Low inflation and interest rate in the European Union

1

Low inflation and interest rate in the European Union

David Baas

Student ID: 10783334

Supervisor: dr. Péter Foldvari

Submitted: 31 January 2017

Academic year: 2017-2018

2

Statement of Originality

This document is written by Student David Cornelis Christiaan Baas who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents

.

3

Table of content

3.2 Augmented Dickey-Fuller test……….………7

3.3 Johansen Cointegration test……….………7

3.4 Vector Autoregression……….7

3.5 Granger Causality………7

3.6 Varstable………8

3.7 Lagrange-multiplier test……….8

3.8 Cumulative Impulse Response Function………..8

4. Test results and analyses………..8

4.1 Varsoc……….9

4.2 Augmented Dickey-Fuller test………9

4.3 Johansen Cointegration test……….10

4.4 Vector Autoregression………..10

4.5 Granger Causality……….11

4.6 Varstable………12

4.7 Lagrange-multiplier test………..13

4.8 Cumulative Impulse Response Function………13

5. Conclusion………14

6. Discussion………15

7. References………..16

4

1. Introduction

The European Central bank (ECB) (2003) clarified that in the pursuit of price stability it aims to maintain inflation rates below, but close to, 2% over the medium term. The ECB has three reasons for this policy. Firstly, they want to keep a large margin to avoid the risks of deflation, because if the inflation rate becomes too close to zero and the interest rate is already low, it is hard to change the negative direction of inflation. It is not possible to cut the interest rate without limits. The second reason is about the HICP (Harmonised Index of Consumer Prices) inflation. The ECB takes into account the possibility of HICP inflation slightly overstating true inflation as a result of a small, but positive, bias in the measurement of price level changes using the HICP. The ECB (n.d.) denotes the term “harmonised” as the fact that all the countries in the European Union follow the same methodology. This ensures that the data for one country can be compared with the data for another. As a third reason, the ECB states that it avoids that individual countries in the euro area have to structurally live with too low inflation rates or even deflation.

Despite the ECB’s inflation policy, the inflation rate dropped far below the 2% during the last economic crisis. Moreover, the inflation rate in the Netherlands became negative at the beginning of 2015. Because of the low inflation rate in Europe, the ECB started an open market operations (OPO) and tried to raise the inflation rate. As a consequence of the ECB’s OPO, the interest rate dropped to a level of almost zero, whereas the deposit facility rate is negative since July 2014. Trading Economics (n.d.)

Because of the low interest rates, investors and individuals stopped saving by depositing their money in the bank and started investing in the stock market. This resulted in a strongly rising (very bullish) stock market and an increase in production. A negative consequence of the low interest rate is in the housing market. The housing prices are rising strongly in the big cities, what makes it hard for starters to buy their first house. Furthermore, the banks have no real incentive to take on the risks of lending. Instead, they put their money into safe assets such as Treasury bills, which yield almost as much as loans would.

Considering the negative effects of the low interest rate, this research will research the possibility of another policy which the ECB could maintain. Since there are large cultural and economic differences between countries in the euro area, this research will only focus on the Netherlands. Therefore, the research question of this article is: ‘How will the GDP growth in

the Netherlands change when the ECB accepts low inflation while at the same time raises the

long-term interest rate with one percentage point?’

The structure of this research will be as follows. In section 2, related literature about the effect of interest, inflation, export and import on the economy will be presented. Section 3 describes which tests will be used in this research. In section 4 the results and analyses of the tests will be shown. Finally, the results will be discussed in sections 5 and 6.

2. Literature review

According to Bradford Delong and Sims (1999) our ability to predict and control the inflation rate in the long-term is low. The available policy instruments are strong, but inaccurate. Moreover, other sets of circumstances than a general decline in goods-and-services prices alone -in particular, a sharp decline in asset prices- could set in motion the economic processes

5 that we fear from deflation. However, Akerlof, Dickens and Perry (1996) found that our labour market insures us against deflation. If the unemployment rate drops below the natural rate of unemployment, the inflation rate will rise. Otherwise, if the unemployment rate rises above the natural rate, the inflation rate will drop limitlessly. Thus, if the unemployment rate is at the same level as the natural unemployment rate, the level of inflation will remain constant.

Fisher (1993) concluded that a reasonably low rate of inflation and a small budget deficit is conducive to sustained economic growth. Barro (1995) found that an increase in the average inflation rate by ten percentage points per year is estimated to lower the growth rate of real GDP per capita by 0.2 – 0.3 percentage points per year.

Fisher (1930) suggested that expected interest rates change in proportion to the changing expected inflation, or expected real interest rates are invariant to the expected inflation rates. Mishkin (1992) explored that the Fisher effect (a high correlation between the interest rates and inflation) only happens in the long-run (and not in the short-run), but only if the interest rate and inflation rate have a common stochastic trend when they exhibit trends. Saymeh and Abu Orabi (2013) investigated the effect of interest rate, inflation rate and GDP on real economic growth in Jordan. They found out that GDP was directly affected by the interest rate and real growth rate was affected by inflation rate.

Tyler (1981) found that countries who neglect their export, tend to have a lower economic growth rate. Furthermore, the results of this research show that export performance was important, along with capital formation, in explaining the intercountry variance in GDP growth rates during the 1960–1977 period. The results of the research of

Saeed and Hussain (2015) show that there is unidirectional causality between exports and imports and between exports and economic growth. These results provide evidence that growth in Tunisia was propelled by a growth-led import strategy as well as export-led import. Import is thus seen as the source of economic growth in Tunisia. Yuhong et al. (2010) did time-series co-integration analyses with 28-year statistical data of east China from 1991 to 2009 of import, export and GDP growth. The results suggest that long-term or short-term causality exists between GDP and total export and import. Minella (2003) used a Vector Autoregression (VAR) to investigate monetary policy and basic macroeconomic relationships involving output, inflation rate, interest rate, and money in Brazil. He found that monetary policy shocks have significant effects on output but do not cause a decrease in the inflation rate in the first two periods.

Keynes’ created, in his book ‘General Theory of Employment, Interest and Money`(1936), an identity 𝑌 = 𝐶 + 𝐼 + 𝐺 + (𝑋 − 𝑀) to declare the change in the the dependent variable Y. The variable Y stands for total output/aggerate demand/total income.

C denotes the level of consumption which consists of an autonomous part and a part which

depends on the level of Y. So, when Y increases the total consumption increases as well. Because of the higher level of consumption, the number people needed for the aggregate production increases, which reduces unemployment and raises total income. This effect will lead to a higher level of consumption, and so on. This circle effect is called the multiplier effect. Both variables have effect on the other. Besides the level of income, the inflation rate also affects the rate of consumption.

Furthermore, I denotes the value of investments which is equal to the savings. Just like consumption, investment exists of an autonomous part and a part which depends on the level

6 of Y. This means that investment also has a multiplier effect. The rates of investment and saving are also affected by the interest rate. However, Gylfason (1981) declares that consumption depends not only on investment but also on interest rate and inflation rate. The empirical results reported in his paper indicate that the aggregate propensity to consume in the United States varies inversely with nominal interest rates and directly with the expected rate of inflation. Therefore, both variables will be used in the regression.

In Keynes’ identity the G denotes the government expenditure. Although government expenditure is not directly influenced by the ECB's policy, it is still important for the other variables in the identity. The letters X and M are for export and import. The import and export rates will be used as variables because of the high import and export rates in the Netherlands. In 2016 the Netherlands’ export to GDP rate was 82.4% and import to GDP rate was 71.4%, while the averages in the euro area are 44.05% and 39.97% (World Data Bank). This data shows that the Netherlands is an open economy, when compared to other EU countries. Just like the investment and consumption, there is also a multiplier effect between export/import and Y.

For each variable, the growth rate in percentage relative to the quarter before has been used. Therefore, instead of being covariance stationary, this time series appears to be “first-difference stationary”. This means that the level of a time series is not stationary, but its first difference is. Rahbek and Mosconi (1999)

3. Research method

In this thesis the Vector Autoregression (VAR) and Cumulative Impulse Response Functions (CIRF) will be used to answer the research question. Before doing VAR and CIRF, some assumptions have to be made and checked.

To answer this question, the databases of the CBS (Central Agency for Statistics of the Netherlands) and of Trading Economics will be explored. The CBS will be used because it is the official statistics agency of the Netherlands, Trading Economics is one of the largest website in statics. Because the database of Statistics Netherlands with these variables does not go any further than 1996, will in this research the quarterly data series from 1996 until the second quarter of 2017 of the export growth rate, the import growth rate, the consumption growth rate, long-term interest growth rate, government rate, investment rate and the HCIP (Harmonised Index of Consumer Prices) be used.

3.1 Varsoc

The Vector Autoregressive model cannot be done if the ideal length of the lags are not defined. To check how much lags should be used, the Stata command Varsoc will be used. In Stata, a lag-order selection statistic (pre-estimation) will be performed, which gives five criteria for the perfect time lag. The test gives five criteria: LR, FPE, AIC, HQIC and SBIC. The lag with the most asterisks (*) is the best lag to use for the Vector Autoregression. (Stata Manual)

The result will be used as the lag value for the Johansen Cointegration test and Lagrange-Multiplier test as well.1

7

3.2 Augmented Dickey-Fuller test

Before making use of the VAR model, all the variables have to be stationary, they have to be integrated of the same order. To check the status of all the variables, an augmented Dickey-Fuller test will be used. The augmented version of the Dickey-Dickey-Fuller test will be used, because it removes all the structural effects (autocorrelation) in the time series and then tests using the same procedure as the non-augmented version. The null hypothesis for the test is that the variable contains a unit root, the alternative hypothesis is that the variable was generated by a stationary process. (Stata Manual)

This test can be done in three different kind of styles:

1. With no intercept and no time trend: ∆𝑌t= 𝛿𝑌t-1+ ∑𝑝_𝑡=1𝛼𝑡∆𝑌t-1 + 𝜀1

2. With intercept and no time trend: ∆𝑌t= 𝛼 + 𝛿𝑌t-1+ ∑𝑝_𝑡=1𝛼𝑡∆𝑌t-1 + 𝜀1

3. With intercept and time trend: ∆𝑌t= 𝛼 + 𝑌t+ 𝛿𝑌t-1+ ∑𝑝_𝑡=1𝛼𝑡∆𝑌t-1 + 𝜀1

For this test, style 1 will be used. All the tests will be done with a lag of 0. However, to ensure the strength of the test, the variable interestlongterm will also be tested with a lag value of 1. Whether the test is valid or not can be seen in the sign of the coefficient. The coefficients except the trend and constants have to be negative. The tests will be done with an alpha of 5%.

3.3 Johansen Cointegration test

If not all variables are stationary, a Johansen Cointegration test must be performed. But since Johansen Cointegration test checks for both cointegration and a long-term equilibrium relationship, this test will be carried out anyway. In other words, this test checks whether the levels regressions are trustworthy (a situation called "cointegration".). If the asterisk (*) is placed at rank 0, the H0 hypothesis has to be accepted. If the asterisk is placed at rank 1 to 6, the H0 hypothesis has to be rejected. The rank where the asterisk is placed represents the number of cointegration relationships. The hypothesis will be checked with the trace statistics and the max statistics. The Johansen cointegration test will be done with an alpha of 5%.

3.4 Vector Autoregression

If all the criteria for the VAR have been met, the Vector Autoregression can be executed. In this research the VAR will be used because all the variables have effect on each other. A ‘normal’ regression is in this case not possible. The Vector Autoregressive model regresses all the variables on each other. Each variable has an equation explaining its evolution based on its own lagged values, the lagged values of the other model variables and an error term.

3.5 Granger Causality test

After the Vector Autoregression is done, the causality between the variables will be checked. Causality between variables is called the Granger Cause. Granger (1969) explains this concept as follows: if some series Y, contains information in past terms that helps in the prediction of X, and if this information is contained in no other series used in the predictor, then Y1 is said

to cause Xt. The Granger Causality test will determine whether one time series is useful in

8 term exists. The test will be done with an alpha of 5%.

3.6 Varstable

Except for causality, the VAR must also be checked for stability. The stability of the VAR will be checked with the Stata command Varstable. The Varstable checks the eigenvalue stability condition after estimating the parameters of a Vector Autoregression. If all the eigenvalues are below one, the Vector Autoregression is stable. This is the most important test for the validity of the VAR, if the results show instability the VAR outcome is not reliable and cannot be used.

3.7 Lagrange-Multiplier test

After doing the Vector Autoregression, the regression has to be checked against autocorrelation and normality in the errors. If autocorrelation exists, the variables are correlated with an error term. This error term represents a random “shock” to the model, or something that is missing from the model. However, the actual error term can never be seen. Therefore, we use the error term observations or residuals to check for autocorrelation. If they follow a pattern, this pattern is evidence of autocorrelation. To check for auto correlation a Lagrange multiplier test will be done. For the lag value, the same value will be used as in the VAR and Johansen Cointegration test. The test will be done with an alpha of 5%.

Furthermore, the residuals have to be normally distributed. But, no test is needed for this requirement because normality in the residuals can be assumed by the Central Limit Theorem.

3.8 Cumulative Impulse Response Function(CIRF)

After doing the Vector Autoregression, the results of the regression will be used to estimate the change in GDP. This will be done with a Cumulative Impulse Reponses Function. By doing CIRFs the effect of a one percentage point change in the long-term interest rate on consumption, export, import, government expenditure, investment and long-term interest rate itself will be measured. The CIRFs are statistically significant if the confidence interval (or band) does not contain zero (horizontal axis). But before calculating the change in GDP, the share of the variables in GDP in the period 1996-2017 has to be calculated. This is for consumption 48%, for export 69%, for import 60%, for government expenditure 26% and for investment 17%.2 _{These percentages will be used with the CIRFs to calculate the change in}

GDP.

To calculate the percentage change in the GDP the following formula will used:

𝛿 ln 𝑌 = 𝛼 𝐶 ∗

+ 𝛼 𝐼 ∗

+ 𝛼 𝐺 ∗

+ (𝛼 𝑋 ∗

− 𝛼 𝑀 ∗

)

4. Test results and analyses

In table 1 are the means and standard deviations.

9

Table 1 4.1 Varsoc

Table 2 shows that for two criteria, HQIC and SBIC, the ideal lag value has to be 1. The criteria LR and AIC say the ideal lag value has to be 7. They are identified with an asterisk (*). In this research a lag value of seven will be used for the Vector Autoregression, the Johansen Cointegration test and for the Lagrange-Multiplier test.

Table 2

4.2 Augmentend Dickey-Fuller test

The following hypotheses apply to the dickey fuller test: H0: interestlongterm got unit root(non-stationary)

H1: interestlongterm is stationary

Table A33 _{shows that the test statistic -5.414 is lower than the 5% critical value of -2.902.}

Therefore the H0 hypothesis has to be rejected and and this means that the variable interestlongterm is stationary. The same test, with the same hypotheses, will now be done with a lag value of 1. And also in this case the test statistic -5.451 is smaller than the 5% critical value of -2.903. This means the H0 hypothesis has to be rejected and the variable interestlongterm is stationary. For import, export, inflation, government expenditure and investment the H0 hypothesis have to be rejected as well.4 _{The test statistics are smaller than}

the 5% critical value. These variables are just like interestlongterm stationary. However, for the variable consumption the H0 hypothesis has to be accepted. The test statistic is larger than the critical value. This means consumption is a non-stationary variable.

3 _{See appendix}

10

4.3 Johansen Cointegration test

Using the results of the Lag-order selection statistic (pre-estimation) the Johansen Cointegration test will be done with a lag value of 75_.

The hypotheses are:

H0: no cointegration among variables H1: cointegration among variables

Table 3

In table 3, looking at the trace statistics, it can be seen that the asterisk (*) are at rank 3. Because the asterisk is not at rank 0 the H0 is rejected. Looking at the second part of the table the same conclusion can be drawn. The max statistic is higher than the 5% critical value for rank 0 and for rank 1. So for both tests the H0 hypothesis is rejected. This means, in the long run the variables move together and cointegration exists.

4.4 Vector Autoregression

Because the VAR criteria have been met, the Vector Autoregression can be executed. Table 4 shows that all the variables except government expenditure have a probabilty value lower than 5%. The means that these equations are significant. Government expenditure has a probabilty value higher than 5%, so this equation is not significant.

11

Table 4

4.5 Granger Causality test

With the results of the VAR in consideration a Granger Causality test will be done. The hypotheses for the test are:

H0: the joint effect on the variables is 0 H1: the joint effect on the variable is not 0

12

Table 5

For import, export, consumption, inflation, investment and long-term interest rate the H0 is rejected. The probability value is lower than 5%. But for government expenditure, the probability is 39.2%. This is higher than the critical value of 5%. So H0 is accepted. From these results it can be concluded that government expenditure is an exogenous variable and the other variables are endogenous variables.

4.6 Varstable

Table 6 shows that all the moduli have a lower value than 1. This means the VAR is stable but this also means that the CIRFs make sense.

13

4.7 Lagrange-Multiplier test

After doing the Vector Autoregression the regression has to be checked for autocorrelation. The hypotheses for the test are:

H0: no autocorrelation at lag order H1: autocorrelation at lag order

Table 7

For all the lags the probability is higher than 5%. This means that H0 is not rejected. We accept H0 and can conclude that there is no autocorrelation at lag order.

4.8 Cumulative Impulse Response Function (CIRF)

All the criteria have been met and the VAR is stable, so the CIRF may be executed. Table 8 shows the effect of 1 unit raise in interest on the other variables. For import, it can be seen that the first quarter, after a 1 unit raise in interest, rises. But then the import rate drops for two months, after which it rises again two quarters later. Thereafter it responses negative again. However, it is insignificant in all the periods. Export rate reacts positively in the first quarter, after which it reacts negatively for one month. The next six quarters are positive again. But just like the import growth it is in all the periods insignificant. Consumption and investment have an increasing positive response in all the quarters. But where consumption is insignificant in all periods there, investments are still significant in the first period and then become insignificant. Long-term interest rate has varying positive response in all the periods and it is significant in all the periods. Furthermore, government expenditure has an increasing negative response as a consequence of a permanent 1 percentage point rise in interest rate. But it is insignificant in all the periods.6

14

Table 8

The calculation for the first quarter is shown below, for the other quarters, the answers are provided in table 9. 𝛿 ln 𝑌 = 0.0022 ∗ 0.48 + 0.0177 ∗ 0.17 + −0.0071 ∗ 0.26 + (0.025 ∗ 0.69 − 0.0108 ∗ 0.60) = 0.0194 Quarter 1 0.019469 % Quarter 2 -0.0035 % Quarter 3 0.079779 % Quarter 4 0.052988 % Quarter 5 0.060543 % Quarter 6 0.049792 % Quarter 7 0.071263 % Quarter 8 0.068933 % Table 9

5. Conclusion

The aim for this test was to check what the effect of a raise in one percentage point in interest rate on the GDP is. The implemented variables are derived from the identity of Keynes. To check whether the VAR model may be used, some tests had to be done. First, the ideal lag value was investigated with the Stata demand Varsoc. The outcome showed that for the Vector Autoregression, Johansen Cointegration test and Lagrange-Multiplier test a lag value

15 of 7 had to be used. The Augmented Dickey Fuller test (ADF) was used to determine the stationarity of the variables. All the variables, except consumption, are found stationary at the first difference. However, because the variables are the first difference, all the variables are stationary on assumption. Furthermore, the Johansen Cointegration test confirmed there is cointegration between the variables at rank level 3. This indicates a long-term equilibrium relationship between all the variables exists. This means all the criteria for the VAR have been met and the VAR model has passed the diagnostic check.

The Granger Causality test confirmed there is short-run causality between all the variables, except government expenditure. Between the other variables is a unidirectional relationship. This means that if interest rate raises, it will influence all the other variables. But at the same time the other variables will also influence the interest rate. Because government expenditure has no causality with the other variables, it is an exogenous variable. The Varstable shows all the moduli have an eigenvalue below 1. All the eigenvalues are inside the unit circle, so the VAR satisfies the stability condition. According to the Lagrange-Multiplier test there is no autocorrelation at lag order.

By doing the CIRFs, the effect of a one unit raise in interestlongterm on all the variables is measured. In the first quarter the GDP will raise with 0.01%, in the second quarter the GDP will drop with 0.003%. In the following 6 quarters GDP will grow with circa 0,05 – 0,08%. However, because the CIRFs of importgrowth, exportgrowth, consumptiongrowth, governmentexpendituregrowth and all most all the periods of investmentgrowth are insignificant, can be concluded that long-term interest rate has no impact via the other variables on GDP growth. So, GDP growth will not change due to 1 percentage point increase in long-term interest rate.

6. Discussion

In this research the effect of a raise of one percentage point in the interest rate on the GDP is investigated. Although the GDP has grown the last couple years, the negative effects of the low interest rate, as discussed in the motivation, could have been reduced by raising the interest rate with one percentage point. CBS (n.d.) However, it seems that interest rate has no such impact on consumption growth, export growth, import growth, investment growth and government growth that is affects GDP growth. The underlying idea behind this may be that the long-term interest rate has barely moved the last eight years, it has been fluctuating between the -0.3% and 1.5%. The economy has adjusted the low and barely moving interest rate but already expects a change in the interest rate. So, for the just mentioned variables, the ‘shock’ will not have a direct effect.

Moreover, between government expenditure and the other variables no causality exists. This means that government expenditure is an exogenous variable. The reason for the lack of causality is that the government expenditure depends not only on monetary policy but especially on politics. This explains also the insignificance of the government expenditure equation in the Vector Autoregression.

Although there is no significant evidence that interest rate influences GDP growth through the named variables. I still suggest a raise in the interest rate, but gradually. In that case, for example, the stock market has time to adjust and the shock will be less intense. The same argument applies to the housing market, the housing prices will fall but people will not directly get in trouble with their mortgage. Moreover, starters get an opportunity to buy their first house.

16

References:

Akerlof, G., Dickens, W., & Perry, G. (1996). Low inflaton or No Inflation: Should the Federal reserve pursue complete price stability? Taylor & Francis, Ltd. (39)5, 11 - 17

Barro, J. (1995). Inflation and economic growth. Annals of Economics and Finance. Society

for AEF. 14(1), 121-144

Bradford Delong, J. & Sims A. (1999). Should we fear deflation? Brookings Institution Press. (1999)1, 225-252

Dfuller — Augmented Dickey–Fuller unit-root test. (n.d.) Retrieved from

https://www.stata.com/manuals13/tsdfuller.pdf

Engel, R. and Granger, W. (1987) Co-Integration and Error Correction: Representation, Estimation, and Testing. Econometrica. (55)2, 251-276

Fisher, I. (1930). The theory of Interest. Macmillan.

Fisher, S. (1993). Macroeconomics factors in growth. Journal of monetary economics. (32)3, 485-512

Granger, C. (1969). Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica (37)3, 424-438

Gylfason, T. (1981). Interest Rates, Inflation, and the Aggregate Consumption Function. The

Review of Economics and Statistics. (63)2, 233-245

Keynes, I. (1936) General Theory of Employment, Interest and Money. Retrieved from http://books.google.com

Measuring inflation – the Harmonised Index of Consumer Prices (HICP). (n.d.) Retrieved from https://www.ecb.europa.eu/stats/macroeconomic_and_sectoral/hicp/html/index.en.html Minella, A. (2002). Monetary Policy and Inflation in Brazil (1975-2000): A VAR Estimation.

Brazilian journal of economics. (57)3,

Rahek, A. & Mosconi, R. (1999). Cointegration rank inference with stationary regressors in VAR models. The econometrics Journal. (2)1, 76-91

Saaed, A. & Hussain, M. (2015). Impact of Exports and Imports on Economic Growth:

Evidence from Tunisia. Journal of Emerging Trends in Economics and Management Sciences. (6)1, 13-21

17 Saymeh, A. & Orabi, M. (2013). The effect of interest rate, inflation rate, GDP, on real

economic growth rate in Jordan. Asian economic and financial review. (3)3, 341-534 The definition of price stability.(n.d.) Retrieved from

https://www.ecb.europa.eu/mopo/strategy/pricestab/html/index.en.html

Tyler, G. (1981). Growth and export expansion in developing countries: Some empirical evidence. Journal of development economics. (9)1, 121-130

Varsoc — Obtain lag-order selection statistics for VARs and VECMs. (n.d.) Retrieved from

https://www.stata.com/manuals13/tsvarsoc.pdf

Yuhong, L., Zhongwen C. & Changjian San (2010) Research on the Relationship between Foreign Trade and the GDP Growth of East China-Empirical Analysis Based on Causality.

Modern Economy. (1)1, 118-124

Vec intro — Introduction to vector error-correction models. (n.d.) Retrieved from

18

Appendix A:

19

Table A5 import

Table A6 export

Table A7 consumption

Table A8 inflation

20

Table A10 investment:

Graph A1 Graph A2: -.2 -.1 0 .1 .2 0 2 4 6 8

CIRF1, interestlongterm, importgrowth

95% CI cumulative irf step

Graphs by irfname, impulse variable, and response variable

-.1 0 .1 .2 .3 0 2 4 6 8

CIRF1, interestlongterm, exportgrowth

95% CI cumulative irf step

21 Graph A3 Graph A4 Graph A5 -.05 0 .05 .1 0 2 4 6 8

CIRF1, interestlongterm, consumptiongrowth

95% CI cumulative irf step

Graphs by irfname, impulse variable, and response variable

0 .05 .1 .15

0 2 4 6 8

CIRF1, interestlongterm, investmentgrowth

95% CI cumulative irf step

Graphs by irfname, impulse variable, and response variable

-.06 -.04 -.02 0 .02 0 2 4 6 8

CIRF1, interestlongterm, govermentexpgrowth

95% CI cumulative irf step

22 Graph A6 .5 1 1.5 2 2.5 0 2 4 6 8

CIRF1, interestlongterm, interestlongterm

95% CI cumulative irf step