Bachelor Thesis
Name: Junnan You
Student number: 10256393
Specialization: Economics and Finance
Field: Monetary Policy
Number of credits thesis: 12
Title of thesis:
The impact of central bank independence on unemployment in
developing countries from 2000 to 2010
Contents
1.INTRODUCTION...3
2. THEORETICAL FRAMEWORK & LITERATURE REVIE...5
3. MODELS AND METHODS...8
4. DATA...12 5. RESULTS PRESENTATION…………...15 6.DISCUSSION………...19 7. CONCLUSION ...22 8. REFERENCE LIST...24 9. DATA SOURCE…...25 10. APPENDIX 1...…….……….……….………26 11. APPENDIX 2...…….……….……….………27
1. Introduction
This paper aims to explore the impact of central bank independence on unemployment in developing countries during the period of 2000 and 2010. The central bank plays an important role in the economic life of a nation, including setting the interest rate, controlling money supply, regulating bank reserve, and more. The recent financial crisis pushed the central bank to the frontier of combating deflation and market liquidity issues with its conventional or unconventional tools. After 2008, quantitative easing and other expansionary monetary measures have been excessively used together with fiscal policies to save the economies from recession. The purchase of corporation and government debt raised the issue of moral hazard, leading to the concern about central bank independence during the crisis. According to Walsh (2005), central bank independence can be defined as “ the freedom of monetary policymakers from direct political or governmental influence in the conduct of policy”. The controversial bailout measure of the Federal Reserve during the 2008 crisis is viewed as the sign of losing autonomy. Its extended liquidity provision stands on enormous risk of moral hazard and its purchase of treasury bills may be pressured by political demand (Kacperczyk & Schnabl, 2009). Many other central banks also follow the routine of the Federal Reserve, making the doubt about central bank independence even stronger.
One reason that scholars pay much attention to central bank independent is its close relationship with inflation change. The Barro-Gordon model argues that the inflation bias arise due to the problem of time inconsistency. If the central bank expects the total domestic output to surpass its natural level, then it may not be able to commit a low inflation rate, leading to inflation bias. Inflation change is defined as the difference between actual inflation rate and expected inflation rate; so if the central bank fails to keep its commitment, inflation change will increase as well. In Barro-Gordon model, the central bank can curb inflation fluctuation by maintaining its independence because an independent central bank would not pursue excessive economic growth but focus on inflation stability. On the other hand, the influence from governments can alter the original intention of the central bank, impairing central bank independence. Therefore, there is a negative relationship between central bank independence and inflation change in the sense that the more independent a
central bank is, the less inflation change there will be (Barro & Gordon, 1983).
The influence of central bank independence goes far beyond inflation change. Inflation targeting was introduced as part of monetary policy framework to lower inflation rate and its fluctuation, but not restricted to one objective. Bernanke & Goodfriend (2005) asserted that the main objection to inflation targeting is the narrow concentration of Fed on inflation at the expense of output and employment. This objection indicates that inflation change is connected to employment, or unemployment if you view it from reverse aspect. A more clear interpretation is demonstrated by the Phillips curve: the change of money wage rates (the change in inflation) can be explained by the unemployment variable. When the change in inflation increases, the unemployment decreases, vice versa (Phillips, 1958). Intuitively, a booming economy can possibly bring more job opportunities, decreasing unemployment. Meanwhile, a booming economy demands more currency in circulation so that it can satisfy the needs of increased demand for commodities. Increasing money supply can result in higher inflation given a constant interest rate. Consequently, we can expect a negative relationship between inflation change and unemployment.
Based on the theories discussed above, it is reasonable to connect central bank independence and unemployment via inflation change, which functions as a bridge. It is indicative to say that central bank independence is positively related to unemployment, meaning an independent central can incur higher unemployment. Baccarco & Rei (2007) investigated the relationship between unemployment and central bank independence during the research on the institutional determinants of unemployment in OECD countries. However, the country spectrum of this paper is restricted. Because of the need to test the results to a larger range, also of the important role of developing countries in world economic recovery, it is worthwhile to do the research among developing countries as well after financial crisis. This leads to the research proposition of this thesis:
The impact of central bank independence on unemployment in developing countries from 2000 to 2010
This research proposition helps us mainly focuses on central bank independence as the selected determinant of unemployment and on developing countries as the proliferation of previous research.
The model applied in the paper of Baccarco & Rei (2007) will be used in this thesis. Nevertheless, due to the unavailability of some factors in developing countries and special situations, such as the 2008 financial crisis, in recent years, the model will be simplified and modified to investigate a
rough relationship between central bank independence and unemployment in developing countries. The standard for being a developing country is based on the categorization developed by the IMF in its world economic outlook (IMF, 2014). The discussion about the measurement for central bank independence stands on the paper of Dincer & Eichengreen (2013). In order to make the results as precise as possible under the condition of limited data, the test for relationship between central bank independence and unemployment starts with the panel test without central bank independence variable. Then central bank independence data is added to check if the model can explain unemployment better in developing countries from 2000 to 2010. Afterwards, the coefficient and significance of central bank independence variable will be checked. The country data for different determinants is obtained for World Bank Database (2014).
This paper is formed in this way. Theoretical framework and literature review are discussed after introduction. Models and methods explanation is presented as the third part. The fourth part of this paper is data interpretation, closely followed by results presentation. The sixth part is discussion of the data. Eventually, a conclusion is drawn based on the previous analysis.
2. Theoretical framework & literature review
Much of the literature for central bank independence sheds light on its influence on inflation change and the problem of inflation bias. The literature could be put together with the literature of Phillips curve to discuss the impact of central bank independence on unemployment. Moreover, some scholars discovered the connection between central bank independence and unemployment from empirical test for the determinants of unemployment in developed countries. Although different in the range of countries, many of the experiment procedures are still replicable in developing countries.
An early involvement of inflation change in central bank independence analysis occurred in the paper of Barro & Gordon (1983). The authors established a simple model to interpret the time inconsistency problem. The problem of time inconsistency occurs when the central bank fails to maintain long-term low inflation fluctuation but pursues a short-term economic growth with higher inflation fluctuation. If time inconsistency happens, the society will suffer from the loss caused by the central bank. The loss function for society is
𝐿 = !
! [𝑎 𝜋 − 𝜋
In equation (1), L stands for loss, a stands for the coefficient of the difference between real inflation and initial equilibrium inflation, π means real inflation, π* means initial equilibrium inflation, y means total output, k
is the coefficient of initial equilibrium total output, and y* means initial equilibrium. Coefficients a and k vary from country to country.
The ultimate goal of the central bank, according to this equation, is to minimize the loss function by targeting inflation rate. Therefore, high central bank independence can be measured by minimized loss. Lucas aggregate supply function helps to reveal that the difference between real output and natural output positively correlates with the difference between real inflation and expected inflation. If the real inflation is replace by other variables in Lucas aggregate supply function, we can reach:
𝐿 = !! [𝑎𝜋!+ ((1 − 𝑘)𝑦∗ + 𝑏(𝜋 − 𝜋!))!] (2)
In equilibrium (2), b stands for coefficient of the difference between real inflation and expected inflation (inflation of last period), and πe stands for
expected inflation. The meaning of other symbols is the same as explained in equation (1). After minimizing this lost function, the new function about inflation can be expressed as:
𝜋 = !!!! !(𝑏𝜋! + 𝑘 − 1 𝑦∗) (3)
Here, Barro & Gordon (1983) assumed that people foster their expected inflation according to the inflation in the preceding period. If the central bank sticks to its commitment of low inflation change, then the change of inflation will always be zero; if the central bank deviates from its commitment, the expected inflation will rise to the inflation of previous year continuously until it equals a new equilibrium inflation rate. Comparing these two situations, the former scenario has smaller loss function, which implies a more independent central bank. Thus, it is plausible to conclude that central bank independence has negative relationship with inflation change.
The empirical evidence found by Phillips (1958) manifested a negative relationship in inflation change and unemployment. By adding more controlled economic variables, such as the price of labor service, change of retail price, import price, and productivity, Phillips was able to find a non-linear negative relationship between change of money wage rates
and unemployment, with the exception of the years that are after a dramatic increase of import price. Afterwards, the neoclassical version of the Phillips curve was introduced with the connection of Okun's law. Unemployment moves reversely with the change in inflation given natural unemployment level, forming a downward curvature graph for unemployment and inflation change (Mankiw, 2009). Regarding inflation change as an intermediate variable, we can connect central bank independence and unemployment with a positive relationship.
Another paper about institutional determinant of unemployment in OECD countries (Baccaro & Rei, 2006) provides more insights about the impact of central bank independence on unemployment. The focus of this paper is on unemployment, so the choice of regressors is much broader than merely central bank independence. Besides, this paper pays major attention to the labor market, which has a significant influence on unemployment. Many perspectives of OECD labor market are discussed in this paper. High trade union density and collective bargaining coverage rates are two common characteristics of OECD labor market. These features protect workers from economic shocks on one hand, but push the payroll to a higher level one the other hand. From the view of time series, unemployment converges in different countries from 1970s for European countries, forming a homogenous movement. Univocal OECD labor market and transparent data lead to regression results with satisfactory significant level. From the established model and calculated results, it is clear to see that real interest rate, union density and central bank independence play significant role in five-year unemployment observation. Here, the authors discovered a positive relationship between unemployment and central bank independence with distinct statistical computation.
Different from preceding findings, Alesina & Summers (1993) concluded in their paper that the fluctuation and level of inflation could be reduced by central bank independence, but the macroeconomic performance, seems not be affected in terms of overall benefits and losses. The regressions for economic variables, including inflation, GNP growth, unemployment and real interest rate, indicate that the central bank has no impact on economic performance. The explanation for these results lies in the possibility that a country can still obtain benefits of sustainable low inflation without setting monetary discipline. This explanation stands on the ground that discipline based performance is more essential than discretionary performance. As a result, the hypothetical relationship between central bank independence and unemployment is also denied according to his paper. However, the conclusion in the article is not conclusive due to the
limitation on data, premises of assumptions and endogeneity of central bank independence. Further investigation still need to be done on this paper.
Grounded on the data of OECD countries, Cukierman & Lippi (1999) reached a unique conclusion regarding to central bank independence and labor market variables. Cukierman & Lippi (1999) addressed centralization of wage bargaining, unemployment, and inflation in most parts of the paper. They examined the effect of central bank independence, centralization of wage bargaining and their interaction on unemployment, inflation and real wage. Statistical analysis reveals the hump-shaped relationship between institutional variables and central bank independence will slowly disappear. Finally central bank independence increases as centralization of wage bargaining rises straightly. This relationship indicates that a highly independent central bank and decentralization of bargaining tend to decrease unemployment, inflation and real wage. Central bank independence and unemployment here have a negative relationship given other institutional factors. This paper also lists some possible channels through which central bank independence could influence unemployment. If inflation matters to trade union, the central bank can influence the perception of union thus affecting their decision on wage setting. Moreover, the central bank can affect trade unions through competition for real wage. Therefore, trade union, as an important determinant of unemployment, is deeply affected by central bank’s commitment on inflation. One shortcoming of this paper derives from the simplification of labor supply elasticity, which needs further research.
3. Models and Methods
The paper of Baccarco and Rei (2006) provides much inspiration for the formulation of methods and models. With the aim of analyzing the impact of central bank independence on unemployment precisely and plausibly, two simple models with economic variables can be built. Details of the models, including the choice of variables would be explained further in the following. Moreover, three statistical methods are applied to deal with statistical anomaly during the process of analyzing. The first one is binary variable analysis. Statistical summary reveals that some variables are highly volatile over time, causing statistical insignificance. Therefore, it is necessary to introduce dummy variables to stabilize volatility. The second method is panel data analysis. The data for 67 developing countries over eleven years are available, so panel data analysis is conducted according to the country
set and time series. The third method applied is the comparative method, which compares statistical results with and without a certain variable (central bank independence) to see if this variable contributes to higher R squared.
Before actually setting the variables for regression, Baccarco and Rei (2006) first discussed the relationship between unemployment and labor market institution. The trend graph shows that the fluctuation of unemployment among OECD countries looks quite different. From the authors’ point of view, these different fluctuations can possibly result from labor market rigidity. Here, labor market rigidity is regarded as a channel for labor market to affect unemployment, especially including trade union density and collective bargaining coverage. Channels play a pivotal intermediate role for the effects to take place. Therefore, Baccarco and Rei (2006) started to discuss the channels through which labor market can influence unemployment. According to previous literature and their own work, Baccarco and Rei (2006) came up with ten hypothetical channels in their model. These ten channels are: employment protection, union density, benefit replacement rate, tax wedge, wage bargaining condition, central bank independence, real interest rate, change in the inflation rate, change in terms of trade, and change in labor productivity.
For the research among developing countries, these ten channels can also be utilized as variables of the research, but need to be modified. Some channels are hard to be measured quantitatively, such as employment protection; hence these channels will not be included in the analysis for developing countries. Some other channels cannot be applied because the data is not available in developing countries. The data with easy access for developing countries is economic data. As a result, tax wedge, central bank independence, change in the inflation rate, change in terms of trade, and change in labor productivity are left for regressors. Two things need to be noted here. Firstly, real interest rate is a common economic figure but it is not chosen. The reason lies in the absence of interest rate data in many Islamic countries due to religious reason1. The data that needs to be imputed is so significant that the results may be biased. Secondly, the channel of labor productivity is replaced by GDP growth for the sake of measurement and accuracy.
Although the basic variables are set with the help of Baccarco and Rei (2006), some abnormal figures occur in inflation change, especially for the
1 Islamic countries do not set interest rate due to religious reason, so real interest rate is not obtainable for Islamic countries.
year of 2000, 2008,2009 and 2010. Inflation change of these years is observably higher or lower than other years, leading to possible statistical bias and error. Referring to the economic condition during these years, it is not hard to find that financial crisis was prevalent by then. From 1997 to 2000, Internet bubble stroke the market, leading to enormous damage to the economy; the 2008 financial crisis and the quantities easing measure afterward led to large injection of liquidity in the market (Kacperczyk & Schnabl, 2009). These two crises fundamentally affected the inflation change of different countries. Thus, adding a binary variable of financial crisis can make the results more accurate and neutral. If a year falls within the set of 2000, 2008, 2009 and 2010, it is recorded as 1, otherwise 0.
The second method applied is panel data analysis. The row data from Internet database is quite miscellaneous. Cross-sectional data together with fixed time serial is obtained from database. With the purpose of giving insights to the multidimensional data, panel data analysis with fixed effects model is chosen to investigate the relationship. Fixed effects regression is an approach to control for omitted variables in panel data when omitted variables change from entity to entity but remain constant over time. The fixed effects regression model contains different intercept for every country. A serial of indicator variables are used to express the intercepts. As a result, the effect of each omitted variable that derives from the difference between one entity to the next is absorbed by indicator variable (Stock & Watson, 2010, pp354-369). In addition, the results of Wu–Hausman Test for panel data also imply that fixed effects model is a proper choice. In Wu–Hausman Test, random effects model is more consistent and efficient under the null hypothesis; fixed effects model is more consistent and efficient under the alternative hypothesis. The results in Appendix 1 demonstrate that the probability for a real value to have greater Chi squared is 0, rejecting the hull hypothesis.2 With regards to the research sample, the researched countries are highly diversified within the sample. Every developing country has its unique economic and culture situation and this situation tends to stay constant over time. Accordingly, it is more reasonable to consider country specific effects as fixed effects so that they can be isolate from other economic variables. Panel data analysis with fixed effects regression can not only reveals a complete regression that is overtime and cross-country, but also isolates country specific effects from other variables. Consequently, panel data method with fixed effects model is chosen in this paper.
2 Role of thumb: if Chi squared is smaller than 5%, the null hypothesis is rejected, otherwise not.
The third method used in this paper is comparative method. The first step needed is to run the regression without central bank independent. The R squared value, which denotes the total sum of explained variance, can be found. Afterwards, the model with central bank independence variable will be tested to obtain a new R squared. By comparing the R squared before and after, the effect of adding additional variable can be explicitly observed.
From the channels and methods discussed above, the original model of this paper can shaped like this:
𝑈𝑁𝐸!,! = 𝛽!+ 𝛽! ∙ 𝐺𝐷𝑃𝐺𝑅!,! + 𝛽!∙ 𝑇𝑅!,! + 𝛽! ∙ 𝑈𝑁𝐸!,!!! + 𝛽!∙ Δ𝐼𝑁𝐹!,! + 𝛽!Δ𝑇𝑂𝑇!,! + 𝛽!∙ 𝐹𝐼𝑁𝐶𝑅𝐼𝑆!,! ∙ Δ𝐼𝑁𝐹!,! + 𝛼! + 𝜀!,!
The meaning of the variables from left to right is unemployment rate (UNE), GDP growth rate (GDPGR), tax rate (TR), lagged unemployment rate (UNE-1), change in inflation rate (ΔINF), change in terms of trade
(ΔTOT) and financial crisis dummy (FINCRIS) respectively. Besides, α represents entity specific intercept, ε represents error terms, i represents country and t represents time in this formula. All the changes in this paper refer to absolute changes.
If we add central bank independence (CBI), the comparative model therefore is:
𝑈𝑁𝐸!,! = 𝛽!+ 𝛽!∙ 𝐶𝐵𝐼!,!+ 𝛽!∙ 𝐺𝐷𝑃𝐺𝑅!,!+ 𝛽!∙ 𝑇𝑅!,! + 𝛽!∙ 𝑈𝑁𝐸!,!!!+ 𝛽! ∙ Δ𝐼𝑁𝐹!,!+ 𝛽!Δ𝑇𝑂𝑇!,! + 𝛽!∙ 𝐹𝐼𝑁𝐶𝑅𝐼𝑆!,!∙ Δ𝐼𝑁𝐹!,!+ 𝛼!+ 𝜀!,! It is important to demonstrate that the problem of perfect multicollinearity will not happen even though the change of inflation rate is included as a variable. Although the change of inflation rate is used as intermediate variable to connect unemployment rate and central bank independence, the relationship is not exact linear. Especially in the relationship between inflation change and central bank independence, country specific figure k, which is mentioned in equation (1), also plays a crucial role. Actually, including the change of inflation rate can help to detect how important inflation change is in the relationship between central bank independence and unemployment rate.
These two models are pivotal for the analysis and will be mentioned again in the part of results presentation. More characteristics and source of the data are elaborated in the following part.
4. Data
To unitize the model mentioned above, data set for every independent and dependent variables is necessary. One of difficulties in data collection lies in measuring central bank independence. The measurement can be subjective and partial. The criterion to measure central bank independence still has not been settled yet. Along the way of measuring central bank independence, disparity starts to emerge. The most commonly used methods will be discussed in the following part of this paper together with the explanation of data summary for other variables.
In order to turn qualitative judgment on central bank independence into quantitative data, Cukierman, Webb and Neyapti (1992) developed an index measurement for central bank independence based on four major categories with different weights. These four categories are chief executive officer, policy formulation, objectives and limitations on lending to the government. To be specific, the authors applied 16 criteria under the four categories, coding from 0 to 1 respectively. The higher the numeral coding, the more independent the central bank is. During the research, qualitative research methods, such as questionnaire, were used to transform subjective perception into weighted data. The time span of this research is about 40 years, which provides ample observations to calculate central bank independence index. Due to the political and economic issues in some countries, especially in developing countries, the data for some years are missing, thus the time span varies marginally from country to country.
This measurement is extended by Dincer & Eichengreen (2011) to capture the new changes in central bank independence. Instead of focusing on the time period between 1950 and 1989, Dincer & Eichengreen (2011) studied recent 13 years, from 1998 to 2000. The scope of country is also enlarged to 120 countries or regions. It is worth noting that the index used in this paper are from the work of Dincer &Eichengreen (2011), and the index is unweighted central bank independence index for developing countries from 2000 to 2010. This index is used later in regression as data source for central bank independence. The innovative work in this paper is its endeavor in searching for the determinants of central bank independence despite the criteria used before. The authors ran regression on past inflation, openness, financial depth, GDP per capita, IMF lending, legal origin and so on. The results suggested that financial depth, IMF lending and legal origin are three most significant factors. This test, although did not include unemployment
as a factor, can help further tests which are relevant to central bank independence avoid the problem of inconsistent ordinary least square.
After checking the availability of all the data, 67 developing countries are left for analyzing. The definition of developing countries is based on world economic outlook (IMF, 2014). The list of developing countries is in Appendix 2.
Data for other variables is procured from World Bank Database (2014). Excluding unambiguous variables such as unemployment, lagged unemployment rate and financial crisis dummy, other economic variables have to be clearly defined to avoid confusion. First of all, GDP growth data mentioned in this paper is measured in constant local currency. The exchange rate risk is avoided in this measurement. Moreover, the tax rate data used is total tax rate of commercial profit. This tax rate is specifically towards business rather than consumption. Besides, the inflation calculated by World Bank is based on consumer price index (CPI), which is different from GDP deflator measurement. Finally, terms of trade is short for net barter terms of trade index. The way to compute terms of trade is dividing export value unit value index by import value index relative to 2000 benchmark. The change in terms of trade represents the shocks in import price and export price, which can possibly affect unemployment.
One thing to stress is the usage of the change in inflation rate instead of inflation rate. The paper of Baccaro & Rei (2006) argues that inflation change can represent possible trade off between unemployment and inflation in Phillips curve. The change in inflation helps to show how far each economy is away from its equilibrium unemployment in explaining cross-country patterns of unemployment (Glyn etc, 2003). A major doubt for the usage of inflation change comes from country selection. Phillips curve derives from developed countries. In developed country, a low inflation change means a low inflation in general, but this relationship may not hold in developing. Therefore, it is necessary to regress the relationship between average inflation rate and standard deviation of inflation (inflation change) in different researched countries. The regression shows that one percent increase in average inflation will lead to 1.225 percent increase in standard deviation. The constant is -4.05 and significant at 1% level. Based on this result, it is reasonable to say that average inflation rate and standard deviation of inflation (inflation change) are positively related in developing countries as well. Hence, features of Phillips curve can also be applied to the researched developing countries.
The summery of data from World Bank is illustrated in Table 1. In total, seven variables are summarized.
Table 1
Variable Mean Std.Dev. Min Max Obs
UNE Overall 10.4357 7.4283 1 38.7 N=737 Between 7.1791 1.5636 34.0636 n=67 Within 2.0835 2.1993 24.9630 T=11 GDPGR Overall 5.0134 5.1720 -33.1008 54.1578 N=737 Between 2.4263 0.6380 14.9079 n=67 Within 4.5764 -32.8979 54.3607 T=11 TR Overall 45.7876 35.0727 8.3 272 N=737 Between 33.7749 9.3 258.6182 n=67 Within 10.2398 -34.6306 112.7694 T=11 UNE-1 Overall 10.4994 7.4523 1.2 38.7 N=737 Between 7.2096 1.7273 34.1 n=67 Within 2.0653 2.1358 25.2358 T=11 ΔINF Overall -1.2771 12.1512 -172.4358 76.8010 N=737 Between 4.4038 -25.9948 1.2967 n=67 Within 11.3368 -152.4652 96.7716 T=11 ΔTOT Overall 2.7916 14.3767 -104.2218 76.3654 N=737 Between 5.0135 -3.3744 16.3508 n=67 Within 13.4868 -115.7902 64.797 T=11 CBI Overall 0.4627 0.19330 0.09 0.83 N=737 Between 0.18947 0.1 0.83 n=67 Within 0.04421 0.10543 0.67452 T=11
From the table showed above, the abnormally large extreme values for inflation change are quite conspicuous. These extreme values all happened during the years of financial crisis. Consequently, it is worthwhile to add a dummy variable.
It is also important to check Pearson correlation between variables so that none of two variables have correlation higher than 0.7 (rule of thumb). If the correlation between two variables is above 0.7, there will be the problem of multicollinearity. The problem of multicollinearity can possibly result in erratic change in dependent variable due to small change in correlated independent variable. The results of Pearson Correlation test are shown in table 2.
Table 2
UNE GDPGR TR UNE-1 ΔINF ΔTOT CBI GDPGR 1.000 TR 0.0735 1.000 UNE-1 -0.0682 -0.1526 1.000 ΔINF -0.0100 -0.0986 0.0320 1.000 ΔTOT 0.0211 -0.0326 -0.0313 0.0273 1.000 CBI 0.0586 0.1105 0.1388 -0.0077 -0.0193 1.000
According to table 2, none of two variables are highly correlated; hence the consideration for multicollinearity problem is not necessary. The models established in this paper are still valid although some regressors seem to correlate. More results of regression will be presented in part five.
5. Results Presentation
An overview of the regression results is demonstrated first. As stated in part three, the comparison of model with and without central bank independence is put into one table to see how significant the differences are. Then the regression results will be explained in detail. Next, the comparison of two models will be analyzed via the interpretation of coefficients, residuals and R squared. Finally, the test for Barro & Gordon model and Phillips Curve will be conducted to inspect the relationship between central bank independence and unemployment through inflation change.
After capturing the whole picture of variables, we can conduct the panel data analysis. The panel independent variable is country, coding from 1 to 67; and the time variable is year, coding from 2000 to 2010.The chosen model
type in is fixed effects model in panel regression analysis. The table for panel data analysis is shown as table 3:
Table 3 Model 1 Model 2 GDPGRi,t -0.0357 ** (0.0149) -.0300 ** (0.0151) TRi,t -0.0029 (0.0066) -0.0031 (0.0066) UNEi,t-1 0.5713 *** (0.0328) 0.5765 *** (0.0329) ΔINFi,t -0.0036(0.0081) -0.0042(0.0082) ΔTOTi,t 0.0048 (0.0050) 0.0046 (0.0050)
FINCRISi,t*ΔINFi,t 0.0060
(0.0119) 0.0067(0.0119) CBIi,t / / -3.1865 ** (1.5468) Constant 4.7344*** (0.4561) 6.1330 *** (0.8173) 𝛔 𝛂𝐢 3.0331 3.1473 𝛔 𝛆𝐢,𝐭 1.8165 1.8121 R2 : within 0.3142 0.3186 between 0.9979 0.9754 overall 0.9321 0.9121
Note: * means at 10% significant level; ** means at 5% significant level; *** means at 1% significant level.
From the table, it is clear that central bank independence is a statistically significant variable in model 2, which has a negative relationship with unemployment. The protuberant variables in both model 1 and model 2 are GDP growth rate, lagged unemployment rate, and constant. In the results of regression for both model 1 and model 2, GDP growth, tax rate, inflation change, and the interaction term of financial crisis dummy have negative relationship with unemployment; while lagged unemployment, change in
terms of trade, and constant are positively related to unemployment. Theoretical explanations of these results are displayed in the sixth part of this paper.
𝛔 𝜶𝒊 in the table represents the standard deviation of residuals for
fixed effects (country specific effects) and 𝛔 𝛆𝐢,𝐭 means the standard deviation of residuals for overall terms. Clearly, standard deviation of residuals for fixed effects increases after adding central bank independence variable; while standard deviation of residuals for overall terms decreases after adding the new variable. This means that data for central bank independence does not represent idiosyncrasy of different countries but it helps to reduce error overall. Without considering country specific effects, the new model explains the data better by reducing the error; however, if country specific effects are taken into consideration, the overall explanatory power still needs to be checked by R squared overall.
For R squared, different measurements show distinct trends. R
squared within, which is obtained by running ordinary least squares on the
transformed data, increases after adding central bank independence variables.
R squared within does not take fixed effects into account; thus the results are
consistent with lower standard deviation of residuals for overall error terms.
R squared between is based on only fixed effects. The drop in R squared between indicates that central bank independence data cannot reveal
idiosyncrasy of different countries as well as other variables. Finally, R
squared overall combines both standard deviations of overall errors and
fixed effects residuals. R squared overall declines after adding central bank independence, indicating that the drop in fixed effects residuals overwhelms the increase in overall explanatory power (lower residuals for error terms means higher overall explanatory power). Most relevant R squared here is R
squared overall because it shows the overall trend of goodness to fit for
individual countries after adding central bank independence variable. As a result, the missing idiosyncrasy of different countries hampers the increase trend in R squared, showing a lower R squared overall after adding central bank independence.
Previous results show the direct relationship between central bank independence and unemployment, but it is also crucial to check the relationship through the change in inflation. Theoretical relationship derives from the combination of Barro & Gordon model and Phillips Curve. The following steps can be applied to examine the relationship between central bank independence and unemployment through inflation change. Firstly, check the impact of central bank independence on the change in inflation. Then examine the relationship between inflation change and central bank
dependence. The results are shown in table 4. In order to avoid the situation that inflation change is mainly explained by the interaction term of financial crisis and inflation change, the interaction term is omitted in the first step. Table 4
CBIi,t→ ΔINFi,t ΔINFi,t →UNEi,t Variables Dependent Variable: ΔINF
i,t
Dependent Variable: UNEi,t
GDPGRi,t 0.1176(0.0988) -0.0357 ** (0.0149) TRi,t -0.0574 (0.0429) -0.0029 (0.0066) UNEi,t-1 -0.0852 (0.2153) 0.5713 *** (0.0328) ΔINFi,t / / -0.0036(0.0081) ΔTOTi,t 0.0434 (0.0325) 0.0048 (0.0050)
FINCRISi,t*ΔINFi,t /
/ -0.0060(0.0119) CBIi,t -6.8548 (10.1360) / / Constant 4.7068 (5.3564) 4.7344 *** (0.4561) 𝛔 𝛂𝐢 4.7838 3.0331 𝛔 𝛆𝐢,𝐭 11.8825 1.8165 R2 : within 0.0074 0.3142 between 0.0177 0.9979 overall 0.0054 0.9321
Note: * means at 10% significant level; ** means at 5% significant level; *** means at 1% significant level.
Table 4 reveals that the change in inflation and unemployment has a negative and insignificant relationship; the relationship between central bank independence and inflation change is also negative and insignificant. This means the function of inflation change in connecting central bank independence and unemployment is still statistically uncertain. Accordingly,
theoretical inference based on Philips curve and Barro & Gordon model may not hold. This issue will be discusses further in the following part.
6. Discussion
The results of regression for both model 1 and model 2 are discussed in this part to check the impact of central bank independence on unemployment rate in developing countries. This part starts with the theoretical explanation of regression results, closely followed by the comparison with former research results. Then the coefficient of central bank independence variable is specifically discussed with connection to part two. Finally, three problems in the analysis procedures are listed.
Results from the last part fits the theoretical expectation in general. A booming economy will enjoy high GDP growth and more job opportunities, thus lowering unemployment rate. Moreover, increasing change in terms of trade means that a country has higher export price relative to import price. In this situation, the country will import more and export less, leading to less working opportunities. Therefore, the negative relationship between change in terms of trade and unemployment rate is reasonable. Furthermore, an economic boom also tends to result in higher inflation when more job opportunities are generated, thus having a negative relationship with unemployment. Finally, the unemployment rate is a sticky variable by its nature. Working contract can protect the rights of workers and make careers more stable. Consequently, it is not surprising to find positive relationship for unemployment rate and lagged unemployment rate.
The explanation for the tax rate is quite complex. Higher tax is definitely an extra cost for a company, but whether the tax burden has negative effect on employment or not depends on who actually shoulder the tax burden, namely the bargaining power of labors. If tax burden is transferred to workers, the labor demand in labor market will not be affected. Companies pay workers lower wage after tax to retain their profitability. Nevertheless, the supply of labor will be affected. If the supply of labor increases to compensate for the drop in wage, then unemployment rate and tax rate are negatively related. If the supply of labor declines due to strong bargaining power of workers, then unemployment rate will increase, indicating a positive relationship (Baccaro & Rei, 2007). Back to the regression results, the negative relationship shown in the results suggests the former explanation. The negative relationship also implies that bargaining
power of workers in developing countries is relatively weak so the supply of labor increases in the end.
Another issue in the regression is the insignificance of inflation change, tax rate and change in terms of trade. Insignificant inflation change implies that inflation channel may not an important channel to affect unemployment in developing countries. Other exogenous channels may dominate the relationship between central bank independence and unemployment, weakening the role of inflation. For the tax rate, the failure in including data for bargaining power of labors in the model might have led to insignificance. For change in terms of trade, its impact on unemployment may also be determined by other factors such as exchange rate and trade openness. In the simple model of this paper, many potentially influential variables mentioned above are not investigated, hence resulting in insignificance.
The results of regression also provide clues to answer the research topic of this paper. In the results, central bank independence has significant negative relationship with unemployment rate. Compared with preceding model, the R squared overall remains at a high level, but decreased a bit after adding central bank independence variable. The decrease of R squared
overall mainly derives from fixed effects. The regression results of this
paper differ from the research results found by Baccraro & Rei (2007). Their paper shows a positive relationship between unemployment rate and central bank independence for developed countries, despite similar models and methodology applied in both paper.
Actually, a distinct feature of Baccraro & Rei (2007) paper is taking real interest rate into consideration. Their paper argues that the influence of central bank independence is free of the influence of real interest rate. As a result, central bank independence relates to real interest rate in a weak way. Therefore, goodness to fit may improve if exogenous variable real interest rate is added. Additionally, independent monetary authority may cause a short-term increase in unemployment and then it probably turns into permanent increase due to hysteresis intervenes. Their paper suggests in the final that hysteresis intervenes seems to be more conspicuous explanation for OECD countries and the inclusion of real interest rate variable affects the coefficient of central bank independence. For the model developed in this paper, real interest rate is not included; therefore the influence of real interest rate variable on central bank independence cannot be checked.
However, the negative relationship between unemployment rate and central bank independence is still explainable. One of the most important aspects that needs to be interpreted is bargaining structure. If labor unions
are resistant to inflation, unemployment and centralization of wage bargaining tend to be more correlated as central bank independence decreases (Cukierman & Lippi, 1999). For developing countries with comparatively low central bank independence, monetary authority can hardly affect the decision of labor union on wage setting because of low credibility and accountability. Besides, the labor unions in many developing countries are not able to centralize, leading to competitive market for labor union. Due to the situation of the central bank and characteristic of market for labor unions, each labor union struggles for its best interest by requiring a higher nominal wage because they are afraid of low real wage caused by higher inflation. The request for higher wage soon decreases the demand of labor, leading to higher unemployment. Accordingly, central bank independence has negative impact on unemployment given decentralized labor unions in developing countries. In these complex relationships, labor unions require higher wage as a way to combat inflation and maintain real wage. It is an essential channel through which central bank independence can affect unemployment.
Previous explanation also uncovers three problems in the analysis process. The first one is exogenous variables. The relationship that derives from Barro & Gordon model and Phillips curve is not reliable if bargaining structure of labor union also plays a dominating role. The unreliability is partly explained by the statistically insignificant coefficients for relevant variables in Table 4. The effectiveness of inflation channel may be seriously weakened by bargaining structure of labor union channel. Moreover, other potentially protuberant factors, such as real interest rate, productivity change and tax rate, are not considered as variables in the models, thus incurring bias reasoning.
The second potential problem in this paper is endogeneity problem. The cause of endogeneity problem lies in simultaneous causality of independent variables and dependent variables. For instance, GDP growth is considered as a determinant of unemployment in model 1 and 2, but on the other hand, the change in unemployment can also affect GDP growth. Due to simultaneous causality, many regressors are correlated to the error term, leading to endogeneity problem. However, the magnitude of endogeneity problem is still uncertain. Maybe the negative relationship between central bank independence and unemployment rooted in significant endogeneity problem. Consequently, further investigations still need to be done on the problem of endogeneity.
Eventually, the change of R squared also implies the problem oin central bank independence index. A general high R squared overall indicates
that the models of this paper capture most effects of independent variables on dependent variables. However, the decrease in R squared overall results from missing idiosyncrasy of different countries in central bank independence index. Central independence index grades from 0 to 1, which has limited space to fluctuate. Moreover, many factors, such as turnover rate, tend to be constant during investigated years, leading to homogenous index. Consequently, fixed effects are not explained better after adding central bank independence variable and R squared overall decreases. A lower R squared
overall indicates that central bank index data also needs to be improved to
represent country fixed effects.
7. Conclusion
From the explanation and discussion in previous sections, the answer for the research topic has emerged. With limited variables in the model, the negative impact of central bank independence on unemployment is conspicuous in developing countries from 2000 to 2010. The regression results are significant at 5 percent level, indicating the reliability of the results. One plausible interpretation for these results lies in the bargaining structure of labor unions in developing countries. Low central bank independence may hamper the ability of monetary authority to influence wage-setting decisions of labor unions. In the situation of decentralized labor union in developing countries, each individual labor union will try to set higher nominal wage to avoid higher inflation. Although real wage may be protected, higher nominal wage will incur shrinking demand in labor market. Unemployment rate will then increase.
Three problems are also found during the analysis procedures. One major problem is exogenous variables. The main difference in methodology between this paper and preceding literature is the choice of variables. Factors, such as bargaining structure of labor union, real interest rate, exchange rate and union density are not used due to unavailability of some
economic and social data in developing countries. However, limited regressors may lead to the problem of biased results. The second potential problem is endogeneity. The change in dependent variables may affect independent variables as well, resulting in the correlation between regressors and error term. Protuberant endogeneity problem may change the coefficient of central bank independence into negative. The final problem exits in central bank independence index. The decrease of R squared overall in regression results denotes that central bank independence index data does not contain enough information about idiosyncrasy of different countries, which means insignificant fixed effects.
In order to improve the accuracy of this paper and provide more indications for developing countries, new models that include bargaining structure of labor unions should be established and relevant data should be collected as well. With the help of new model, the explanation for the negative impact of central bank independence on unemployment rate may be verified. Besides, real interest rate, exchange rate, union density, and more factors can also be incorporated in the model to eliminate potential omitted variable bias. Moreover, the endogeneity problem should be checked and solved by using more advanced econometric tools such as instrument variables. Last but not least, a new index about central bank independence can be introduced to make the models more explanatory. New central bank independence index should have larger range in coding and more idiosyncrasies of different countries. The final results may change as more factors and new data are taken into consideration. The most difficult part for further investigation tends to be collecting economic and social data for developing countries.
Reference List:
Alesina, A. Summers, L. H. (1993). Central Bank Independence and Macroeconomic Performance: Some Comparative Evidence. Journal
of Money, Credit and Banking, 25(2), pp. 151-162.
Baccarco, L. Rei, D. (2007). Institutional Determinants of Unemployment in OECD Countries: Does the Deregulatory View Hold Water?
International Organization, 61(3), pp. 527-569
Barro, R. J. Gordon, D. B. (1983). A Positive Theory of Monetary Policy in A Natural Rate Model. The Journal of Political Economy, 91(4), pp.
589-610
Bernanke, B. S. Goodfriend, M. (2005). The Inflation-Targeting Debating.
National Bureau of Economic Research, p. 312
Cukierman, A. Lippi, F. (1999). Central Bank Independence, Centralization of Wage Bargaining, Inflation and Unemployment: Theory and Some Evidence. European Economic Review 43, pp. 1395-1434
Cukierman, A. Webb, S. B. Neyapti, B. (1992). Measuring the Independence of Central Banks and Its Effect on Policy Outcomes. The World Bank
Economic Review, 6(3), pp. 353-398
Dincer, N. N. Eichengreen, B. (2013). Central Bank Transparency and Independence: Updates and New Measures. Bank of Korea Working
Paper No.2013-21, pp. 1-52
Glyn, A. Baker, D. Howell, D. Schmitt, J. (2003). Labor Market Institutions and Unemployment: A Critical Assessment of the Cross-Country Evidence. Department of Economics. Discussion Paper Series. ISSN
Kacperczyk, M. Schnabl, P. (2009). When Safe Proved Risky: Commercial Paper During the Financial Crisis of 2007-2009. NBER Working
Paper NO.15538, pp20-24
Mankiw, N. G. (2009). Macroeconomics (7th Edition). Worth publisher, pp.388-400
Phillips, A.W. (1958). The Relationship between Unemployment and the Rate of Change of Monetary Wage Rates in the United Kingdom.
Economica, New Series, 25(100), pp. 283-299
Stock, J. H. Watson, M. W. (2010). Introduction to Econometrics (3rd edition). Addison Wesley Longman, pp.354-369
Walsh, C. E. (2005). Central Bank Independence Prepared for the New Palgrave Dictionary, pp1
Data Source:
International Monetary Fund. (2014).World Economic Outlook. Pp. 160-163. Accessed at 24th, June 2014. Available at:
http://www.imf.org/external/pubs/ft/weo/2014/01/pdf/text.pdf
World Bank Database (2014). Accessed at 24th, June 2014. Available at:
http://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG http://data.worldbank.org/indicator/SL.UEM.TOTL.ZS
http://data.worldbank.org/indicator/FP.CPI.TOTL.ZG?display=map http://data.worldbank.org/indicator/TT.PRI.MRCH.XD.WD
Appendix 1: Wu-Hausman Test for fixed effects model and
random effects model in panel data
.
(V_b-V_B is not positive definite) Prob>chi2 = 0.0000
= 166.63
chi2(7) = (b-B)'[(V_b-V_B)^(-1)](b-B) Test: Ho: difference in coefficients not systematic
B = inconsistent under Ha, efficient under Ho; obtained from xtreg b = consistent under Ho and Ha; obtained from xtreg CBI -3.186504 .0460556 -3.232559 1.500073 Interaction .006708 .0108769 -.0041689 .0009953 ChangeTOT .0046347 .0033998 .0012349 . ChangeINF -.0041948 -.0048034 .0006086 .0021411 UNElag .5764633 .9585385 -.3820752 .03134 Taxrate -.0030598 -.000738 -.0023218 .0062065 GDPgrowth -.0299868 -.0566487 .0266618 .0057763 fixed random Difference S.E.
(b) (B) (b-B) sqrt(diag(V_b-V_B)) Coefficients
Appendix 2: Researched Developing Countries’ List
Country List 1
Country List 2
Albania Malaysia
Angola Maldives
Armenia Mauritius
Azerbaijan Mexico
Bahamas, The Moldova
Barbados Mongolia
Belarus Mozambique
Belize Namibia
Bhutan Nigeria
Bosnia and Herzegovina Oman
Botswana Papua New Guinea
Bulgaria Peru
Cambodia Philippines
Chile Poland
China Romania
Colombia Russian Federation
Croatia Saudi Arabia
El Salvador Sierra Leone
Fiji Solomon Islands
Georgia South Africa
Guyana Sri Lanka
India Sub-Saharan Africa Indonesia Syrian Arab Republic
Iraq Tanzania
Jamaica Thailand
Jordan Trinidad and Tobago
Kenya Tunisia
Kyrgyz Republic Turkey
Lao PDR Uganda
Latin America & Caribbean United Arab Emirates
Lesotho Venezuela, RB
Lithuania Yemen, Rep. Macedonia, FYR Zambia Malawi
