by the economists. Fidrmuc(2003) concluded in his paper, that democracy firstly influences economic liberalization, and the last one has positive impact on growth in transition process. Fidrmuc studies the relation by examining the influence of standard macroeconomic variables and special transition indices on economic growth. The important issue of the paper is that democracy, without considering liberalization can have negative effect on growth at the beginning of transition. But the impact of democracy on growth, when the liberalization is taken into consideration and used in the regression, appears to be positive or negative. But the main thing to emphasize is that democracy has influence on reforms and through it on growth. This strengthens the idea of using democracy as determinants of reform in the country.

Interesting results are obtained in the paper by Lawson & Wang(2004). Authors study different policy impact on growth - the level effect and speed effect. The prediction of the study defined structural reforms positively influencing the output of the country. The concluding results are different from predictions. Inflation and the ratio investment to GDP(Gross Domestic Product) have more significant and congruous impact on growth than different reform policies.

To finish the general part of reviewing transition process and the growth-reform linkage from the perspective of reforms’ influence on growth, Babeckýa & Havránek(2013) studied the structural reforms and growth in transition by meta analysis. All the studies they used in the paper examined the reforms in 26 transition countries. The result was as expected: in the long run future period, reforms had positive impact on growth. Evolving the costs of implementing the reforms for short time period gave the small negative effects on the growth. These results were summarized by partial correlation coefficients to capture the significance of the impact. Also the differences in reforms gave different significance and speed of the effect on growth.

All the above discussed papers and a lot more existing in economic literature, prove the importance of reforms for transition economies. Though the results are different in each paper, the influence is obvious and studying of the topic is inevitable. Transition countries from 1990s are on the way of becoming market oriented developing countries. General knowledge in the society is that reforms are implemented in the world with different degrees of success. The paper focuses on reforms in transition economies. Precisely, to study how they are implemented and what determines their succes or failure. The studies of the literatures including as the main sphere of analyses reforms, will be reviewed below. The economic measures that will be used in this paper as reform determinants are chosen according to those papers and works, as well as the measures of reforms.

Kim & Pirttilä(2006) examine the reform dependence on political constraints, that include: public support for reforms – a proxy formed from the results of CEEB(Central and Eastern EuroBarometer) surveys and system in the country –democracy, measured by the degree of freedom in the country (source: freedom house). Also they used the following macroeconomic variables: growth, inflation, unemployment and income inequality, measured by Gini coefficient. Their target of analysis were 14 transition countries in years 1990-1997. The focus of the study – reform index is taken as weighted average of three indeces from World Bank data, they use EBRD data for estimation and for checking and comparing the results by using different reform measures. Though the period of study is not long, the results are important. All the determinants in regression and their influence on reform progress is in line with predictions. To summarize, the public support for reforms has positive and significant influence on reforms. Low unemployment rate gives rise to reform progress and low economic growth is also translated into more progress in reforms. Great importance has the political freedom – democracy in the country – on reforms pace. The results show positive relation between democracy and reforms. The more democratic is a country, the more rapid is progress and implementation of new reforms. In case of inflation, the negative influence on reform progress is captured, as well as inequality in the society is reflected on reforms negtively as the supoort for reforms is weakend. The fact that the results can be used for deciding the determinants of reforms is strengthened in terms of economic significance of the results.

Public support that is trust from the public and institutions for reforms is adressed in the work of Heinemann & Tanz(2008) as determinant of reform progress. Authors provide theoretical proves and the empirical testing that trust should be included and considered as determinants of reform as well as for changes in economic insitutions. The measure of trust is obtained from surveys measuring the level of trust in public. In relationship with reforms trust is related to certain or uncertain information how successful they will be. The results from testing are different according to type of policy and reform. The positive impact is obtained in case of reforms in legal system, labor market and privatization. Combining the empirical results and theoretical background, trust is essential and important for reform success but is difficult to measure.

Another important issue is that the growth or fall in output caused by different conditions of the country has great impact on reforms. The sphere of interest is crisis that is immediately expressed in decrease of growth or in negative growth of the economy. It was fact in 1990s in transition economies and Fidrmuc & Tichit(2013) in their study try to show positive impact of crises on economy, especially on reforms. The paper is based on the infrmation mainly from post-communist countries. Authors examine the long-term impact of crises on reform, that in transition economies is market oriented. The measure of crises is output, its fall at the beginning of transition process. Alongside with output fall, the following variables are used as explanatory ones in the paper: lagged reform index, democracy index, dummy for war. All the variables are used in lagged form, except dummy. The results are in-line with predictions. The past level of democracy and the main variable of interest - output fall - have positive effect on reform progress. The last coefficient of war dummy is also significant, fully proving the idea that countries respond to severe crises with higher progress in economic reform. Overall result is that crises is translated into growth in output and into advanced economic performance- more reforming.

The regime in the country, system of governance, defines all the above mentioned indicators of the economy. An interesting work of how democracy works for reforming by Intriligator (1998) presents important features of the theme. The author discusses and brings examples of political systems and implemented policies in post-communist countries. The necessity of democracy for reforming is put under doubt, as the practice shows existence of two different systems in post-communist countries and two different development patterns. In Table 2.1 is presented the four distinct examples:

Table 2.1 Expanding economy Collapsing economy

illustration not visible in this excerpt

Here china is vivid example that without democracy success is achieved and Russia, on the contrary, proves that with democratic system collapsed economy can be also obtained. Two large countries give a concept of no need of democracy. While sufficiently small countries experienced the opposite. The author introduces important concept how the results were attained. In successful cases of the mentioned countries the influence from developed countries played a great role. All transition economies are supported from more developed countries and international organizations. That may give us purpose to believe that democratic systems in transition countries will work in a right direction. Again including democracy in our model can be strengthened with the ideas from paper by Intriligator (1998).

The effect of growth and political system of the country on liberalization is addressed in the work by Pitlik(2008). The author uses sample of developed and developing countries for examining the interrelation and impact of economic performance and democracy (autocracy) on reforms. To specify, Pitlik’s(2008) purpose is to show how system works for or against the reform during growing economy or in crises period. Reforms implementation is often followed by opposing from different parties in the country. That is main driver for prediction that in autocratic country, where one’s word rules, the implementation process would be smooth. But the concluding results after testing and checking for significance show that democracy is catalyst of reforming. It is related to more liberal reforms and especially, during fall of economic growth appears to be more responsive political system, than autocracy.

Transition process involves changes in political, economic and social lives in the country. All the transformations are kind of different reforms towards free market economy. Volker(1999) discusses and examines the reasons of unemployment fall or rise in CEE(Central and Eastern Europe) countries during transition. First is discussed the possible forces that defined the evolution of the unemployment rate. The most influential is U- shaped development of GDP path – rapid decline in output. That was followed by labor productivity fall. But the unemployment changes itself could be led by three possible mechanisms. The author proposes three forces presented in Blanchard(1997). The first is reallocation of labor from state to private sector – part of privatization process. Second is disorganization: the disappearance of central planner made vague who could be the supplier and buyer, as before there was only one type of firms- state firms, and one supplier-buyer – state. The final mechanism can be restructuring. This is directly connected to the forming of private companies instead of state firms. The process involves not only the change of owner and style of governance, but the fundaments: the quality of product, the size of firms, quality of integration inside the firms and the attitude towards labor.

These factors, proposed in economic literature, really defined the unemployment rise. But the interesting fact is the impact of unemployment on transition process. It can develop in two scenarios: become an obstacle during transition, if it reaches too high level, or become a product of transformation and catalyst of reforming. The model in the paper proved the second approach of unemployment role, but the real facts from several transition countries strengthened the first one. Volker(1999) concluding his study, sees more negative impact of unemployment on the whole process of transition. These results are interesting, as unemployment will be used in my work as reform determinant. I will try to show which impact – positive or negative- is more.

The study of transition economies covers mostly 26 transition countries. The patterns of development, results of reforming and factors determining reforms are almost the same. If the difference appears, still we see groups of countries that move forward with same economic measures. The groups in the economic literature are named as CEE countries, or European Union(EU) member transition countries, countries in Caucasus region, etc. The question and interest rises how and why it happens.

Fidrmuc & Karaja (2013) present the impact of informational spillovers on reforms. They use the gravity model and data of post-communist countries during 1990-2000s. This period can be characterized as boom of reforming in these countries after collapsing the communist governance. The study begins with explanation of the impact of uncertainty on implementing reforms, not only economic, but political and structural. The paper presents comparing of two situations. One is when completely new policy is implemented in the country. The second is the situation when already successfully implemented reform in another country is taken to be introduced in the specific country. Comparing offers different scenarios. Different expectations from public are obtained and it is quite understandable.

To test the hypothesis that informational spillovers matter Fidrmuc & Karaja (2013) use democracy or economic reform index as measures of political or economic reform. They offer three types of distance measures: geographic distance, dummy for having border between countries and same history. They test the model with each measure separately, than together. Their model verifies the hypothesis that the previous outcome from reforms lowers the uncertainty for its future benefits in another country. And the probability of repeating that particular reform elsewhere increases. Different reforms happen actually in waves and the closer are the countries the more rapid is spillover. It is not surprising and moreover, it is widely used in political economy by reformers to gain the support from public for particular reform.

The concept of importance of the information from neighboring economies and economic co integration between countries is examined by Elhorst at al. (2013). They use for estimating the co integration in financial liberalization index spatial panel data model and the data from the paper from 2008 of Abiad et al. They also examine the effect of changes in a specific reform determinant in one country on the reforms in another country. It really has impact and countries co integrate, as the lack of information about important policies implemented in the country makes necessary for states to consider information and outcomes in another countries. The use of spatial econometrics in the paper of Elhorst at al. (2013) is interesting point and important one, that played a great role for choosing the model specifications for this work. Using of spatial models counts long history, but mostly it was used in science and geography.

In Miller (2004), is presented the connection between spatial analysis and Tobler’s first law. He starts with the phrase: “Everything is related to everything else, but near things are more related than distant things” (Tobler 1970). The author discusses the fact that with the passing of centuries technologies make distance smaller in terms of time. And relation of closer things, in our case closer economies, their influence on each other is of high interest to study. It is a good point to start geographic investigations states the author, but the same rule might work in economics and the use of spatial modeling explains and proves it.

In Elhorst at al. (2013) using this spatial econometric, authors take financial liberalization index and test the relation of different economic variables. They take lag of dependent variable, and also include country and time fixed effects. To measure the interaction between reforms is used weights matrix –W- describing how countries are situated in spatial dimension. The authors also use spatial first differences model and then error correction model presentation of the first model. The approaches were used to obtain stable model. The estimations show that evolution of reform in one country depends on the reforms implemented in other countries, as well as on the previous reform index in the same country. But as the authors used data from 2008 of Abiad et al, they compare past and their own results. The authors find differences in significance of main variables of our interest in the study: the financial reform influence in different countries is smaller; democracy and governance are also insignificant. To conclude, the results show less importance of presented relation and interaction of neighboring countries, but it still remains of high interest to farther researches.

Greatly considerable results are presented in the paper by Campos & Horváth (2012), where authors study reform determinants. They focus on three reform indexes: internal liberalization, external liberalization and privatization reforms. The main difference of their work from other related literature is the use of new measures of these indexes. The new measures are obtained by them after collecting the data necessary for calculating the indexes. The authors compare their measures with existing measures of World Bank and EBRD. The reason of constructing new measures was the lack of information how the previous indexes were generated, intention to reflect more the process of transition and its characteristics and the prediction that in comparing two studies on determinants of reform, the new results would be more precise.

The estimation results, we can say, repeat the development patterns presented in related literature. The difference is in better dynamics and the size of impact of a particular determinant on cumulative or separate reform indexes. To specify, the effects on growth are higher and on overall reform are smaller than before in related literature. Also democracy appears significantly affecting the all three reforms, while economic growth and political power are significant only for external and internal liberalization. The process of reforming presented by estimation using new reform measures becomes less smoothed. Campos & Horváth(2012) measures can be seen as criticism of existing measures, and an obstacle to use EBRD measures in new studies. But also can be seen as better measures, still replicating the important results obtained by using those old measures.

This part of the paper tried to discuss the part of wide literature on reforming in transition economies. Connect their results and see why different authors studied the concept, why they used particular determinants, what was their prediction and actual results. All the references affected the choice of determinants for my work, and the way hypotheses are constructed, questions are stated and possible ways of answering are generated. The process of transition started after enormous changes in political and economic life of many countries is itself one great reform. It is multidimensional as Campos & Horváth(2012) state in their paper is affected by different mechanisms and forces. This will be one new study of this complicated relation and interaction, to show development path in somehow different way.

3. Hypotheses

The theme of the work is wide and different macroeconomic measures can be considered as reform determinants. The testing and checking of their impact on reforms is of high importance and different implications can be drawn out of the results. The choice of the main hypotheses that will be addressed in this work will cover only part of possible multiple relations around the theme. Somehow our questions can have the similar form as in other related literature, and through the process large amount of questions will arise, but the main purpose is to examine the four following hypotheses:

Hypothesis #1: Neighborhood effect is positive for reform undertaking – distance matters for information spillovers Here the Tobler’s law “Everything is related to everything else, but near things are more related than distant things” (Tobler, 1970) is driving force. Transition economies move to the same direction and almost the same mechanisms are catalysts or barriers for them separately and together. But the impact of each and particular country on others is significant, and the closer are those countries the more obvious is this relation. Proving this hypothesis will prove the idea that “neighbors” are more influential for closer environment. The speed and level of impact is higher whether it is in positive or negative direction.

Hypothesis #2: Region and union membership does matter for different development pattern of reform indicators (case for Caucasus region and EU member transition countries).

This is partially continuation of the first hypothesis. The informational spillovers will be examined on smaller subsamples. One sample will be made according to geographically neighboring target countries and the second one will consist of EU member transition countries. Will be examined the idea that, for example, that EU member TCs are characterized with higher spatial dependence than other transition countries influence each other.

Hypothesis #3: The influence of determinants is different on the results of a reform – democracy should be most influential.

This hypothesis serves the idea that political system of the country can greatly define the reform path, and the progress of reforms. Despite the democracy each of other determinants are characterized with similar level and pattern of influence on the reforms tested by different authors. Democracy is considered as two-sided determinant, difficult to emphasize the one positive or negative influence on reforms. In this paper the predicted result is considered that democracy in its true form only supports development and progressive, market oriented reforms.

Hypothesis #4: The determinants have different influence on different reform indexes (large scale privatization, governance and enterprise restructuring, etc).

This hypothesis will show what determinant is of high importance for separate reform indexes. Moreover, will be checked how robust those results will be. We will took the six reform indexes and check which determinant is important for each of them separately. Then will be done the same for indexes measured by Campos and Horvath (2012) and comparing them will discuss the robustness of our results. Prediction is that the results using the main data will be significant and robust. It will define the reliability of the study contribution.

4. Data and Methodology

4.1. Data sources and description

The data used in this work comprises of reform measures and the measures of reform determinants. The data is collected for 24 transition countries from which 10 countries are European Union(EU) member countries . Majority of economists use the data of these countries for studying different concepts of transition – majority of the works discussed in the literature review part of the paper. Several authors use more or less countries, depended on their sphere of interest and availability. For this study is chosen the countries for which all the necessary data is available. The big share of not-available data is a real barrier as many countries’ statistical data on important economic forces are still unavailable. For that purpose I excluded 5 countries: Bosnia and Herzegovina, Mongolia, Montenegro, Serbia and Turkey. The time-period of the study is the interval 1995-2012 years. The desired period for the study was from the very beginning of transition process till nowadays, but here the problem of necessary data arises, and accordingly the early transition years are not addressed.

The choice of data is depended on determinants of reforms considered by the related authors (Elhorst at al. (2013), Campos & Horvath (2012), Fidrmuc (2003), Fidrmuc & Karaja (2013)). Some determinants are the specific economic measures. Some of them are overall indexes containing different measures of the concept. In this work are used four main sources of data: The EBRD, The World Bank, The Freedom House and CEPII(Research and Expertise on the World Economy) data banks. The sources are considered to be reliable by the high frequency of use of the data from those institutions. Below, in the Table 4.1 is presented the data, their sources, measurement unit and expected sign of influence.

Table 4.1 Data Description

illustration not visible in this excerpt

FDI(Foreign Direct Investment) Net inflows, % of GDP The World Bank +

It needs to be mentioned, that in the model is used the overall index of reform as well as the separate reform indexes. The overall reform index is computed as average of five EBRD indexes presented in the Table 4.1. The source is EBRD, which is the main source of data in lots of economic papers and in the papers discussed in the literature review part. EBRD by constructing transition indicators assesses the transition results for specific year and country comparing to industrialized countries. The level of development of different spheres and reforms in the country are rated between intervals from 1 to 4+, 1 is least developed sphere and 4+ the most developed one. The plus or minus signs are defined as 0.33 points more or less than rounded index. EBRD indicators are often criticized for ambiguous measurement methods, though better measures are not yet provided by other institutions or economists for the chosen period of this study. Previously was mentioned the new reform measures by Campos & Horvath (2012) and their reliability. Those measures are computed only for three similar EBRD indexes and only till 2001 year. But they are used for comparing and robust checking of obtained results in this work accordingly by using the similar EBRD indexes.

This paper is different from other related works by including spatial weights matrix in the model, as measure of distance among countries and as an instrument for capturing the spillover effects among countries. It is used to construct spatially lagged variables (endogenous and exogenous) in the model for addressing the spatial dependence. Longitude and latitude coordinates of the capital cities of each TC taken from CEPII are used for structuring inverse distance weights (W) matrix. The choice of using inverse weights is determined by the work of Elhorst at al. (2013) where authors used different distance weights in their model and compared the obtained results to find proper measure and form of weights matrix. The detailed process how W matrix is formed is explained in methodology part.

Another important issue is the democracy index. The source is Freedom House scores reflecting freedom in the world by measuring the political right and civil liberty level. In the paper is used the overall measure - average of these two sub-indexes. The measures have values from 1 to 7. Countries with the indexes close to 1 considered as free and democratic, while with the score close to 7 considered as autocratic. This measure of democracy was used by Fidrmuc & Karaja(2013) in their work studying policy reforms and informational spillovers. However I used the original measures, while Fidrmuc & Karaja(2013) used rescaled indexes. Must be admitted that many doubts exist, on the one hand, how democracy is recognized by the governors and society, and on the other hand, if the measures from the Freedom House are appropriate and reflect the reality. But again this is the source used most frequently in economic researches and no other institution provides democracy measures for the sample period used by me.

The rest of the variables are directly taken from the sources. GDP per capita growth is important determinant. It is direct measure of economic growth. For reforming growth is essential, it acts like catalyst for reform progress in the country. This is proved by lot of papers and their results, as GDP per capita is used in reform studies presented also in the literature review part. The source is World Bank and annual data is taken for 18 years.

Unemployment rate in the total labor is considered to have also positive impact on reforms, though it, the high rate of unemployment, is not a good issue itself. The data from World Bank is also annual for 18 years. The logic behind considering unemployment as reform determinant is that high unemployment requires significant changes in policy implementation that is reforming.

Inflation is another measure of the economic conditions and its impact is of high importance to examine. High inflation periods as well as high unemployment rate is considered to cause more reforming by Fidrmuc & Karaja(2013). Their estimated model coefficient reflects the positive influence of inflation on financial liberalization index. Though Campos & Horvath (2012) while sensitivity checking on three reform areas obtain negative relation with privatization and external liberalization and positive with internal liberalization. However in this paper is expected negative influence as high inflation for long period works negatively on the progressing economic conditions.

FDI is most important determinant for the transition economy. The majority of reforms are implemented with the help of FDI-s. The high rate of FDI inflows in the country is considered as directly connected to more reforms. Jan Svejnar (2002) provides graphical representation of FDI inflows to transition countries. Those TCs that from the early periods characterized with high rates of FDIs now appear as leading TCs. The former is logic for considering FDIs as determinant of reform.

Another measure of economic performance and developed market economy is level of export of the country. The relation and impact on reforms is pondered to be also positive. Process of privatization, price liberalization and trade & FOREX system can be considered as supportive for rising export rates. But the opposite relation is of high interest to examine and the expected influence is positive. The results from empirical study show which of the predictions are true by using the spatial econometric modeling.

However it should be mentioned that for testing the hypothesis different subsamples are used. In all samples Asian TCs are also excluded as in most other works. In the Table 4.2 is provided the descriptive analysis of the full data sample for all TCs and eighteen years. It provides the statistical characteristics of the observations:

Table 4.2 Descriptive Statistics

illustration not visible in this excerpt

N stands for number of observation, SD-Standard Deviation, FDI-Foreign Direct Investment, LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System.

From the Table 4.2 we see that data is strongly balanced that is first main condition to estimate the chosen model in the software Stata. Sample is not large, overall 432 observations were collected for the given period and countries. For the reform indexes we see that through the years and countries data varies from minimum score 1 to maximum 4+. Same is proved for constructed democracy index; presented sample countries are fully free as well as autocratic.

To move on, data distribution can dictate us whether the estimation results are reliable or not. Though in real world data normality is hard to find and gain. Skewness and Kurtosis statistics presented in the Table 4.2 are far from the best options for assuming the normality of the data. Most of the skewness measures are far from 0. Majority of kurtosis measures are also far from normal level of 3. To eliminate the non-normality I used data transformation in logarithms but the problem still was presented while checking normality. Accordingly, the original data set is used in the models. In the next section is described the methodology used for estimations.

In the appendix A are presented descriptive statistics for other three samples used for hypotheses testing. Table A.3, A.4, A.5 represent the statistics results. The differences in statistical characteristics of sample mainly come from the number of observations and length of the sample period.

4.2 Methodology and Model Forming

The topic of this paper gives broad possibility of using different econometric models. However, considering the specification of the work - studying the determinants of reform addressing the spatial dependence among sample countries leaded to the choice of spatial econometric models. Why should we care about spatial dependence can be explained by two main reasons. The first is the important statement already mentioned above in the literature review part. It is about the distance and the larger influence of closer entities on each other (Tobler 1970). The second reason is broadening of the first one, and giving some economic intuition to the spatial characteristics among countries. The term of open economy reflects that the country development process is not a sole one. In the traditional econometric models the influence and interconnection of different economic variables among different countries, cities or other geographical entities is not considered. Especially, when we study the reform determinants, a specific sphere, where practice shows that neighbor or countries with similar development patterns adopt successful reforms, we should include spatial correlation in the model. Spatial econometric is already broadly used and trusted methodology. Authors Lesage(1999), Anselin(1999), Elhorst(2011) and Vega & Elhorst (2013) provide theoretical description how the methodology and models work presented below in this chapter.

LeSage(1999) in his book “The Theory and Practice of Spatial Econometrics” gives detailed description of the differences between traditional and spatial econometric. They are reflection of spatial dependence and spatial heterogeneity. These two are ignored in traditional methods of econometric modeling as they violate Gauss-Markov assumptions about fixed independent variables in the sample and about existence of the single linear relation on the sample. Dependency among the observations in the sample means that territorially different measures suffer from influence from each other or from different economic processes from another location. Formulation can be as following:

illustration not visible in this excerpt

In this equation y is specific economic indicator and i & j admit different geographical units. Lesage(1999) brings good example of how unemployment and labor force measure can be wrongly measured. If we take into consideration that laborers are mobile not only inside the country but across the borders, becomes vivid that labor force or unemployment rates measured according to where people live can give misleading information on country’s unemployment rate, and accordingly shows spatial dependence.

Following the above mentioned spatial spillovers play an important role on development pattern of modern economy. Heterogeneity herewith reflects the variation of observations and the relation among them through time and space. If we acknowledge the existence of spatial dependence, then we should consider that the results of these relations are different. This assumption is opposite of traditional regression methods assuming the constant relation through the whole sample of observations.

To move on, answering the question of how can distance and spatial characteristics be represented in the model – is important. It is done with the inclusion of spatial weights (W) matrix in the model. The W matrix is the main part where all the economic intuition discussed above should be put. It creates the spatially lagged variable - dependent, independent or the error term - in the model for capturing the spatial effects. It computes weighted averages where the closer units have more weights. Have to be admitted that the critics existing towards the spatial modeling is connected mainly to the W matrix. As Corrado&Fingleton(2011) in “Where is the Economics in Spatial Econometrics?” refer that in majority studies spatially lagged variable is significant leading to the existence of spillover mechanism, though it can be taking the effect of other omitted spatial variable and overall giving the misleading results. Even though, the spatial modeling by including W matrix is already widely used methodology and we follow the model forming process in the following paragraphs.

According to Lesage(1999) to create the matrix itself two measures of distance can be used. One is locations on Cartesian system- longitudes and latitudes. Another one is contiguity- this is information about the location of one entity relative to another in the space. Both measures give finally the same results that should confirm the ideas that distance matters. However the choice for forming the weights matrix is longitude and latitude coordinates. Distance based weights are functions of distance between observation j and all other ones forming the diagonal W matrix. It is decaying function, where θ is parameter used for decaying influence connected to distance and dj is distance observations. Formulation is as following:

illustration not visible in this excerpt

Using the W matrix frequently used models are spatial lag and spatial error models. The mixed model is used with the same success as separate ones. However the standard model at Paelinck(2012) is assumed to be spatial lag model. Which is at Lesage(1999) described as The first-order spatial AR model(FAR). The formulation is following:

illustration not visible in this excerpt

Where each matrix has the following general form defined at Drukker at al. (2013):

illustration not visible in this excerpt

The focus of our interest is diagonal weights matrix where we have zeros on the diagonal as replication that a country can’t be on any distance to itself. The non-diagonal parameters are weights, in our case they are computed using longitude and latitude coordinates of the capital cities of particular TC located on ith row and jth column in the matrix. W in the model is normalized inverse weights matrix where the increasing value of non-diagonal parameter means decreasing of the distance between units. The weights calculated are row normalized as to sum up to unity across the rows.

Further on models we should admit that the FAR model shows dependence of one economic measure only on the same measures from spatially close or farther located entities. However in this work we are also interested to study the determinants of reforms that are different economic variables. The model needs to have additional X matrix of explanatory variables. As maintaining the reliable model and accordingly – reliable results, the spatially lagged exogenous variables are of interest to be included in the model. Considering the above discussion our main working models will be the following ones, Lesage(1999):

1. The mixed spatial autoregressive-regressive model (SAR) – the name comes from the combination of standard regression variables and spatially lagged endogenous variable. The matrix X is added to the previous given matrixes with dimension nxk that will be given below with the specification of spatial Durbin model:

illustration not visible in this excerpt

2. The spatial Durbin model (SDM) – that is exponential form of SAR model by including the additional set of exogenous variables, now forming the spatial leg of all the observations in matrix X with the W matrix. These variables are constructed as averages from similar measures of neighboring countries.

illustration not visible in this excerpt

Another difference from theoretical models in the model presented in this paper is the constant term in the regression. Depending on this information X matrix will contain first column of ones and the general form of the X matrix will be as following:

illustration not visible in this excerpt

SDM model is considered more preferable in studying global spillover effects by Vega & Elhorst(2013), though in the empirical part I have presented results from the models considered as more appropriate by comparing AIC(Akaike Information Criterion) and BIC(Bayesian Information Criterion) statistics. As the study involves panel data sample, the used estimation method for the above discussed theoretical econometric models is method of maximum likelihood estimation. As simple least square estimation in case of panel data gives biased and inconsistent estimates. Lesage(1999), similar to Anselin(1999), brings steps for estimating ML regression for SAR model. It starts with two OLS regressions on standard model:

illustration not visible in this excerpt

The next step is calculation of residuals for both models:

illustration not visible in this excerpt

With given residuals we find the ρ maximizing likelihood function:

illustration not visible in this excerpt

After getting estimated ρ, next is computing beta coefficient and the variance of ε:

illustration not visible in this excerpt

For SDM model the process works in the same steps. Though β_1 and β_2 are expressed with the following notation:

illustration not visible in this excerpt

And the log-likelihood function gets the following form where C is not nesseccary constant. Given the ρ ̂ maximizing log-likelihood function next is computation of the beta coefficients and σ ̂^2:

illustration not visible in this excerpt

Maximum likelihood estimation gives possibility to use several options for estimating. Both models can be estimated by Random effects estimator or by fixed effects estimator. The preference between former two will be made based on Hausman test results. Final interpretation will be based on the estimations from the preferred model.

Above described theoretical part gives general idea about the model and methodology used. The following paragraphs present the models specific for testing hypotheses. The first main hypothesis of this work studies the overall spillover effect through the whole sample of countries. Two different models are used for its testing with different options. Finally the four different models estimates are obtained to compare and analyze the hypothesis question. Equation below presents a complex model where all kind of spatial interactions are included.

illustration not visible in this excerpt

Where 〖OI〗_ct stands for the overall index of reform in country c at time t, GDPGR_ct is growth of GDP per capita (in country c at time t), UN_ct stands for the unemployment rate in the country, INF_ct is measure of inflation, EXP_ct stands for the measure of goods and service export, FDI_ct is variable depicting foreign direct investments level in the country and DI_ct is average democracy index. Z_ct is used for depicting all the exogenous variables that will be estimated with W matrix. Depending on our assumptions and interest spatial coefficients – ρ,λ,ζ - in the equation (4) will take zero value and transform to specific models – SAR or SDM, accordingly. For testing the first hypothesis the model is estimated on the full data sample. Moving on the second hypothesis the sub sample for EU countries and countries in Caucasus region is used. Focusing mainly on the results from EU countries as the spillover effect depended on Caucasus region short sample cannot be considered fully reliable.

Different variables appear in the model and testing of fourth hypothesis. It is robust checking of the obtained results including EBRD reform indexes. The robustness is examined against the model with three Campos-Horvath reform indexes. Campos and Horvath(2012) constructed reforms indexes in three sphere- Internal liberalization, external liberalization and privatization. The results including their internal liberalization index is compared to the same model estimated by EBRD price liberalization index. The choice is according to similarity in underlying objectives of what they measure. By the same logic external liberalization index and EBRD Trade & Forex system index are used in the model used for comparing. The last one is privatization. Campos-Horvath privatization index covers large scale and small scale privatization processes; while EBRD measures them separately. To make the obtained results comparable is used the simple average index from EBRD LSP and SSP indexes.

For sensitivity checking I estimate additional models. As explained above, in the initial model are included all the explanatory variables. According to the significance of obtained coefficients and detected correlation among exogenous and endogenous variables the new models are formed by excluding less significant variables or on the contrary, including new one – lagged dependent variable together with the spatially lagged one. The results are presented in chapter 6.

5. Model Estimation and Results

5.1 Determinants of Reform

This part of the work presents the empirical results obtained using the data given in the chapter 4.1 and applying the methodology described in chapter 4.2. In the discussion below is presented the results that help to show the spatial interaction among reforms and other economic indicators through majority of transition countries. According to the study question the econometric models, mainly SDM, provide the estimates presenting classical coefficients capturing direct effect and also coefficients capturing spatial influence on reforms in the country. To test the first hypothesis equation (4) from chapter 4.2 was estimated. First by referring the coefficients: λ,ζ to be equal to zero leading to SAR model. Assuming that in the scope of our interest, together with the direct determinants of reform, is the spatial impact of reforming in other countries on the reforms in the country of our choice. Secondly, SDM model is estimated referring only ζ to be equal to zero and already estimating the spatial influence of exogenous determinants of reform among sample countries.

In the Table 5.1 are presented results from the SAR and SDM models by using the methodology of random and fixed effects estimators with overall reform index as dependent variable. Needs to be admitted that AIC and BIC statistics showed the preference of SDM model to SAR (the Table A.1 in appendix A provides the statistics), but in the Table 5.1 are reported all the results to compare how the coefficients change when spatial inference of exogenous variables is also included in the model. In the chapters 4.1 and 4.2 is described the W matrix, its construction methodology. To understand and see what the matrix looks like and how it works in the appendix A Figure A.1 shows the graphical representation of W matrix for the full sample. The white boxes in the diagonal admit zero, as each country capital is on zero distance with itself. Non- diagonal boxes are in different colors switching from white to black. The darker is the box the closer the capital cities of particular countries are situated and the higher is the parameter value standing on that color.

Table 5.1 Determinants of Reform

Dependent variable: Overall Reform Index

illustration not visible in this excerpt

Source: Author’s computations - Stata output. Method: RE-Random Effects, FE-Fixed Effects, FDI stands for-Foreign Direct Investment, W in front of exogenous variables depicts coefficients capturing spatial interaction.

Before moving to the interpretation part, the expected signs of spatially lagged variables should be stressed out. As the main purpose of studying the spatial interaction in the sphere of reform is that countries located closer adapt good practice of reforming from each other and implement it, the expectation of the coefficient of spatially lagged reform index is to be positive and significant. For the spatially lagged exogenous variables expected signs are in line with classical coefficient signs for the relevant explanatory variable.

In column 1 and 2 are reported estimations from SAR model. Random and fixed effects estimations are almost same in value. The significance of each coefficient is also unchanged. In columns 3 and 4 coefficients from SDM estimation are slightly higher in values and coefficients standing for inflation are not significant any more. Expected signs of the coefficients are met with actual results in all models except unemployment and export coefficients. GDP growth and FDI have positive signs. Inflation and average democracy index have negative signs. The main coefficient of our interest - spatially lagged reform index is positive and significance. While coefficient of spatially lagged GDP growth has negative sign.

Interpreting obtained results the hypothesis is proved - about spatial informational spillover. Spatial coefficient is positive in all models and significant, though in the SDM model where spatial interaction among reform and exogenous variable is also included the value decreases. The logic behind spatial coefficient is that the shorter is distance, stronger is informational spillover effect. The reform implemented successfully in closer country is spread to particular country by 0.22 (FE estimation) successes than in the country of origin. Reforming rises by 0.22 of joint impact of distance and neighboring reforms. In TCs reality it is vivid. The experience of success or failure is accordingly adapted by others. Though the era of technology and globalization works in favor of informational spillovers, neighboring and distance still matters according to obtained results. The higher value in SAR model can be explained by the omitted interaction that is eliminated in SDM model by considering spatial exogenous variables.

The spatial coefficients of exogenous variables reflect indirect effects on reform. GDP growth in one country has impact on the reforming process in another country. However the relation is negative and significant, while we expected positive coefficient as economic growth in country i should incite more reforming in country j in order to reach similar economic growth. 0.69 and -1.60 are accordingly direct and spatial coefficients. The direct effect is rather small in value, yet it is proving that GDP growth is positive determinants for reforms in the country. Spatial coefficient, on the other hand, is unexpectedly high in negative value covering direct effect.

However all the other coefficients of spatial variables are matching with expectations. The rise in unemployment in country i of TC group instigates country j to implement policies to avoid similar rise of unemployment. Yet the coefficient is not significant at all. While in direct effect coefficient – when unemployment in country i should foster reforming in the same country – is negative opposing intuition. And for all models coefficient is insignificant. The opposite is for inflation. High inflation in particular year terminates reforming in country i. We see that it terminates reforming in other countries, and the closer it is located by -0.16 more the process is slowed down. The spatial effect is again more than direct effect.

The unexpected sign has direct coefficient for export, yet it is significant, showing the accuracy of estimation. This variable is not widely used by economists for studying reforms. I cannot bring comparing from previous studies, but from economic theory the interaction can be defined as positive, while high export is good for economic growth and therefore, should instigate more reforms. Spatial interaction is in line with intuition with quite high coefficient. 1.21 is huge effect; impact on closer country’s reforming is more than one of joint effect of distance and export level. As well as, case of unemployment, further study on export as reform determinant is reasonable.

The next variable is FDI. Its inclusion in the model is mainly supported only from assumptions based on the economic theory that it should have positive impact on reforms. As FDIs flow into more industrialized and liberalized countries, than to those with a lot of barriers. And itself they are directed to more development and more reforming. Obtained coefficients are in line with intuition. Both direct and spatial effects work for more reforming. Again spatial coefficient is higher in value, though it is insignificant. Interpretation is as following, country i obtaining information about rising FDIs in country j works more on reforming. It rises by 0.63 of joint impact of distance and FDI level in j country.

Strong and important determinant of reform is average democracy index. It was expected to be negative, as the measure of democracy used in the paper takes values from 1 to 7, the lower the value the higher is the level of democracy in the country. The negative coefficient works in the way that the more democratic is a country the less it terminates reforms. Coefficients are all of almost same values that emphasize the importance of civil and political freedom in the country for real and successful reforms. And even more importance has free and democratic system in neighborhood.

Overall the hypothesis on average reform index and its determinants considering spatial spillover effects is proved; as reported results empirically strengthen the idea. Not considering the spatial effects countries are able to reform, but in open economy informational spillovers do exist and do have certain impact on reforming processes.

5.2 Determinants of Separate Reform areas

Reforming covers different spheres and EBRD and other institutions working on it are constructing and studying separate spheres of reforms. In this paper are used 5 reform measures, five indexes measuring progress and level of development in different fields. Studying reform determinants, therefore, involves studying determinants of each separate indicator of reforming. The idea behind is that significance of determinants and level of influence should be different on each of them. Staehr (2005) refers the sequencing together with the speed of reforms important factor for studying different relations, he addresses growth determinants. Campos & Horvath (2012) studying reform determinants connect the issue also to reform sequencing; as reforming in separate spheres did not started at one time. To refer the question of which determinants are most influential for separate reforms Table 5.2 below presents results from the model estimation with the same exogenous variables on the left hand side, but on the right hand side having each reform index as dependent one.

Examining determinants of large scale privatization and small scale privatization starting point is direct effect coefficients. The coefficients are reported in 1 and 2 column. Only coefficients for INF, FDI and Di are with the same positive effect as on overall measure of reform. Except inflation, the other two coefficients are significant. Flow of investments in the country and democratic system is supportive for privatization; while high inflation terminates privatization process. Surprisingly, export has negative coefficient addressing that it interrupts privatization processes. Still the coefficient is significant and we should consider it as important determinant for small and large privatization processes. Different direction of influence is found from GDP growth and unemployment on LSP and SSP. On small scale privatization growth has positive and significant effect, 1% growth leads increasing privatization by 1.68. This is the highest value of influence of GDP growth and it was expected as growth is catalyst for reforming, though for LSP we detect negative relation. In the same logic, unemployment has the expected influence on SSP. Rise of unemployment leads to more privatized small scale entities.

Table 5.2 Determinants of Separate Reforms. Model: SDM

illustration not visible in this excerpt

Standard errors in parentheses below each coefficient.*** p<0.01,** p<0.05,* p<0.1. Source: Author’s computations -Stata output. Method: RE or FE, depended on Hausman statistics. W in front of exogenous variables depicts coefficients capturing spatial interaction. FDI-Foreign Direct Investment; LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System.

But for large entities, higher unemployment is not supportive. The intuition behind can be that state owned large entities can settle the unemployment rate on some level depending on the situation as state can decide to employ more or less people. However, it brings country far from industrialization. To move on spatial coefficients, first should state that the spatial affect of LSP and SSP from other countries are positive; emphasizing the important role of adapting successful precedents. The most influential spatial determinants for both type of reform are inflation and export. The less expected spatial relation is obtained from GDP growth. For these two reform areas spatial determinants replicate the results as in case of overall reform.

From the rest three reform areas, two – price liberalization and trade & FOREX system – can be considered as types of internal and external liberalization. I discuss them from this viewpoint and compare which determinants they have in common, and how it is explained. From direct effect coefficients GDP growth and export are important and significant. For TFS only growth is significant and in line with intuition. Growth is unconditionally main determinants of any kind of reform. The same is true for democracy index, though for separate reform areas it lost significance but kept positive sign. FDIs are of same importance for both reform areas. The only different determinant is inflation. Inflation is conducive for price liberalization as coefficient says, though in economic reality inflation is destructive for price liberalization. The process losses effectiveness when high inflation is presented. From spatially lagged determinants democracy in closer countries is supportive for price liberalization in particular country, as expected, though it works in opposite direction for trade & FOREX system. The former needs further analyzing.

Reported results in column 3 are for government and enterprise restructuring. Empirically the most important determinant of this reform area is democracy. Economic theory and expected importance is met with coefficient. This reform area works directly for government restructuring and democratic and free system is main catalyst for it. The spatial effect coefficient for democracy is of same importance, its significant determinant for reforming in area of government and enterprise restructuring. For it democratic neighborhood is most affecting.

5.3 Reform determinants: EU and Caucasus region sample

Studying reform determinants by considering spatial interaction and spillover effects leads to reasonable results and conclusions. As empirical estimations are matching expectations and distance matters for determinants to have stronger impact on reforming. However, in this paper is given hypothesis that together with distance membership of European Union or locating in the same region is important influence, strengthening and increasing the previously described and analyzed relations. To visualize location and distance characteristics for this sample of countries, in appendix A Figure A.2 represents W matrix graphically. By estimating the same model as for the main hypothesis but only on the shorter samples I expect the coefficients to be more significant and higher in value.

In the Table 5.3 are reported estimation results for average reform index as well as for separate reform indexes, respectively in columns 1 to 6. From direct effect coefficients on overall reform only inflation is presented with increased impact. FDI is left almost with same value and unemployment show again negative influence, though in reduced value. Reduced influence is showed by the growth and democracy coefficients. The scope of our interest spatially lagged reform index and its coefficient is still significant and referring higher influence on reforming in closer countries. The explanation of this increased effect can be that EU member TCs are characterized with faster informational spillover and more adaption of successful reforms from related country’s experience. Moreover, membership of EU evolves and means sharing of main ideas and policy implications for future development path. The economic indicators, their spatial effect should be intensified by membership. Though only democracy and unemployment work in this direction, and growth, that has better value than in case of whole sample, though is still negative, that opposes predictions.

Table 5.3 Determinants of Reform. Model: SDM

illustration not visible in this excerpt

Standard errors in parentheses below each coefficient. *** p<0.01, ** p<0.05, * p<0.1 Source: Author’s computations - Stata output. Method: RE or FE, depended on Hausman statistics. W in front of exogenous variables depicts coefficients capturing spatial interaction. FDI- Foreign Direct Investment,OI-Overall Reform Index, LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System.

To move on separate reform areas, not all the determinants improved in terms of value and significance. For large scale and small scale privatization most considerable are changes in direct and spatial coefficients of LSP determinants. GDP growth changed the sign from negative to positive and increased significantly in value. The intuition can lead to one explanation – in EU member TCs growth has more impact on privatization, as relation is bilateral. Privatization brings more growth of GDP per capita, while GDP growth dictates more reforming in privatization. Inflation and FDIs follow the logic. In spatial inference, democratization shows much improved influence and determination of reform in privatization. And spatially lagged LSP reform refers to the power of informational spillovers in EU.

The rest three reform areas report most considerable improvement and conformity with the expectations and predictions. GDP growth appears with highest impact determinant of government and enterprise restructuring, it has higher value than in all TCs sample estimation. The unstable variable in our sample - unemployment for all three reforms shows positive sign of influence. This variable is mostly tested in reform studies together with growth and democracy mainly obtaining stable positive results. The instability in my estimations should lead to further studies or improved specification of the models. Inflation is in line with prediction, except GER it proves the more influence on reforms in EU member TCs together with democracy. Yet coefficients capturing spatial interaction depict less improvement and less conformity to the hypothesis idea. For trade & FOREX system the spatial rho coefficient has changed sign, that is liberalizing trade and FOREX conditions in closer countries incite the particular country to lower the TFS reforming by 0.17. This is in controversy with predictions.

Considering above discussion in this chapter obtained empirical results are accurate. The following explanations are intended to show that situating in one region has similar effect as membership of e.g. EU. The Table B.1 in appendix B provides the results. Understanding that 3 country and overall 54 observations cannot lead to reliable results, I decided to bring the results as additional estimation that will show the point of continuing or beginning studying this issue with more reasonable methodology. Still the results are worth of discussing. Again most important is corresponding coefficients when dependent variable is overall reform index.

The improved values of coefficients are found in the same coefficients as in EU sample estimation. GDP growth considerably changed for large scale privatization, only government and enterprise restructuring suffers from negative influence from gross domestic product growth. Inflation characterizes with most considerably increased influence. Showing that in Caucasus region countries high inflation incites more reforming in any area, except GER. For inflation that is surely true, as Georgia, Armenia and Azerbaijan from early periods of transition suffered from high inflation, and correspondingly with more incentives for reforming. Government and enterprise restructuring separately does not have specific determinant for this sample estimation. Only spatial interaction from GDP growth can be considered as more or less significant. It is notable that spatial rho coefficient is negative for all reform areas, except GER. It is controversial to the spatial spillover effects that in small region where countries are located closer to each other, accordingly their capitals too, should be found higher informational spillover for reforming and higher degree of sharing experience.

Above is already mentioned that, this chapter served checking of additional determinant of reforming and faster informational spillovers according to two characteristics: membership to EU and situating in one region- in this case Caucasus region. The results suffer from small number of observations that hardly makes results reliable, though considerable conclusions can be still arrived at. First, being part of any union means being part of merely similar development plan and path towards one point of democratization, industrialization or reforming. This is founded in results as predicted, though different specification of model rather than only inclusion of specific sample can very likely bring to more reliable results. The similarity is vivid when countries are located in one region, the region characterizes with some geo-political features that lead naturally to adapting already achieved success or avoiding failure considering neighbor’s experience.

6. Robust and Sensitivity Checking

6.1 Reform determinants: Comparing EBRD and Campos-Horvath indexes

In this chapter I refer to the criticism towards EBRD reform measures that are used in this work for main empirical models. The critics are mainly towards ambiguous measurement methodology. That the scores assigned to some level of reforming in the country finally have maximum measure pointing to the perfectly industrialized country and economy. However reforming does not have limits and is not stopped after reaching the point e.g. 4+. Campos and Horvath (2012) criticize variables behind each reform index and the transparency of how each aggregate index is obtained. Authors constructed new indexes for three reform area that I use for robust checking of results obtained by including EBRD indexes n the model.

In the Table 6.1 are reported results from six models. Three estimated by EBRD indexes and the rest three by Campos- Horvath indexes. The data sample for estimation was shortened to 7 years - 1995-2001. As the paper covers years from 1995 till 2012, while Campos-Horvath indexes are available for the period 1989-2001. Again the SDM methodology was used. As explained in chapter 4.2 average privatization index from EBRD and privatization index of Campos-Horvath(CH) are assumed to measure similar concepts. In the same logic are used price liberalization and internal liberalization indexes, and trade & FOREX system and external liberalization indexes.

The results indicate interesting issues to be discussed. First, looking at privatization index and determinants that influence it, differences are found. GDP growth is positive determinants for privatization measured by EBRD, while it is negative and less significant for CH measure of privatization. However, in their paper relation of CH index and GDP growth is positive as predicted by the authors. Difference influence is found testing inflation as determinant of both reform measures. While for

Table 6.1 Determinants of Reform. EBRD - CH Comparing. Model: SDM

illustration not visible in this excerpt

Standard errors in parentheses below each coefficient. *** p<0.01, ** p<0.05, * p<0.1 Source: Author’s computations - Stata output. Method: RE or FE, depended on Hausman statistics. W in front of exogenous variables depicts coefficients capturing spatial interaction. FDI-Foreign Direct Investment, PRV-Privatization, Pl-Price Liberalization, IL-Internal Liberalization, TFS-Trade and FOREX System, EL-External Liberalization. CH-stands for Campos-Horvath Indexes.

EBRD privatization it is in line with expectation and for CH it is not. Moving to export coefficient, both models report negative impact on privatization reforms. What is good and totally matching our expectations for both reform measures are coefficients of unemployment, FDIs and democracy. In values they are close to each other, though in CH model all of them are significant too. Moving to spatial coefficients GDP growth coefficient is again for both measures of reform negative. This type of relation is found almost in every model that is affecting the stability of our results and indicating that either specification of the model or size of the sample needs to be reviewed for further studies. Another determinant – democracy index- shows unexpected relation, as well. If we look at the effect, that autocracy in other countries refers to more privatization reform in particular country, seems still reasonable, though in general, opposite is expected. The rationale behind can be the willingness of the country not to become autocratic and possible way out of it is considered to be privatization.

The rest spatial coefficients and the most important – spatial rho – are significant. In value all coefficients regressed on EBRD index, except spatial rho is higher. Spatial effect of reforms implemented in closer countries is higher in CH models. But for concluding comparing of these two models evolving one area of reform, can point out that the results from models including privatization indexes in this work are quite robust and depict the estimations close to real relation among variables.

Moving to second reform area – price liberalization similar to internal liberalization, the different determinants appear important for each of them, reported correspondingly in column 3 and 4. The similar coefficient has only FDI from direct effect coefficients. The value is high, equaling 0.91 and 0.85 that is almost equal in value, indicating that FDI is determinant of reform, and its inclusion in the model is rational. From spatial coefficients inflation and again FDI is in line with expectation for both measures of reform. Inflation leads increase in reforming of price liberalization measured by both indexes. While FDI shows larger influence on reforming measured by EBRD than measured by CH. Itself coefficient equaling to 4.48 is too high comparing to the same coefficient from other models in this paper. Unexpectedly, spatial rho is far from expected sign.

The 5 and 6 columns represent results for reform area of trade &FOREX system measured by EBRD and external liberalization measured by Campos-Horvath. For this particular reform GDP growth, inflation and democracy are positive determinants checking on both types of reform measure. The largest difference in value is between democracy coefficients, though both are significant. The rest direct effect coefficients appear to be negatively influencing, moreover all are insignificant.

The spatial interaction is well described by both models. Spatial influence from GDP growth on reforming in the area of trade is high – 2.56 is huge impact of joint distance and GDP growth from closer countries. While on CH index we obtain negative coefficient again, though considerable results are obtained for unemployment rate. Significant influence equaling to, respectively, 6.59 and 3.22 is again huge effect, yet positive influence was expected. From rest of spatial effect coefficients all work in favor to our hypothesis. Spatial rho has different estimates by sign, though in line with expectations is coefficient from regression on CH index. Considering the difference in estimates but not the high volume of it, concluding remarks still work for robustness of original estimates.

6.2 Sensitivity Checking: Estimation of Several Models

This work presents model that is used for first time addressing the topic - determinants of reform. It is of high importance and interest to have results that are reliable and bringing strong ideas about importance of each particular determinant and showing that distance matters and spillover effects are also strong. Robust checking is one way proposed in the chapter 4.2. for testing the results. The second method is sensitivity checking by stating the initial model with the main determinants and estimating several times by including each time one different additional determinant of reform. The results for sensitivity checks with initial model including overall reform index and two determinants – GDP growth and democracy index are presented in the appendix B Table B.2 The decision to state initial model in this specific form depends on the coefficient estimates through all models reported in above parts. Another way is to eliminate variables according to correlation between exogenous variables, though as reported in appendix A table A.2 no strong correlation was detected among them.

In the initial model all the coefficients are significant and match with expected sign except spatial coefficient of GDP growth. Spatial rho is positive and in line with prediction. Initial model is starting option for testing the main hypothesis. In the second column of Table A.2 is reported results from model after adding unemployment rate as determinant. We see that GDP growth and democracy index still are left positive and significant, almost with the same values. The same is true for spatial effects coefficients of GDP growth and democracy. While new determinant in direct effects part has negative sign opposing the economic intuition and predictions of possible positive influence. Spatial coefficient is positive, not significant. Spatial rho is left also with positive impact and significance, with almost no change in value. Inclusion of unemployment rate did not bring additional good effect.

Next is third model in column 3, including inflation as additional determinant. Inflation coefficient itself, direct and spatial, is met with predicted sign and impact on reforms. Yet it brought to reducing values of other determinants still staying significant. It shows that inflation absorbs some part of other interactions or before its inclusion the model suffered from omitted variable bias. Cannot be said the same for fourth model including export as new additional determinant. The initial determinants stayed again significant with almost equal values as in initial model. Spatial growth coefficient is again negative. The new determinant indicates that direct effect of export level on the same country’s reforming is negative, and spatial effect from other countries’ is positive. It is not completely in line with expectation. The last model is with including foreign direct investment as another determinant. As inclusion of inflation, inclusion of FDI made model more complete, as all the other relations stayed unchanged with same significance and almost same values of influence. And FDI itself appeared important determinant directly, as well as, in spatial interaction.

Compared to initial model each variable brought new effect in the model with few changes in initial interaction. Comparing each model and coefficients to first model results from Table 5.1 (results from SDM model column 4) the sign of influence is replicated only with larger values. The only variable changing the values considerably is inflation. However, before concluding on overall results another model for sensitivity checking is presented in Table B.3 in appendix B, where initial model includes lag of reform as main determinant.

Inclusion of lagged reform variable means taking into account reforming from previous year. Expected impact on current reforms is positive as successfully implemented reform or progressive process is leading to continued reforming. Empirical results show that in each model lagged reform and spatially lagged reform coefficients are significant and in line with prediction. But even in initial model inclusion of lagged reform variable induced loss of significance of all other determinants’ corresponding coefficients and decrease in their values - translating into decreased level of influence.

In this model adding other determinants is not resulting in considerable change of initial model. None of them, except export coefficient, is significant. However, export and inflation appear with opposing signs of influence. Direct effect stays positive for unemployment and FDIs, though spatial coefficients indicate on negative relation. The fact that lagged reform has the highest value (highest influence on reforming in current year) and correspondingly, other determinants have significantly lower values can be explained by one reason. In previous year level of reforming was determined by different economic indicators that are included in models presented in this work. Include these variables together results in absorbing larger part of interaction by lagged variable than by other direct coefficients.

Comparing again to the first model (reported in Table 5.1), the direct determinants those are stable through different estimations: GDP growth, democracy index, FDI and inflation replicated their results in robust and sensitivity checking. Unemployment and export stayed unstable. From spatial coefficients spatial rho, spatial FDI, DI and inflation coefficient showed stable performance, while again unemployment, export and GDP growth are unstable.

Finally, we can address the results more robust and less sensitive to changes while evolving different determinants. To be precise, sensitivity checks indicate that the initial models are better including few variables, than those including additional determinants of reform.

7. Concluding Remarks

This part of this work presents summation of results presented in chapter 5 after empirical testing of the four main hypotheses. Following the obtained results according to their significance provided below, the related policy implications and further, possible, improvements needed in the specification are also discussed. The paper serves to show, using different methodology than used in related works, the determinants of reform. The use of spatial econometric brought new dimension of measuring and checking reform determinants and it is the focus of the first hypothesis.

The model estimation on the whole sample checking for spillover effect and importance of distance showed significant relationship in line with expectations. Spatial coefficient that indicates on spatial interaction in all models is of high significance. First, in the model, empirically testing only the spatial effect of endogenous variable, obtained coefficient is equal to 0.71. This is quite large impact, proving that distance matters and closer located TCs adopt successful reforms from neighbors. In the expanded model including spatial effect from exogenous variables the significant coefficient decreases in value, though influence is still indicating that informational spillovers are part of development in transition process. The spatial impact of exogenous variables replicates the same direction of influence as direct effect coefficients.

The direct continuation of the first hypothesis, focusing on specific sample and larger spatial effects is tested on EU sample countries and Caucasus region countries. In the first case, expectations are empirically proved; giving purpose to declare that model works and spatial dependence is strengthened by additional uniting characteristic that successful reforming is more transmitted. Here the membership of European Union is itself an indicator that member countries should choose similar direction of development and on that path reforms implemented can be same and adopted from the overall successful practice. These countries still have a space for improvement by more focusing on exporting and entering EU market, reforming towards liberalizing business formation and towards reaching common standards of producing. Still being in the list of transition countries means the governors should work to use the advantage of union membership more productively. However, testing same hypothesis on Caucasus region countries, spatial coefficients are far from predictions, indicating that in the region successful reforms lead to less reforming by neighbors. Here another recommendation for further studies arises. The obtained results can be misleading, as the used sample is very short and better econometric approach can bring to proving or rejecting presented results.

The above interaction was tested on each separate reform area. Firstly, the aim was to prove that results obtained using overall reform index works for every separate reform area. Secondly, I intended to test the fourth hypothesis and check whether the same economic indicators are significant determinants for separate reforms. The importance and influence of spatial interaction was proved for all reform areas except price liberalization, in value it is close to zero. As for determinants of separate reforms, GDP growth is significantly influencing small scale privatization, price liberalization and trade and FOREX system reforming. Unemployment is positively affecting only small scale privatization. Inflation works as economic predictions were for all spheres except price liberalization that is explained by high pressure on prices during inflation. FDI appears important for all areas, but mainly it affect largely reforming in large scale privatization that is fully applicable, as investments firstly are directed to large and medium size fields. Another considerable determiner of LSP reforming as well as other areas is democracy index.

Continuing former relation of democracy index and reforming processes should admit that as reflected in the third hypothesis democratic system is vital for reforming and successful ending of any kind of reform in every TC, for any sample of countries and any length of the study period. The coefficient appeared stable in every estimated model. Robust and sensitivity checking proved the general estimation results that democratic countries reform more and react on informational spillovers. The issues related to different understanding of the values of democracy or ambiguous measurement can put under question the reliability of the results. On the other hand, it can be purpose for further studies by using democracy measures provided from other institutions, nowadays available only for shorter sample periods.

Questions arose regarding reform index measurement. It is discussed in the chapter 6.1 where robust checking was done by testing main hypothesis – defining main reform determinants on two sample including two different reform measures. Estimations resulted in similar results. Determinants significant in estimation using Campos & Horvath reform indexes are significant and repeating the same pattern of influence in the estimation using EBRD indexes. Also in data description and methodology part is presented intuition behind using the particular economic measures as exogenous variables in the reform determinants econometric estimation such as GDP growth, Inflation, Unemployment. While there are works also discussed in literature review part where reforms are used for determining GDP growth or Inflation level in the country. This is causing doubts for bi-directional relationship that is vivid from the economic development that reforming leads to higher levels of economic measures and on the contrary, growth of certain economic indicators catalyzes more reforming in the country. In the work data used is gathered from independent sources that should eliminate any correlation among the variables. Real world data can be far from normally distributed and testes performed prove this but as they are taken from official statistical sources any assumption regarding bias from the beginning was considered to be omitted from the estimations.

Provided concluding remarks are quite reasonable as we see from their significance and robust and sensitivity checking. As results meet the predictions, concluding remarks can be drawn related to policy implications, as well. Firstly, policy makers evaluating some economic indicators can predict the reforming process considering direct effects estimated from models. Secondly, existence of spatial interaction is signal for policymakers to guide by the practice of neighbors. This interaction is good signal for policy makers to adopt successful reforms, to guide by the tested practice. It is more important to consider failures as informational spillover can work positively, as well as, negatively.

8. Conclusion

Reforms are implemented in transition countries to continue the process of forming market economy. Considering importance of the topic this paper addressed the determinants of reform. Previous works provided specific economic indicators that determine how reforms will end. One of the main purposes of this study was to approach the former concept by considering spatial interaction among countries, meaning that economic processes in one country do affect the same or related fields in other countries and at the same time, testing the direct and spatial determinants of reforms on different samples and different areas of reforming. In every model I included reform indexes as dependent variable and different economic indicators directly and jointly as independent variables with spatial W matrix for catching the influence of distance.

Overall, 4 data samples were used. The first one was estimated for 24 TCs and 18 years resulting in significant coefficient of democracy index, equal to -0.05633 and significant spatial rho coefficient equal to 0.71 by SAR estimation. Idea behind figures is that democratic and free governing system in the country works for more rapid and successful reforming. The minimum negative influence comes from the nonexistence of perfect democracy that leaved space and ability for terminations in reforming. From other determinants GDP growth, inflation and FDI also have coefficients in line with expectation. Incoming investments in the country create new possibilities and build new ways for reforms and development of different spheres. For the samples of EU member countries with 10 TCs again for 18 years and Caucasus region transition countries, coefficients for the first case replicate the full sample results with higher values for spatial rho coefficient equal to 0.57 and coefficient of -0.04 for democracy, while the second sample for 3 countries results in results out of prediction. Spatial rho proves that economic measures in different countries do influence each other, especially successful reforming. The fourth sample estimated is for shorter period of 7 years. For robust checking of the results were estimated new models for the period of 1995-2001 years including first EBRD reform indexes and then Campos- Horvath indexes. The results are matching for similar reform areas for both measures of reforms, proving the correctness of used methodology and similarity of indexes in terms of correct measuring and capturing the right relationship.

The results can be suggestive for policy makers tracking and controlling reforming in the country. They can consider how rise of GDP growth, inflation, FDI and governing system affects reform levels directly and through spillovers. Unstable coefficient for unemployment and export variables, and not in line results for Caucasus region sample indicate for further research necessity, to be proved the accuracy and significance of results for all types of models.

Bibliography

- Anselin, Luc, 1999, Spatial Econometrics, Bruton Center, School of Social Sciences, Richardson, TX 75083-0688
- Babecka Jan & Tomáš Havránek, 2013, Structural Reforms and Growth in Transition: A Meta-Analysis, Institute of Economic Studies, Faculty of Social Sciences, Charles University in Prague.
- Blanchard, Olivier 1997, The Economics of Transition, Oxford.
- Byung-Yeon Kim & Jukka Pirttila, 2006, Political constraints and economic reform: Empirical evidence from the post-communist transition in the 1990s, Journal of Comparative Economics 34 (2006) 446–466.
- Campos Nauro F. & Fabrizio Coricelli, 2013, Economic growth in the transition from communism, Handbook of the Economics and Political Economy of Transition. Routledge, pp:421-428
- Campos Nauro F. & Roman Horváth, 2012, Reform redux: Measurement, determinants and growth implications, European Journal of Political Economy 28 – 227–237
- Castanheira Micael & Gerard Roland, 2000, The Optmal Speed of Trnsition: a General Equilibrium Analysis, International Economic Review, Vol. 41, No. 1,
- Corrado Luisa & Bernard Fingleton, 2011, Where is the Economics in Spatial Econometrics? Serc discussion paper 71.
- Drukker David M., Hua Peng, Ingmar R Prucha. & Rafal Raciborski, 2013, Creating and managing spatial-weighting matrices with the spmat command, The Stata Journal, 13, Number 2, pp. 242–286
- Elhorst, J. Paul, 2011, Spatial Panel Models, University of Groningen, Department of Economics, Econometrics and Finance
- Elhorst, Paul, Eelco Zandberg & Jakob De Haan, 2013 The Impact of Interaction Effects among Neighboring Countries on Financial Liberalization and Reform: A Dynamic Spatial Panel Data Approach, Spatial Economic Analysis, 8:3, 293-313, DOI:10.1080/17421772.2012.760136
- Falcetti Elisabetta, Tatiana Lysenko & Peter Sanfey, 2005, Reforms and growth in transition: re-examining the evidence, EBRD. Working paper No. 90, Prepared in March 2005.
- Fidrmuc, Jan, 2003, Economic reform, democracy and growth during post-communist transition, European Journal of Political Economy , Vol. 19, 583–604
- Fidrmuc Jan & Ariane Tichit, 2013, How I learned to stop worrying and love the crisis, Economic Systems 37, 542–554, Economic Systems journal homepage: www.elsevier.com/locate/ecosys
- Fidrmuc Jan & Elira Karaja, 2013, Uncertainty, informational spillovers and policy reform: A gravity model approach. European Journal of Political Economy, 32 - 182–192
- Hare, Paul & Gerard Turley, 2013, Introduction to the handbook, Handbook of the Economics and and Political Economy of Transition, pp: 1-14
- Heinemann Friedrich & Benjamin Tanz, 2008, The impact of trust on reforms, Journal of Economic Policy Reform, 11:3, 173-185, DOI: 10.1080/17487870802405375
- Intriligator, Michael D., 1998, Democracy in Reforming Collapsed Communist Economies: Blessing or Curse? Contemporary Economic Policy (ISSN 1074-3529), Vol. 16, 241-246
- Lawson, C. & Wang, H. (2004) Economic transition in Central and Eastern Europe and the former Soviet Union: Which policies worked? Working Paper. University of Bath, Bath, UK, Link to official URL (if available): http://www.bath.ac.uk/cpe/workingpapers/transition-5th.pdf
- LeSage, James P. (1999), The Theory and Practice of Spatial Econometrics, Department of Economics, University of Toledo
- Miller, Harvey J., 2004, Tobler’s First Law and Spatial Analysis, Department of Geography, University of Utah, Annals of the Association of American Geographers, 94(2), pp. 284–289
- Paelinck, Jean H.P., 2012, Specification and Identification in Spatial Econometric Models, George Mason University, Estadística Española, Volumen 54, número 177 / 2012, pp. 35-52
- Pitlik, Hans, 2008, The Impact of Growth Performance and Political Regime Type on Economic Policy Liberalization, Kyklos, Vol. 61, No. 2, 258–278
- Roland, Gérard, 2002, The Political Economy of Transition, Journal of Economic Perspectives—Volume 16, Number 1, Pages 29–50
- Staehr, Karsten, 2005, Reforms and Economic Growth in Transition Economies: Complementarity, Sequencing and Speed, The European Journal of Comparative Economics, Vol. 2, n. 2, pp. 177-202, ISSN 1824-2979,
- Svejnar, Jan, 2002, Transition Economies: Performance and Challenges, Journal of Economic Perspectives—Volume 16, Number 1, Pages 3–28
- Tobler, W. R., 1970, A Computer Movie Simulating Urban Growth in the Detroit Region, Economic Geography, Vol. 46, International Geographical Union, pp. 234-240
- Treier, Volker, 1999, Unemployment in reforming countries: Causes, fiscal impacts and the success of transformation, BERG working paper series on government and growth, No. 29, ISBN 3931052109
- Vega, Solmaria Halleck & J. Paul Elhorst, 2013, On Spatial Econometric models, spillover effects and W, ERSA conference papers ersa13p222, European Regional Science Association

Appendixes

Appendix A: Descriptive specifications

Figure A.1 Graphical representation of spatial weight matrix for the whole sample-24 countries.

illustration not visible in this excerpt

Figure A.2 Graphical representation of spatial weight matrix for EU member transition countries – 10 countries

illustration not visible in this excerpt

Figure A.3 Graphical representation of spatial weight matrix for Caucasus Region countries – 3 countries

illustration not visible in this excerpt

Table A.1 List of Transition countries

illustration not visible in this excerpt

Table A.2 SAR-SDM Model Comparing

illustration not visible in this excerpt

Table A.3 Correlation Matrix

illustration not visible in this excerpt

Table A.4 Descriptive statistics for EU sample

illustration not visible in this excerpt

N stands for number of observation, SD-Standard Deviation, FDI-Foreign Direct Investment, LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System.

Table A.5 Descriptive statistics including CH indexes

illustration not visible in this excerpt

N stands for number of observation, SD-Standard Deviation, FDI-Foreign Direct Investment, PRV- privatization, PL-Price Liberalization, TFS-Trade and FOREX System, IL-Internal Liberalization, EL-External Liberalization, CH- Campos Horvath Index

Table A.6 Descriptive statistics of Caucasus Region Sample

illustration not visible in this excerpt

N stands for number of observation, SD-Standard Deviation, FDI-Foreign Direct Investment, LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System.

Appendix B: Regression results

Table B.1 Caucasus Region Sample

illustration not visible in this excerpt

Standard errors in parentheses below each coefficient. *** p<0.01, ** p<0.05, * p<0.1 Source: Author’s computations - Stata output. Method: RE or FE, depended on Hausman statistics. W in front of exogenous variables depicts coefficients capturing spatial interaction. FDI- Foreign Direct Investment, OI-Overall Reform Index, LSP-Large Scale Privatization, SSP-Small Scale Privatization, GER-Government and Enterprise Restructuring, PL-Price Liberalization, TFS-Trade and FOREX System. All indexes are from EBRD

Table B.2 Sensitivity analyses on Whole Sample

illustration not visible in this excerpt

Standard errors in parentheses below each coefficient. *** p<0.01, ** p<0.05, * p<0.1 Source: Author’s computations - Stata output. Method: RE or FE, depended on Hausman statistics. W in front of exogenous variables depicts coefficients capturing spatial interaction. FDI- Foreign Direct Investment, All indexes are from EBRD

Table B.3 Sensitivity analysis on Whole Sample. Model with lagged reform

illustration not visible in this excerpt

Frequently asked questions

What is the purpose of this document?

This document is a comprehensive language preview that includes the title, table of contents, objectives and key themes, chapter summaries, and key words. It is intended for academic use, analyzing themes in a structured and professional manner.

What is the main topic of the research?

The research focuses on reform determinants for transition countries, particularly using spatial econometrics to analyze the impact of various factors on reform progress.

Which countries are included in the study?

The study includes data from 24 transition countries. Some analyses focus on specific subsets, such as EU member transition countries and countries in the Caucasus region.

What is the time period covered by the data?

The data primarily covers the period from 1995 to 2012. However, some analyses using Campos-Horvath indexes have a shorter time frame (1995-2001) due to data availability.

What are the key data sources used in the research?

The primary data sources include the European Bank for Reconstruction and Development (EBRD), The World Bank, The Freedom House, and CEPII (Research and Expertise on the World Economy).

What are the main variables used in the analysis?

The main variables include reform indexes (overall and specific areas like privatization, price liberalization, trade & forex system, government and enterprise restructuring), GDP per capita growth, unemployment rate, inflation, foreign direct investment (FDI), export levels, and a democracy index.

What is spatial econometrics and why is it used in this research?

Spatial econometrics is a statistical method used to analyze data with spatial dependence, meaning that observations are not independent of each other. In this research, it's used to capture the spillover effects and influence of neighboring countries on reform processes.

What is the spatial weights matrix (W)?

The spatial weights matrix (W) is a key component of spatial econometric models. It represents the spatial relationships between countries, typically based on distance. In this study, an inverse distance weights matrix is used, where closer countries have greater weight.

What is the Spatial Durbin Model (SDM)?

The Spatial Durbin Model (SDM) is a type of spatial econometric model that includes both spatially lagged dependent variables (reform index) and spatially lagged independent variables (economic indicators). It's used to assess the impact of both direct determinants and spatial spillovers on reform.

What are the main hypotheses tested in the research?

The main hypotheses are:

• Neighborhood effect is positive for reform undertaking (distance matters for information spillovers).
• Region and union membership matter for different development patterns of reform indicators (EU vs. Caucasus).
• The influence of determinants is different on the results of a reform (democracy is most influential).
• Determinants have different influence on different reform indexes (privatization, governance, etc.).

What are some of the key findings of the research?

Key findings include:

• Spatial spillover effects are significant, indicating that reforms in neighboring countries influence reform progress.
• Democracy is a consistently important determinant of reform, with more democratic countries generally exhibiting faster reform progress.
• The influence of specific economic indicators varies across different reform areas.

What is the purpose of the robustness and sensitivity checks?

Robustness and sensitivity checks are performed to assess the reliability and stability of the research findings. Robustness checks involve comparing results using different reform measures (EBRD vs. Campos-Horvath), while sensitivity checks involve estimating several models with different combinations of variables.

What are some policy implications of the research?

Policy implications include:

• Policymakers should consider both direct determinants and spatial spillover effects when evaluating reform processes.
• Focusing on improving democratic institutions is crucial for promoting reform progress.
• Understanding the specific determinants that are most influential for different reform areas can help policymakers tailor their strategies.

What are some potential limitations or areas for further research?

Potential limitations include:

• Data limitations, particularly for some transition countries and earlier time periods.
• Potential endogeneity issues between reform and economic indicators.

Areas for further research include:

• Exploring different measures of democracy and spatial relationships.
• Analyzing the specific mechanisms through which spatial spillover effects occur.
• Examining the role of institutional quality and governance in reform processes.