Do Crises Instigate Economic Reforms?
Emma Sophie Molenaar
*Supervisor: Dr. Richard Jong-A-Pin
June 20, 2014
Abstract
This thesis investigates the relationship between crises and reforms by looking at whether reforms are more likely during crises, and whether democracies and autocracies implement reforms differently in the presence of a crisis. Empirical evidence on these relationships is still very limited. Using panel data on over 150 countries over the period of 1960-2005, this paper finds evidence that economic reforms are more likely to be adopted in democracies, and that countries converge towards a certain level of reforms. These results are robust to the inclusion of several control variables. Our results also show that the level of reforms in neighbouring countries has a positive influence on reforms. The empirical analysis shows that there is an ambiguous effect of crises. Different types of crises appear to differently affect whether reforms are implemented. This analysis differs from others in the sense that multiple dimensions of reforms are investigated for a large sample.
Keywords: Economic reforms, democracy, crisis JEL Classification: G01, G28, H12, P16
* University of Groningen, Faculty of Economic and Business, Groningen, the Netherlands. Email:
1
Introduction
“The best argument against democracy is a five-minute conversation with the average voter.” ― Winston Churchill
As the quote stated above suggests there are some downsides of democracy. Off course there are many advantages that democracies have over autocracies, but this does not mean that democracies are better in every aspect. This especially depends on whether the autocratic regime has a benevolent leader or not. The effects of political and economic freedom have been the subjects of numerous researches. Many political economists are interested in explaining the relationship between democracy and economic reforms, but empirical research on this topic is still very limited. The focus of this paper is on economic reforms and, more specifically, on the differences between regime types in implementing reforms. Economic reforms are defined as a decrease in regulation and thus reflect economic liberalization. Reforms refer to major policy changes that are not made daily, and enhance both stabilizations and structural changes in any form. One of the questions addressed is whether the characteristic differences between autocracies and democracies influence the ability or willingness to implement reforms. There are several reasons unique to regime types why reforms may not be implemented when necessary. For example, in a democracy there are several interest groups of whom a majority supporting the reforms is necessary, whereas a totalitarian leader is able to quickly implement reforms without others being able to prevent this. Some research has suggested that authoritarian governments are better able to impose painful reforms (Brooks and Kurtz, 2007). However, more recently, the literature found that democratic regimes might have an advantage as they are better able to guarantee that the long term benefits of the reforms are properly distributed (Frye and Mansfield, 2004). The fact that there has not been reached a consensus on the relationship between democracy and reforms is interesting, and justifies more research concerning this topic.
the ‘international interpretations of the roots and implications of the crisis resonate with the ideal and material interests of domestic elites and ordinary citizens’.
Also the severity of an economic crisis appears to facilitate whether market-oriented reforms are implemented (Haggard and Kaufman, 1995). The recent crisis of 2008 spurred reforms and many expansionary fiscal and monetary policies were implemented. This is an example of how crises spark reforms and how quick responses are necessary in order to prevent further deterioration. An increase in the costs of the status quo necessitates countries to reform if this is caused by a crisis, which is seen as one of the pro-reform forces (Galasso, 2014). The purpose of structural reforms is to reduce these costs but also to reduce the chance of a future financial crisis. Off course, this must be done in a manner that least hurts economic growth and welfare in general (Claessens and Kodres, 2014).
Evaluating the interaction between economics and politics, and trying to understand the relationships such as those described above is valuable as it can give insight into why some countries are better able to counteract economic downturns. Previous research also tries to link reforms and crises. Abiad and Mody (2005) look at what influences financial liberalization and find that different types of crises influence financial reforms differently. Our dataset allows us to see if this is similarly the case in real sectors as well as in financial sectors. For a long time, research on economic reforms was constrained and empirical evidence on this relationship is also very limited (Drazen and Easterly, 2001; Drazen, 2002). Many case studies have been performed but these have many limitations and cannot be used to generalize the relationship between crises and economic reforms. Therefore more empirical research, which is performed in this thesis, on this topic adds to the literature.
2
Literature Review
Much literature exists on the interaction between the economy and polity. This section will discuss the literature relevant to the topic of the thesis and is used to support the investigation of the main research question: Do crises lead to reforms and does democracy influence this relation?
Most economists agree that a crisis requires immediate responses and that governments must undertake action as soon as possible in order to minimize losses (Congleton, 2009/2012; Rodrik, 1996). However, this does not mean that they all agree on whether in reality crises induce reforms. Congleton (2009) looks at the political and economic determinants of the 2008 financial crisis and the policy responses. He states that during a crisis, government responses are ‘consistent with public choice predictions’ and that in democracies these responses are driven by voter concerns (Congleton, 2009). Brooks and Kurtz (2007) analyse capital account and trade liberalization in Latin America between 1985 and 1999. Part of the analysis examines the relationship between crises and reforms and the role of political dynamics and accountability, and the distributive effects. They find that indeed crises matter, but that they do not always promote liberalization. Thus, not always will crises lead to reforms if the ‘political systems may simply fail to generate a new political-economic consensus, leaving policy vacillation or the status quo in place despite crisis’ (Brooks and Kurtz, 2007). This is supported by Pop-Eleches (2008), who acknowledges that ‘economic need’ caused by crises is no guarantee that reforms will be successfully imposed, and that governments ‘need to have the bureaucratic capacity necessary to cope with the technical challenges of the reform process’ (Pop-Eleches, 2008). The fact that the status quo may remain unaltered in some political systems, and that governments need a certain ‘bureaucratic capacity’ if they want to successfully counteract crises by implementing reforms, indicates how important it is to also look at the interaction between democracy and crises.
Crisis and reforms
A large part of the relevant literature evaluates policy responses to economic crises. Duval and Elmeskov (2005) show that crises are associated with labour market reforms, based on evaluating an aggregate indicator of reforms. Tompson (2009) however, finds that there is no correlation between economic crises and labour market reforms. By analysing inflation and the black market premium Drazen and Easterly (2001) conclude that crises instigate reforms. Abiad and mody (2005) used a new financial liberalization index covering a period of 24 years from 1973-1996, and find that different types of crises influence reforms differently. A balance-of-payment crisis triggered liberalization, but banking crises increase regulation in the financial sector. Brooks and Kurtz (2007) also find that different types of crises influence reforms in several areas of regulation differently, and that this also influences whether crises instigate or impede reforms. During an inflation crisis, liberalization of trade brings about distributional benefits to many by suppressing price increases. On the other hand debt crises may lead to additional regulation, as governments may want to limit the amount of capital flight during a financial panic (Brooks and Kurtz, 2007). All of these studies have used different indexes or indicators but these are restricted to certain sectors, countries or time periods.
order to limit the damage. Governments may want to increase regulations to protect the domestic markets, or on the other hand reduce regulations in order to stimulate economic growth. However, whether a government decides to increase or decrease protection depends on the ideology and political partisanship of the ruling government (Galasso, 2014). Therefore, I also look at the level of democracy and control for the partisan politics. Galasso (2014) shows that the effects of crises on overall reforms are ambiguous as the results show that financial markets become more regulated and product markets are liberalized.
Rodrik (1996) discusses the reforms of the 80’s and the 90’s that were imposed after the global debt crisis in 1982. Latin American countries imposed the strongest and most sustained reforms, with Chile as the country that had the most liberal reform strategy. Rodrik (1996) even states that‘one of the consequences of the global debt crisis that erupted in 1982 was a wave of market-oriented economic reforms.’ This leads to the earlier proposed hypothesis that crises lead to reform, or as Rodrik (1996) formulates it: ‘the natural supposition that crisis is the instigator of reform.’ Brooks and Kurtz (2007) argue that these types of statements are too general. Their hypothesis includes that it depends on the type of crisis and reforms whether crisis indeed cause reforms. Furthermore, the perception of the short term costs, and the consequences that these market-oriented economic reforms will have, play an important role in the implementation, and in turn these perceptions depend on the current political and economic situation in the country (Brooks and Kurtz, 2007).
oppose those reforms, while the benefits of the reforms might only arise after the current rulers’ term. Many politicians are afraid that they will be punished for being reformists but Buti et al. (2009) show that having a strict reform agenda can actually enhance re-elections. In fact, ‘reforms appear to be more likely to receive a positive reception by the electorate if the economy is in a cyclical upturn’ (Buti et al., 2009). If the financial markets are efficient and well-functioning, they allow for better capital flows and consumption smoothing, which advances the benefits of the reforms and allows policy makers to impose reforms at an accelerated pace during good times. However, during an economic downturn when current policies are no longer sufficient reforms become an issue. Governments may postpone reforms if the level of reserves is still sufficient. Only when the economic conditions worsen sufficiently, and costs of sticking to current policies are higher than the costs of reforms, governments will have an incentive to abandon those current policies. Countries often only decide to act when the crisis is severe enough so that the reforms can be justified. When this point is finally reached it is often too late. (Rodrik, 1996)
The relation between reforms and crises is made more complicated when foreign aid comes into the picture. Aid might be given to a country under certain reform conditions. However, when there are no reform requirements bound to the foreign assistance, it may make reforms less likely, as the effect of the crisis is softened. Thus, as Drazen and Easterly (2001) argue, an economic crisis might not impel reforms if it induces foreign financial aid.
Democracy and reforms
more credible (Bresser-Pereira et al., 1993).
If the projected reforms cause powerful interest groups to face high distributive costs, democratic political systems might not be a favourable environment for these types of reforms, as these interest groups may oppose the structural changes. Previous economic success creates such powerful interest groups according to Olson (1982). These existing groups, who are well organized and have vested interests, may be against any policy change that is potentially detrimental to them. As these groups are powerful they have the ability to block certain reforms that are beneficial to the general public. Only a substantial weakening of the power of these groups, caused by a sufficiently severe enough economic deterioration, can prevent them from blocking the reform. A crisis may thus be necessary to “reshuffle” powerful interest groups (Olson, 1982). Or as Williamson (1994) puts it: ‘a sufficiently acute crisis may also create a consensus that the old order has failed and needs to be replaced, leading individuals and groups to accept that their special interests need to be sacrifices on the altar of the general good.’
Many researchers have put forward theories about why reforms are delayed or blocked. An example of a delay in reform implementation was in the 1980’s when Argentina failed to stabilize in the presence of severe inflation due to the political division, and the lack of any interest group being able to exert its power effectively over the other interest groups (Dornbusch and De Pablo, 1989). Stabilization can only occur in a democracy when concessions between the groups are made, or when one group has become politically dominant. In an autocracy, where one person or a group with the same interests are in charge of decision-making, this problem does not occur and stabilization policies can be implemented swifter.
Alesina et al. (2006) use the previously described war of attrition model to contribute to the empirical analysis on the political economy of stabilization policies. They find that the war of attrition model is in line with the so-called “crisis hypothesis”, which states that it is easier to implement reforms in times of crisis than when there are only moderate economic difficulties. Furthermore, they investigate which political conditions have an influence on whether stabilization occurs during economic crises. It appears that when a country has a “strong” government that dominates the political opposition, stabilization reforms are more likely to be implemented. Stabilizations appear not only to be easier to implement but also have proven to be more successful the less institutional veto players exist, e.g. in presidential systems or systems where the ruling government has a majority. The planning of elections also plays a major role in the relation between crisis and stabilization. When new elections are nearby, a government may not want to impose stabilization policies as it can make them unpopular, which gives them an incentive to delay in order to win the upcoming elections. Directly after elections stabilization is more likely to occur, as the ‘new government enjoys a mandate’ (Alesina et al., 2006).
In the description of the war of attrition model above it is evident what might lead to delays in stabilization. However, it may be more interesting to look at what factors initiate these stabilizations? Alesina et al. (2006) identify 5 features that make stabilizations happen:
(2) Crises: An economic downturn may also induce earlier stabilization, as it increases the waiting costs mentioned above. This increases the incentive to concede, especially if the crisis affects one of the groups in particular.
(3) Nature of political institutions: Besides economic costs of delaying stabilization reforms, there may also be political costs. For example, the groups may want to block each other’s reforms at a cost.
(4) Political consolidations and elections: If one of the groups obtains more power it can make it very costly for the opponent to block its stabilization reforms and exercise its veto. If one of the groups loses the elections, it recognizes it is weaker than the other group and may concede.
(5) External factors: Several external factors, such as for example agreements with the IMF or foreign financial aid, may also influence the timing of stabilization. As explained above financial aid softens the blow from a crisis, which may induce a delay. However, if the aid is provided under reform requirements, stabilization reforms are initiated by foreign aid.
Gerring et al. (2011) takes on a case-study approach to evaluate the relationship between how long a country has been a democracy and economic performance. One of the issues central to this relationship is the ability of a democratic government to implement painful necessary reforms. In the implementation of reforms there might be an authoritarian advantage. A dictator, who is not elected, can individually impose policy changes, while a democratic leader must first persuade other interest groups. Especially when the policy changes impose high costs on some well-organized groups may a government face difficulties in instituting reforms. This relates to the war of attrition model of Alesina and Drazen (1991), and shows why it would be plausible that in a totalitarian society reforms can be more easily implemented. Nevertheless, in the case studies on Brazil, India, and Mauritius, it appeared that reforms are very well possible in a democracy, and that a democratic advantage arises as a regime matures.
priority to the general wellbeing but rather to their own personal goals (Giuliano et al., 2013). This problem does not arise in an autocracy where a benevolent dictator does not face these interest groups, can individually shelter the institutions, and can prevent these self-centred interest groups from blocking reforms that improve welfare of the society. Off course this does not hold if the dictator is non-benevolent and only cares about its own welfare, which can also prevent necessary reforms to be undertaken. This predatory behaviour disrupts economic activity and reforms will not be made or will not have the desired outcomes. Democratic rulers on the other hand care about the inhabitants of their country. They are more willing to implement reforms that are in the public interest and necessary to prevent further economic deterioration (Giuliano et al., 2013).
It appears that the best time to introduce reforms is directly after a new government is imposed. Immediately after the new government takes control they still enjoy a ‘honeymoon’, a period in which the public will be more lenient if the incoming government makes a mistake. During the honeymoon phase the incumbent government can blame any costs and difficulties on their predecessors (Williamson, 1994).
Hypotheses
From the literature above some hypotheses can be formed about the relationship between reforms and crises, and the moderating effect of the level of democracy. A crisis will make each interest groups more open to reforms, as the costs of the status quo increase. This will diminish the delay in the adoption of necessary reforms. Hence, the first hypothesis is that crises instigate reforms.
3
Data Description
The main dataset used in this paper was constructed by the Research Department of the IMF. The main goal of this dataset is to evaluate the degree of regulation in several sectors. This is a unique dataset as it covers many countries and several sectors. Most previously existing data sources considered merely a limited number of countries or sectors. The dataset contains information on 150 countries, both industrial and developing, and covers six sectors, both financial and real. For a list of the countries included in the sample see appendix A. An elaborate discussion on the regulation data is given below. Several other variables used in our analysis were also taken from this dataset but originally come from different data sources. For these variables the descriptions below will provide the original data source information. Furthermore, some additional variables are added, which will also be discussed.
Data on regulation
Economic reform is defined as a change in government policies. The data comes from a new dataset of the IMF, which is also used by Giuliano et al. (2013), and explained in detail in Ostry et al. (2009). The dataset covers both financial sector and real sector regulation indices for 150 countries over the period 1960-2005. The financial sector regulation indicators are domestic financial liberalization and capital account liberalization. The real sector indices cover four indicators, i.e. trade reforms covered by tariff rates, current account liberalization, product market reforms and agricultural markets regulations. The reform indicators of each sectors are aggregated into indices and then normalized, which gives values between zero and unity. Higher values of the indices represent lower levels of regulation. It must be noted here however that these indices cannot be compared across sectors; it merely provides information on economic reforms within sectors. Table A.2 in appendix B is a summary from a table in Ostry et al. (2009) and contains short descriptions on all the regulation indicators, and also provides all the data sources for these variables. Below I will discuss the most important aspects.
Financial sectors
regulations on domestic bond and equity markets. The banking sector index is the average of five separate indicators, i.e. interest rate controls, credit controls, competition restriction, the degree of state ownership, and the quality of banking supervision and regulation. The other financial sector indicator concerns regulations on the capital account, and includes information on lending and investment restrictions. Data on these indicators comes from Abiad et al. (2008).
Real sectors
Four real sectors are considered of which two product markets, i.e. the telecommunication and electricity market together, and separately the agricultural market. The first product market indicator, the indicator of the agricultural market, covers the degree of regulation on exports, the amount of public intervention, public ownership, and whether there are managed prices. The second product market indicator, for the electricity and telecommunication market, is quite similar and also looks at the level of regulation, the level of competition, privatization, and whether there is a self-governing regulatory authority in the market. The third indicator looks at regulations in the trade sector measured by tariff rates. Finally, current account regulations are included to see how much restrictions are imposed on trade in visible products and non-visible products such as services.
Data on other variables
Several types of crisis will be included as they may differently influence economic reforms. I will include banking crises, debt crises, currency (balance-of-payments) crises, and inflation crises. All the crises variables are dummy variables taking a value of 0 in a year that there is no crisis and 1 if there is a crisis in that year. The data on banking, debt, and currency crises comes from Reinhart and Rogoff (2009). The missing data on currency crises comes from Laeven and Valencia (2008). The dummy for an inflation crisis is constructed from the dataset of Giuliano et al. (2013) and is equal to 1 if inflation is larger than 40% in a given year and 0 otherwise.
variables the polity score is also a yearly measurement. The index is normalized so that 1 means a complete democracy and 0 indicates a totalitarian government.
Some political variables are also added as independent variables. The data comes from the World Bank’s Database of Political Institutions (DPI) of Beck et al. (2001), which was updated in January 2013. If a country is a democracy then it may matter whether the current government is right-winged or left-winged. The dummy variable left is added to the analysis, which takes on a value of 1 for a left-winged government and zero for a right-winged or ‘central’ government. The information for the left-variable also comes from the DPI, from the variable EXECRLC, which distinguishes between right, left or centre. Another political variable that may influence the choices made on whether to impose reforms or not is how close elections are. This is given by the variable YRCURNT, which gives the years left in current term and is also taken from the DPI. When elections are in the very near future this might provide the current government with election incentives and can cause delayed reform implementation. They might not want to impose necessary stabilization reforms that can make them unpopular right before the elections. To see whether governments that have been in office longer are stronger I also evaluate the number of years in office, given by YRSOFFICE (DPI). In order to test the honeymoon-hypothesis a dummy variable is added, which is equal to one when it is a chief executives first year in office and zero otherwise. This variable is called FIRSTYR. The final political variable that is used is PRESID, which is a dummy variable that takes on a value of one if the form of the government is directly presidential and zero if the system is parliamentary or the president is chosen by an assembly. This variable is added because presidential systems are believed to promote reforms (Persson and Tabellini, 2002).
Table A.3 in Appendix C provides some summary statistics on the reform indices of the separate sectors. Table 1 below shows the correlation matrix of the reforms measured in levels, where it can be seen that the degree of regulation between the different sectors correlates positively. Table 2 present a correlation matrix of the changes in the level of reforms, it gives the correlations between changes in reforms between the sectors. This is an indication of how reform changes in these different sectors influence each other, and again we see that most of the correlations are positive and quite substantial. However, the agriculture market shows only weak correlation, either positive or negative, to any of the other sectors. This implies that reforms in the agriculture market are not as much influenced by reforms in other sectors. Where for example capital account reforms and reforms in the domestic financial market may often be implemented simultaneously (correlation of 0.7), this is not the case for agriculture with any of the other sectors.
Table 1 - Correlation Matrix Reform-Levels Agriculture Market Product Market Trade Capital Account Current Account Domestic Financial Market Agriculture 1 Product 0.27 1 Trade 0.41 0.44 1 Capital 0.38 0.50 0.61 1 Current 0.48 0.49 0.64 0.79 1 Dom. Fin. 0.38 0.65 0.65 0.74 0.73 1
Figure 1, presented above shows how democracy (polity2) and reforms changed over time for the sample period. It shows the overall change in the aggregated reform index (right y-axis), which is normalized between zero and one, and the change in the level of democracy given by the change in the polity score (left y-axis). From the image it is clear that the two, reforms and democracy, move together. However, this does not give more insight into whether changes in the level of democracy cause economic reforms.
Figure 2 below shows the behaviour of both variables for all six sectors; i) the agriculture sector, ii) the product sector, iii) the trade sector, iv) the capital account, v) the current account, and vi) the finance sector. For almost all sectors it appears that the reform index and the level of democracy moved somewhat along the same pattern from 1960 to 2005. Only the trade sector shows during a decrease in the polity score that there was a simultaneous increase in the level of reforms, which is odd as there is plenty literature supporting the view that democracy causes globalization (Rodrik, 1996; Eichengreen and Leblang, 2006). Unfortunately, there is no clear explanation for this observation. However, from about 1970 onwards, the reform index and democracy do move together.
Furthermore, figures A.1, A.2, and A.3 (presented in appendix C) show the same relationship only for the low-, middle-, and high-income countries separately. These graphs confirm that the relationship between democracy and the reform index is present in all types of countries and not just unique or driven by a certain group (developing or developed) of countries.
4
Empirical strategy
This study uses a time-series cross-section (TSCS) analysis. In order to investigate whether democracy, the presence of a crisis, and a combination of the two affect economic reforms, I use panel data with a sample size of 20,123 “sector-country-years”. The panel data is unbalanced, with some countries having more observations than others. The model is constructed in such a way so that I can link the effects of all the explanatory variables on the change in economic reforms. The change in reforms is given by the difference in the index from one year to another in a certain country (c) and sector (s):
(1)
The basic equation that will be estimated can be written as follows:
(2)
The level of the reform index in a certain sector, country and year is given by . The lagged polity score for a certain country is given by , which is the measure for democracy. Furthermore, the lagged reform index is added. Four dummies for the types of crisis are included: Debt, Banking, Inflation, and Currency. θs, δc, and γt represent the sector,
country, and time fixed effects respectively.
Initially, as a simple starting point the equations will be estimated by OLS, of which the results are shown in appendix D. In order for the OLS estimators to be BLUE (Best Linear Unbiased Estimator) some conditions must be met. With autocorrelation the error term depends on its predecessor: . When there is autocorrelation and different error terms are correlated the following assumption is violated: { } Under this violation the OLS estimator is still unbiased but it becomes inefficient, just as with heteroskedasticity. The case where all error terms have the same variance is referred to as homoscedasticity: { } . Thus, in order to be certain that the OLS-regression is an appropriate method for the analysis it must be tested must whether it is correct to assume homoscedasticity and whether can rule out serial autocorrelation in the data. The results of the following tests are shown in appendix D. A Breusch-Pagan test, which tests the null of all error variances being equal, shows that the null of homoscedasticity is rejected. Thus, there is indeed heteroskedasticity present, as is often seen in cross-sectional models (Verbeek, 2012). To test for the presence of first-order autocorrelation a Wooldridge test is performed, which checks specifically for autocorrelation in a panel data set. The null-hypothesis for this test is no first-order autocorrelation. As the test statistic is significant it can be concluded that there is a serial correlation problem. With autocorrelation the error terms in our panel data are correlated. Therefore besides OLS, Prais-Winsten regressions are performed, which takes care of the serial autocorrelation problem.
Reforms may be persistent; therefore, I also add a lagged value of the dependent variable and create a dynamic model. Achen (2000) states that if there is autocorrelation in the error term then adding a lagged dependent variable in the regression will bias the estimates of the coefficients for the independent variables. On the other hand Wilkins (2013) says that including more lagged dependent and independent variables will lead to the least biased estimator. Beck and Katz (2011) supports this view: ‘there is nothing pernicious in the use of a model with a lagged dependent variable’. Furthermore, Beck and Katz (1995) suggest that adding the lagged dependent variable can deal with the serial correlation between the error terms. Others however, state that the inclusion of the lagged dependent variable in a fixed effects model can lead to biased coefficients when panels only cover small time periods, but as T increases the bias becomes smaller (Nickell, 1981, pp1418-1423). This bias suggested by Nickell (1981) is called the ‘dynamic panel bias’. In the statistical computations all models are estimated with and without the lagged dependent variable (results are only shown for the lagged dependent variable is included). Both methods give similar results in term of coefficients of the independent variables and significance. Therefore, apart from the lagged value of the change in reforms having significant meaningful effect, in this case it does not matter whether the LDV is included or not.
As there might be some delay in the effect of crises on reforms I estimate the same equation with lagged values. All other independent variables are also lagged one period in order to reduce the simultaneity bias, as suggested by Allan and Scruggs (2004) and applied in many similar researches (Galasso, 2014; Ostry and Spilimbergo, 2009).
(3)
According to the 2004 World Economic Outlook report of the IMF, countries learn from their neighbouring countries, and cooperation between countries can help them perceive the costs and benefits of planned structural reforms. Therefore, three variables are added that take into account the level of reforms in neighbouring countries. There are three types of neighbours evaluated, i.e. Geographic, Trade, and Political, which will be added as lagged variables as it takes time between observing reforms abroad and implementing them. Adding these variables gives equation 4:
(4)
The final regression equation includes interaction terms of the polity scores and the crisis variables to check whether the relationship between crises and reforms is different in an autocracy or a democracy. This gives us the final equation, equation 5:
(5)
determine which variables are considered as the modifying variables. To answer the research question I will evaluate the marginal effect of crisis on reforms with the level of democracy as the moderating variable. However, it might also be interesting to see what happens if we look at whether the presence of crisis affects democracies differently than autocracies in the implementation of reforms. Normally the marginal effect of any constitutive term is reflected by the coefficient. However, when the variable is interacted with the conditioning variable, the marginal effect must be calculated for each value of the conditioning variable (in the case of crises these values are zero and one) (Brambor et al., 2006).
Endogeneity
Besides the dynamic panel bias there is another problem that may bias the results. It is possible that not only does democracy influence reforms, but also that reforms have an effect on democracy. This means that the independent variable, democracy, is not exogenous. With an endogenous independent variable causality with the dependent variable (reforms) may run in two directions, meaning that the regressors may correlate with the error term. The
instrumental variable approach is one way to deal with this bias. However, as it is very difficult to find proper instruments, this method may also lead to biased results. Therefore, instead of IV 2SLS estimation, an Arellano – Bond dynamic panel GMM estimation is performed as a robustness test.
Causality between democracy and reforms may go in two directions, from democracy to reforms and vice versa. The regressors thus may be correlated with the error term. Another problem with the specifications is that including the lagged dependent variable gives rise to autocorrelation. Normally these problems can be solved by performing a 2SLS (two-stage least squares) estimation. Proper instruments for the endogenous variables are necessary; otherwise the instrumental variable estimation will be biased. Giuliano et al. face the same endogeneity problem and try to solve this by including the level of democracy in
neighbouring countries. They state that the proxy is used because the level of democracy in political allies does have an influence on the level of democracy, but not on the ability or willingness to implement reforms. However, the level of democracy in neighbouring
The difference and system generalized method-of-moments estimator is designed to deal with independent variables that are potentially endogenous (not strictly exogenous), thus variables that might be correlated with the error term. Furthermore, the estimator deals with heteroskedasticity and autocorrelation between individuals, and it deals with fixed effects (Roodman, 2009). This approach was designed to cope with the situation where there are no proper instruments available, as in this case. It is even assumed that there are no good instruments available, leaving as the only option the “internal” instrument (in this case based on lags of democracy). The instruments used in the computation are drawn from within the dataset; it uses lagged levels of the endogenous regressors, making the endogenous variables pre-determined and removing the correlation with the error term (Roodman, 2009). A big disadvantage of this method is that the execution is very complicated and it is likely that incorrect estimates are generated. In implementing this method, I carefully followed the instructions of Roodman (2009). However, the results of this robustness test must be carefully interpreted, as it is easy to make mistakes that alter the results significantly.
In this paper only democracy is instrumented in the robustness test. The robust option is added to render the standard errors robust to heteroskedasticity and autocorrelation.
5
Results
Initially, the idea was to use as the regression method Ordinary Least Squares estimation. However, with panel data the error terms often exhibit correlation. Therefore the equation is estimated with the Praise-Winsten estimator. After taking into account the country and year fixed effects, the regression also gave the results of an F-test checking whether all individual effects u_i are zero. Because the p-value of the test is 0.0000 (see table A.4 appendix D) we are able to reject this, which confirms the significance of including fixed effects, as the composite error terms (ui and ei) are correlated. It can thus be concluded that the fixed effect
more likely to change regulations. These results are robust to adding the lagged level of the reform index. The fact that this variable is negative, smaller than 1 and significant indicates that there is a certain level of regulation towards which countries converge. After adding country and time fixed effects this result becomes even stronger. The inflation crisis estimator is however no longer significant. The effect of the lagged banking crisis becomes significant. It is negative implying that governments are less likely to reform during a banking crisis.
Table 3 – Prais-Winsten Estimation of equation 2, which allows for AR(1) disturbances.
Dependent Variable: Reforms
(1) (2) (3) (4) Lagged Dependent Variable 0.009 0.009 0.025*** -0.204***
(0.008) (0.008) (0.008) (0.008) Debt Crisis -0.016*** (0.004) Bank Crisis 0.004 (0.004) Inflation Crisis 0.014*** (0.003) Currency Crisis 0.006*** (0.002) Polity2 0.010*** 0.010*** 0.019*** 0.010*** (0.002) (0.002) (0.002) (0.004) Lagged Debt Crisis -0.006 -0.007 -0.005
(0.004) (0.004) (0.004) Lagged Bank Crisis -0.007* -0.006 -0.011***
(0.004) (0.004) (0.004) Lagged Inflation Crisis 0.013*** 0.012*** 0.004
(0.003) (0.003) (0.003) Lagged Currency Crisis 0.004** 0.003 0.002
(0.002) (0.002) (0.002) Lagged Reform Index -0.030*** -0.087***
(0.002) (0.003) Constant 0.004** 0.004*** 0.012*** 0.048*** (0.002) (0.002) (0.002) (0.011) Observations 14777 15060 15060 15060 Country FE No No No Yes Year FE No No No Yes Sector FE No No No Yes
Standard errors are between parentheses. The significance levels are indicated by stars.
*** significant at 1% level, ** significant at 5% level, * significant at 10% level
for example the trade sector in the US indeed fluctuate from year to year. The overall level of reforms in the trade sector is quite constant around 0.9 (which represents a very low level of regulation), but in some years there is a small increase in regulations and in other years a small decrease. In all the specifications the lagged banking crisis variable is significantly negative as in the final specification above. In the last three specifications the inflation crisis also has a negative influence on reforms. This negative effect means that during a banking or inflation crisis a country increases regulation. A higher level of the reform index implies lower regulation, thus a decrease in the reform index implies stricter regulation and increased protection of the domestic markets. None of the added political variables appear to have a significant effect.
Table 4 –Add political control variable; allow for AR (1) disturbance. All specifications include sector, country and year fixed effects.
Dependent Variable: Reforms
(1) (2) (3) (4) (5) Lagged Dependent Variable -0.204*** -0.203*** -0.220*** -0.219*** -0.219***
(0.008) (0.009) (0.009) (0.009) (0.009) Polity2 0.012** 0.013** 0.021*** 0.022*** 0.021***
(0.005) (0.006) (0.007) (0.008) (0.008) Lagged Reform Index -0.109*** -0.128*** -0.143*** -0.142*** -0.142***
(0.004) (0.005) (0.006) (0.006) (0.006) Lagged Debt Crisis -0.004 -0.005 -0.005 -0.005 -0.005
(0.004) (0.005) (0.005) (0.005) (0.005) Lagged Bank Crisis -0.008** -0.008** -0.009** -0.009** -0.009**
(0.003) (0.004) (0.004) (0.004) (0.004) Lagged Inflation Crisis 0.004 0.003 -0.000 -0.000 -0.000
(0.004) (0.004) (0.005) (0.005) (0.005) Lagged Currency Crisis 0.002 0.002 0.000 0.000 -0.000
(0.002) (0.002) (0.002) (0.002) (0.002) Left 0.003 0.003 0.002 0.002 0.002 (0.002) (0.003) (0.003) (0.003) (0.003) Presidential 0.000 0.003 0.003 0.003 (0.006) (0.007) (0.007) (0.007) Years to Elections -0.000 -0.000 -0.000 (0.001) (0.001) (0.001) Years in Office 0.000 0.000 (0.000) (0.000) First Year 0.002 (0.002) Constant 0.057*** 0.107*** 0.109*** 0.108*** 0.109*** (0.012) (0.015) (0.017) (0.017) (0.017) Country FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Sector FE Yes Yes Yes Yes Yes Observations 14698 11856 10778 10674 10674
Standard errors are between parentheses. The significance levels are indicated by stars.
*** significant at 1% level, ** significant at 5% level, * significant at 10% level
well geographic as political neighbours significantly influence each other when it comes to implementing reforms. When a neighbouring country, that is either geographically close or politically similar, alters regulation in certain markets, the other country is also more likely to do so. This result confirms our initial thinking that countries may influence each other through peer pressure, but also that similar or close countries imitate each other. In all specifications the results on the lagged dependent variable, democracy, the lagged reform index, the inflation crisis, and the bank crisis give the same results as above.
Table 5 – Reforms and neighbouring variables. All control variables are lagged one period
Dependent Variable: Reforms
(1) (2) (3) (4)
Lagged Dependent Variable -0.213*** -0.218*** -0.217*** -0.212*** (0.009) (0.009) (0.009) (0.009) Polity2 0.021*** 0.021*** 0.021*** 0.021*** (0.006) (0.006) (0.006) (0.006) Lagged Reform Index -0.138*** -0.142*** -0.141*** -0.137***
(0.004) (0.004) (0.004) (0.004) Debt Crisis -0.004 -0.004 -0.004 -0.004 (0.005) (0.005) (0.005) (0.005) Bank Crisis -0.010*** -0.011*** -0.010*** -0.010** (0.004) (0.004) (0.004) (0.004) Inflation Crisis -0.002 -0.003 -0.003 -0.002 (0.004) (0.004) (0.004) (0.004) Currency Crisis -0.000 -0.000 -0.001 -0.001 (0.003) (0.003) (0.003) (0.003) Left 0.002 0.002 0.002 0.002 (0.003) (0.003) (0.003) (0.003) Presidential 0.003 0.003 0.002 0.002 (0.007) (0.007) (0.007) (0.007) Years to Elections -0.000 -0.000 -0.000 -0.000 (0.001) (0.001) (0.001) (0.001) Years in Office 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) First Year 0.002 0.002 0.002 0.002 (0.002) (0.002) (0.002) (0.002) Geographic Neighbour 0.143*** 0.142*** (0.028) (0.030) Trade Neighbour 0.035 0.012 (0.023) (0.026) Political Neighbour 0.017** 0.016** (0.008) (0.008) _cons 0.105*** 0.108*** 0.108*** 0.105*** (0.017) (0.017) (0.017) (0.017)
Country FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Sector FE Yes Yes Yes Yes
Observations 10642 10642 10642 10642
Standard errors are between parentheses. The significance levels are indicated by stars.
*** significant at 1% level, ** significant at 5% level, * significant at 10% level
influences the relation between crises and reforms. The marginal effects of all four types of a crisis should provide information on this matter. Unfortunately, not only are the interaction effects insignificant, but also the marginal effects are not significantly different from zero. The results that come from adding the interaction terms do not provide us with any information that can shed light on how democracy influences the relationship between crises and reforms.
To check whether the polity score becomes more or less relevant during a certain crisis the interaction terms between the polity score and the dummy variables of all the crises are added to the regression. This allows us to evaluate whether democracies and autocracies react differently to crises in terms of reforms. These results are shown in table 6. The coefficient of the polity2 score (β1) gives the effect of when there is no crisis (the value of the
crisis dummy is zero). The marginal effect of democracy on reforms is given by (for debt crisis, replace β15 by β16, β17, or β18 for bank, inflation, or
the interaction terms. Unfortunately, the marginal effect with all the interaction terms included is no longer significant at any of the three evaluated significance levels (1%, 5%, and 10%).
Table 6 – Interaction Terms.
Dependent Variable: Reforms
(1) ME (2) ME (3) ME
Lagged Dependent Variable -0.212*** -0.212*** -0.211***
(0.009) (0.009) (0.009)
Polity 2 0.022*** 0.007 0.021*** 0.020* 0.020*** 0.016* (0.006) (0.017) (0.006) (0.014) (0.007) (0.011) Lagged Reform Index -0.137*** -0.137*** -0.137***
(0.004) (0.004) (0.004) Debt Crisis 0.005 -0.003 -0.004 -0.004 (0.013) (0.014) (0.005) (0.005) Bank Crisis -0.010** -0.011 -0.010 -0.010** (0.004) (0.010) (0.011) (0.004) Inflation Crisis -0.002 -0.002 0.001 -0.001 (0.004) (0.004) (0.009) (0.011) Currency Crisis -0.001 -0.001 -0.000 (0.003) (0.003) (0.003) Left 0.002 0.002 0.002 (0.003) (0.003) (0.003) Presidential -0.000 -0.001 -0.000 (0.006) (0.006) (0.006) Years to Elections -0.000 -0.000 -0.000 (0.001) (0.001) (0.001) Years in Office 0.000 0.000 0.000 (0.000) (0.000) (0.000) First Year 0.002 0.002 0.002 (0.002) (0.002) (0.002) Geographic Neighbour 0.167*** 0.168*** 0.168*** (0.031) (0.031) (0.031) Trade Neighbour 0.013 0.012 0.012 (0.026) (0.026) (0.026) Political Neighbour 0.022*** 0.022*** 0.022*** (0.008) (0.008) (0.008) Polity2*Debt Crisis -0.013 (0.017) Polity2*Banking Crisis 0.001 (0.013) Polity2*Inflation Crisis -0.004 (0.012) Polity2*Currency Crisis Constant 0.104*** 0.105*** 0.103*** (0.014) (0.014) (0.014)
Country FE Yes Yes Yes
Year FE Yes Yes Yes
Sector FE Yes Yes Yes
Observations 10642 10642 10642
Standard errors are between parentheses. The significance levels are indicated by stars.
Table 6 continued
Dependent Variable: Reforms
(4) ME (5) ME
Lagged Dependent Variable -0.211*** -0.212***
(0.009) (0.009)
Polity2 0.023*** 0.014* 0.023*** 0.014 (0.007) (0.008) (0.007) (0.023) Lagged Reform Index -0.137*** -0.137***
(0.004) (0.004) Debt Crisis -0.004 0.004 -0.003 (0.005) (0.013) (0.014) Bank Crisis -0.010** -0.011 -0.010 (0.004) (0.010) (0.011) Inflation Crisis -0.002 -0.003 -0.002 (0.004) (0.010) (0.012) Currency Crisis 0.005 0.000 0.004 -0.000 (0.006) (0.007) (0.006) (0.007) Left 0.002 0.002 (0.003) (0.003) Presidential -0.012** -0.000 (0.006) (0.006) Years to Elections -0.000 -0.000 (0.001) (0.001) Years in Office 0.000 0.000 (0.000) (0.000) First Year 0.010*** 0.002 (0.002) (0.002) Geographic Neighbour 0.167*** 0.167*** (0.031) (0.031) Trade Neighbour 0.013 0.014 (0.026) (0.026) Political Neighbour 0.022*** 0.022*** (0.008) (0.008) Polity2*Debt Crisis -0.012 (0.017) Polity2*Banking Crisis 0.001 (0.013) Polity2*Inflation Crisis 0.001 (0.012) Polity2*Currency Crisis -0.007 -0.007 (0.008) (0.008) Constant 0.103*** 0.102*** (0.014) (0.014)
Country FE Yes Yes
Year FE Yes Yes
Sector FE Yes Yes
Observations 10642 10642
Standard errors are between parentheses. The significance levels are indicated by stars.
*** significant at 1% level, ** significant at 5% level, * significant at 10% level
Furthermore, in the fourth specification of table 6, where the interaction between the level of democracy and the currency crisis is added, the presidential and first year variables become significant, but the effects disappear in the 5th specification where all variables are added. The negative value for the presidential dummy implies that democracies with a presidential system are less likely or able to implement reforms than a democracy with a parliament. The positive effect of the first year dummy implies that there is indeed a honeymoon period in which the government can imply painful reforms, is allowed to make mistakes and more easily gains support for reforms, especially if the regime changed due to failures in current economic policies (Przeworski, 1991). Although I find a significant effect in the fourth specification, it is not possible to say that I have found evidence for the honeymoon hypothesis as the effect disappears in the other specifications.
Next, I look at different categorizations of countries. The difference between developed and developing countries is investigated using the definition of the World Bank to separate our sample into three groups. Countries that have a GNI of $1,035 or less belong to the low income group, middle income countries have an income between $1,036 and $4,085, and countries with a GNI of $12,616 or more are high income countries. The results are shown in table 7.
Table 7 – Difference developed and developing countries
Dependent Variable: Reforms
low middle high Lagged Dependent Variable -0.280*** -0.203*** -0.201***
(0.026) (0.013) (0.015) Polity2 0.028** -0.003 0.031* (0.013) (0.015) (0.014) Lagged Reform Index -0.132*** -0.170*** -0.102***
(0.011) (0.007) (0.006) Debt Crisis 0.032 0.002 -0.018 (0.027) (0.018) (0.016) Bank Crisis 0.003 -0.011 -0.025 (0.019) (0.016) (0.022) Inflation Crisis 0.014 -0.010 0.102 (0.024) (0.015) (0.063) Currency Crisis 0.012 -0.000 0.012 (0.012) (0.010) (0.018) Polity2*Debt Crisis -0.025 -0.008 . (0.053) (0.023) . Polity2*Bank Crisis -0.040 -0.002 0.021 (0.044) (0.020) (0.025) Polity2*Inflation Crisis -0.022 0.007 -0.155** (0.068) (0.018) (0.071) Polity2*Currency Crisis -0.015 -0.002 -0.017 (0.022) (0.013) (0.019) Left -0.008 0.004 0.002 (0.007) (0.005) (0.003) Presidential 0.012 -0.011 -0.002 (0.013) (0.010) (0.008) Years to Elections 0.001 -0.000 -0.001 (0.001) (0.001) (0.001) Years in Office -0.000 -0.000 0.000 (0.000) (0.000) (0.000) First Year 0.003 0.003 -0.001 (0.006) (0.004) (0.003) Geographic Neighbour 0.113* 0.145*** 0.107** (0.064) (0.047) (0.057) Trade Neighbour -0.048 -0.028 0.188*** (0.041) (0.043) (0.054) Political Neighbour 0.046** 0.022* 0.011 (0.023) (0.013) (0.011) Constant 0.017 0.140*** 0.045* (0.027) (0.025) (0.024) Country FE Yes Yes Yes
Year FE Yes Yes Yes
Sector FE Yes Yes Yes
Observations 1460 5184 3980
Standard errors are between parentheses. The significance levels are indicated by stars.
Brooks and Kurtz (2007) state that free market reforms should not be lumped together in an analysis. As I want to look at any changes in reforms I do lump them together in one index, but to see whether the results differ per sector, the model is also estimated for the 6 different sectors separately. These results are shown in table 9 below. The democracy variable does indeed have a different effect on regulations in different sectors. In the trade, capital and current account sectors democracy has a significant and positive influence, implying that liberalization reforms in these sectors are more likely to be made in democracies than in autocracies. Autocracies are often more closed in terms of trade and borrowing, which is confirmed by the higher level of regulation (less liberalizing refoms) that these countries have in these sectors. The reform index remains significant and negative for all sectors, meaning that I can correctly generalize the effect for reforms for all sectors combined, as was done before. The same holds for the lagged dependent variable. Separately for each sector, the bank and inflation crises no longer have a positive influence. For the current account the lagged debt crisis and currency crisis variables do have a significant albeit opposite effect. A debt crisis increases regulation concerning the current account, whereas a currency crisis induces the government to increase current account liberalization.
Table 8 - Separate regressions reforms with all control variables
Dependent Variable: Reforms
Agriculture Market Product Market Trade Capital Account Current Account Financial liberalization Lagged Dependent Variable -0.033 -0.263*** -0.143*** -0.152*** -0.092*** -0.170***
(0.024) (0.023) (0.024) (0.023) (0.026) (0.023) Polity2 0.001 -0.009 0.030** 0.067** 0.058*** 0.015 (0.018) (0.012) (0.014) (0.029) (0.019) (0.011) M.E. Polity 0.015 -0.013 0.052 0.031 0.083* -0.0025
(0.054) (0.056) (0.084) (0.067) (0.06) (0.01) Lagged Reform Index -0.131*** -0.105*** -0.303*** -0.279*** -0.156*** -0.190***
(0.013) (0.012) (0.015) (0.017) (0.015) (0.015) Debt Crisis -0.009 -0.014 0.033 0.068 -0.076** -0.006 (0.032) (0.023) (0.024) (0.049) (0.031) (0.017) Bank Crisis -0.009 -0.010 -0.029 -0.054 -0.014 0.013 (0.024) (0.017) (0.018) (0.037) (0.026) (0.013) Inflation Crisis -0.026 0.012 -0.018 -0.023 0.003 0.016 (0.024) (0.017) (0.019) (0.038) (0.026) (0.014) Currency Crisis -0.004 -0.006 -0.018 -0.001 0.045** 0.012 (0.016) (0.011) (0.012) (0.024) (0.017) (0.009) Polity2*Debt Crisis -0.008 0.012 -0.045 -0.081 0.086** 0.001 (0.043) (0.030) (0.055) (0.064) (0.040) (0.023) Polity2*Bank Crisis 0.002 0.007 0.037 0.043 -0.025 -0.015 (0.031) (0.022) (0.024) (0.048) (0.032) (0.017) Polity2*Inflation Crisis 0.028 -0.026 0.025 -0.024 0.009 -0.001 (0.030) (0.022) (0.030) (0.048) (0.032) (0.018) Polity2*Currency Crisis 0.005 -0.001 0.023 -0.010 -0.020 -0.020* (0.020) (0.014) (0.015) (0.030) (0.021) (0.011) Left -0.002 -0.002 0.001 0.015 0.002 0.003 (0.007) (0.004) (0.005) (0.010) (0.006) (0.004) Presidential 0.015 -0.011 0.027** -0.009 -0.011 -0.007 (0.014) (0.011) (0.011) (0.023) (0.013) (0.009) Years to Elections 0.002 0.001 0.000 -0.001 -0.002 -0.003*** (0.002) (0.001) (0.001) (0.002) (0.002) (0.001) Years in Office -0.001 -0.000 0.000 0.001 0.001 0.001* (0.001) (0.000) (0.000) (0.001) (0.001) (0.000) First Year 0.003 0.004 0.001 0.010 -0.008 -0.004 (0.006) (0.004) (0.004) (0.009) (0.005) (0.003) Geographic Neighbour -0.066 -0.186** 0.225*** 0.298*** -0.050 0.056 (0.066) (0.078) (0.086) (0.099) (0.081) (0.060) Trade Neighbour 0.029 0.071 -0.040 -0.067 -0.081 0.066 (0.063) (0.056) (0.070) (0.084) (0.072) (0.045) Political Neighbour 0.033 0.018 -0.006 0.019 -0.003 0.029** (0.023) (0.025) (0.014) (0.027) (0.018) (0.011) Constant 0.105*** 0.105*** 0.246*** 0.261*** 0.132*** 0.214*** (0.034) (0.026) (0.029) (0.057) (0.037) (0.027)
Country FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Sector FE No No No No No No
Observations 1444 2063 1990 1823 1499 1823
Standard errors are between parentheses. The significance levels are indicated by stars.
6
Robustness tests
To check whether the results in the previous section are consistent with several other specifications, two robustness tests are performed; The Arellano-Bond difference and system generalized method-of-moments (GMM) estimator, and the OLS with Panel-Corrected Standard Errors (PCSE). The results to the first robustness test are shown in table 9 below. As already mentioned in the empirical strategy chapter, these results must be interpreted carefully, as it is very easy to make mistakes with the estimation.
The lagged reform index remains negative and significant in all the specifications. This confirms earlier results, and along the same line of reasoning, I can say that this implies that there is a certain level of regulation towards which countries converge. The lagged dependent variable, however, no longer presents a significant result. The coefficients for the polity score are also no longer significant, which suggests that earlier results on the effect of the level of democracy on reforms were possibly driven by endogeneity and are thus are potentially invalid (the words potential and possibly are used as the results must be interpreted with care and strong conclusions cannot be drawn). The effect of bank crises seems to be confirmed by this result. A banking crisis does seem to have a significant influence on reforms, a finding that was also present in some earlier specifications. A banking crisis thus has a negative influence on the degree of regulation and leads to decreased liberalization, thus an increase in protection. Geographic neighbours have a positive influence on the level of reforms in a country. This result, which is only a confirmation of earlier results, provides evidence for the ‘learning’-hypothesis. Governments observe what goes right but also the mistakes made by others nearby and can learn from these mistakes, which positively influences the level of reforms. It confirms the reasoning that countries influence their neighbours through peer pressure, but also that similar or close countries imitate each other.
Table 9 - Arellano-Bond GMM estimation
Dependent Variable: Reforms
(1) (2) (3) (4) (5)
Lagged Dependent Variable 0.074 0.227 0.227* 0.218 0.233* (0.110) (0.143) (0.135) (0.136) (0.137) Lagged Reform Index -0.198*** -0.235*** -0.271*** -0.269*** -0.272***
(0.011) (0.013) (0.015) (0.015) (0.015) Polity2 -0.008 -0.012 -0.013 -0.004 -0.004 (0.010) (0.012) (0.016) (0.014) (0.017) Debt Crisis -0.001 -0.001 -0.001 0.003 (0.006) (0.007) (0.007) (0.013) Bank Crisis -0.009* -0.010** -0.009** -0.013 (0.005) (0.005) (0.005) (0.010) Inflation Crisis -0.001 -0.013* -0.012* -0.008 (0.005) (0.007) (0.006) (0.014) Currency Crisis -0.001 -0.003 -0.003 -0.006 (0.003) (0.003) (0.003) (0.008) Left 0.002 0.002 0.002 (0.003) (0.003) (0.003) Presidential -0.001 -0.001 -0.001 (0.008) (0.008) (0.008) Years Current -0.000 -0.000 -0.000 (0.001) (0.001) (0.001) Years in Office -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) First Year 0.004 0.003 0.003 (0.002) (0.002) (0.002) Geographic Neighbour 0.124*** 0.124*** (0.042) (0.042) Trade Neighbour -0.010 -0.010 (0.034) (0.034) Political Neighbour 0.011 0.011 (0.009) (0.009) Polity2*Debt Crisis -0.005 (0.018) Polity2*Bank Crisis 0.005 (0.015) Polity2*Inflation Crisis 0.001 (0.019) Polity2*Currency Crisis 0.004 (0.010) Constant 0.147*** 0.178*** 0.163*** 0.132*** 0.133*** (0.014) (0.018) (0.029) (0.022) (0.025)
Country FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Sector FE Yes Yes Yes Yes Yes
Observations 19559 14698 10674. 10642 10642
Standard errors are between parentheses. The significance levels are indicated by stars.
Table 10 - PSCE with autocorrelation term
Dependent Variable: Reforms pcar1 pcar2 pcar3 pcar4 pcar5 b/se b/se b/se b/se b/se Lagged Dependent Variable 0.024* 0.030** 0.037** 0.037** 0.037**
(0.012) (0.014) (0.016) (0.016) (0.016) Lagged Reform Index -0.106*** -0.110*** -0.136*** -0.132*** -0.132***
(0.005) (0.005) (0.007) (0.007) (0.007) Polity2 0.018*** 0.012** 0.021*** 0.020*** 0.022*** (0.004) (0.005) (0.007) (0.007) (0.008) Debt Crisis -0.005 -0.003 -0.003 0.002 (0.004) (0.005) (0.005) (0.012) Bank Crisis -0.010*** -0.010*** -0.009** -0.012 (0.004) (0.004) (0.004) (0.009) Inflation Crisis 0.003 -0.004 -0.003 0.001 (0.004) (0.005) (0.005) (0.010) Currency Crisis 0.002 -0.001 -0.001 0.001 (0.002) (0.002) (0.002) (0.006) Left 0.003 0.002 0.002 (0.003) (0.003) (0.003) Presidential -0.003 -0.004 -0.004 (0.006) (0.006) (0.006) Years Current -0.000 -0.000 -0.000 (0.001) (0.001) (0.001) Years in Office -0.000 -0.000 0.000 (0.000) (0.000) (0.000) First Year 0.002 0.002 0.001 (0.002) (0.002) (0.002) Geographic Neighbour 0.148*** 0.147*** (0.032) (0.032) Trade Neighbour 0.014 0.015 (0.026) (0.026) Political Neighbour 0.021*** 0.021*** (0.008) (0.008) Polity2*Debt Crisis -0.008 (0.016) Polity2*Bank Crisis 0.004 (0.011) Polity2*Inflation Crisis -0.006 (0.014) Polity2*Currency Crisis -0.003 (0.007) Constant 0.042*** 0.059*** 0.000 0.061*** 0.000 (0.007) (0.009) (.) (0.014) (.) Observations 20123 15060 10723 10642 10642
Standard errors are between parentheses. The significance levels are indicated by stars.
7
Conclusion
This paper investigates if crises and the level of democracy influence economic reforms, either caused by the ability or the motivation of a government to alter regulations. Many political economists are interested in why some reforms are delayed, or not implemented entirely, even if they are welfare enhancing. The postponement of reforms may partially be caused by a countries’ ability to implement them, which may depend on whether a country is a democracy or not. This matter has been the subject of many studies, but most of these had several drawbacks and were often limited to short time periods, a few countries, or focused only on liberalization in a particular sector. Besides including a broad measure of reforms I have attempted to look at other influences that may have an impact on regulations. Theories have suggested that a government faces a honeymoon period and can more easily impose reforms shorty after gaining power. Others, on the other hand, claim that a government that has been in office longer is better able to implement reforms. Unfortunately, for neither of these variables the results were significant and no conclusions can be formed on this matter. Some researchers have found that left-governments have different views on market protection and liberalization than right governments. However, our results also do not provide us with a clear answer whether and how government ideology influences reforms. Using our large, elaborate, and unique dataset covering more countries and sectors than all previous datasets, I do find that democracies are more likely to reform, all else being equal. This result proves to be robust to the inclusion of several controls, adding country and time fixed effects, and different estimation methods. Another interesting finding of this paper is that countries appear to converge towards a certain degree of regulation, as the change in the reform index depends significantly on its own lagged value. Furthermore, countries appear to be influenced by the level of reforms in neighbouring countries. Of the three types of neighbours that were investigated, the geographic neighbour appears to be robust in most specifications and I find clear evidence that there is a positive relationship between the degree of regulation in a country and that of their geographic neighbours, especially for the product market, trade sector, and the capital account.
may not be significant in all specifications the results point towards a negative influence of both banking and inflation crises on the adoption of reforms. Apparently, regulations are often increased in face of a crisis. However, as mentioned this result is not robust to all specifications and I am not able to give a clear-cut answer on the relationship between crises and reforms. To go a step further I also evaluate whether the role of a democracy in hindering or fostering reforms changes during a crisis, by adding the interactions terms between the level of democracy and the crisis dummies. Unfortunately, none of these interactions gave a significant result, and still leave these questions unanswered.
Some researchers have argued different crises can influence regulations in several sectors differently. A banking crisis is not expected to have the same effect on the degree of regulation in the agriculture market as in the financial market. Therefore, we also evaluated the influences on all sectors separately and found that for the current account both the debt and currency crisis have significant influences. The presence of a debt crisis appears to be negatively related to reforms, causing a decrease in reforms, meaning an increase in regulation and protection of the market. A currency crisis has the opposite influence, and causes governments to increase liberalization. Whether government reactions (either decreasing or increasing restrictions) improved economic conditions is not evaluated. This thesis does not concern whether it is better to increase or decrease regulation in face of a crisis, but merely looks at what governments did in the past decades.
When looking at financial liberalization I find that the number of years in office matter. This implies that stronger governments that have been in power for a long time are better able to implement reforms in the financial sector. Also the years left in the current term, thus time to the next elections, appear to matter. This provides evidence for elective incentives concerning the financial sector, which is quite logical as financial liberalization is often a very sensitive subject right up to elections and can truly harm a government’s reputation.
