3. Monetary Policy and Global Commodity Prices
3.3. Data and Methodology
3.4.1. Unit Root Tests and Lag Length Selection
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
62
where the term (~=∑ €)) (•C))-: represents the generalized impulse response of 8*K=
with respect to a unit shock to the jth variable at time t. Note that ~= = A:~=-:+ ⋯ + AG~=-G, ‚ = 1,2, … , ~B = 3, ~= = 0 for ‚ < 0 and €) is the n×1 selection vector with unity as its jth element and zero elsewhere and covariance ∑ = •C). Similar to Soytas et. al. (2009) and Nazlioglu and Soytas (2011), the GIRF is used because, unlike the standard approach, this approach does not require orthogonalization of shocks. As the results of the GIRF approach are invariant to the ordering of the variables in the VAR model, it overcomes the orthogonalization problem of the traditional approach.
In addition to the above presented base model, another specification, taking the effect of US monetary policy variable into account, has been estimated. In addition to the seven variables of the base model, the alternative specification considered the US real interest rate as endogenous variable in the VAR model. As this specification did not change the outcome of the analysis on both long- and short-run dynamics between economic activity as well as real interest rate and global commodity prices, results are not reported here.
3.4. Empirical Analysis
The first section of the empirical analysis covers unit root and optimal lag length tests followed by an estimation of the lag augmented VAR models and diagnostic tests in section two. Section three and four presents the results on the long- and short-run dynamics of global commodity prices in relation to economic activity and monetary policy of China.
3.4.1. Unit Root Tests and Lag Length Selection
In the first step, the maximum order of integration LMNO of the variables is determined.
Thereby a variable is said to be integrated of order n when it achieves stationarity
‧
variables, the augmented Dickey and Fuller (1979) (ADF) test is applied. Table 3.10 summarizes the results. All of the unit root tests indicate at the 1% level of significance that LMNO is one.Table 3.10: ADF Unit Root Tests
Level Difference selection criteria based on SIC with maximum lag length of 13.
In the next step, the optimal lag length of the VAR model specified in (1) is
Table 3.11: Lag Selection Tests
AIC FPE HQ SIC
VAR lag order selection
criteria 1 1 1 1
Note: AIC is the Akaike information criterion, FPE is the Final Prediction Error, HQ is the Hannan-Quinn Information Criterion and SIC the Schwarz Information Criterion. 13 lags were included in the lag specification.
‧
3.4.2. VAR Estimation and Model RobustnessOn the basis of the optimal lag length, a VAR(1) model as specified in (1) is estimated. The 7×1 vector of our model jointly considers the prices of the agriculture, energy, industrial metal, livestock and precious metals group as well as economic activity and real interest rate of China, such that 8* = (VW_AY*, VW_Z[*, VW_3[*, VW_73*, VW_W4*, ZA*, 434*)′ . The model further considers a dummy variable capturing the effect of the managed floating- against the fixed exchange rate regime of China. The variable has a value of zero in the periods of fixed exchange rate system and a value of one otherwise. The estimated VAR(1) model satisfies the stability condition so that no roots lie outside the unit circle.
Next, the A4(d) model is augmented to A4(d + LMNO) as specified in (3). The unit root tests indicate that LMNO is one. Therefore the VAR(1) model is augmented to VAR(1+1). As specified in (3), the additional LMNO lags are defined as exogenous variable in the lag augmented VAR model. As the VAR system must satisfy the common regression assumptions to be valid, the residuals of each equation in the model are tested for autocorrelation and heteroskedasticity. Table 3.12 summarizes the results of the diagnostic tests for the augmented VAR model.
Table 3.12: Diagnostic Tests respectively. BG is the Breusch-Godfrey test with H0 of no serial correlation up to lag 2. BPG is the Breusch-Pagan-Godfrey test for H0 of homoskedasticity. WHITE is the White test for the null of homoskedasticity. ARCH is the Engle test for H0 of no autoregressive conditional heteroskedasticity up to lag 1.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
65
The test statistics of the Breusch-Godfrey test (BG) indicate problems with autocorrelation in the CP_EN, CP_PR and EA equation. The other diagnostic tests further indicate problems with heteroscedasticity in the CP_AG, CP_PR, EA and RIR equation. In order to yield valid test results in the subsequent hypotheses testing, the Newey-West Heteroskedasticity and Autocorrelation Consistent (HAC) corrected standard errors are applied to the computations of the respective equations.
3.4.3. Long-Run Dynamics
In the next step, the analysis applies Wald tests on the first lag coefficient matrix of the lag augmented VAR(1+1) model while ignoring the coefficients in the augmented lag matrix. From the Chi-squared statistic of the Wald test, it is possible to infer Granger causality. The results of this test, with the null of no Granger causality, are shown in Table 3.13.
Table 3.13: Granger Causality Tests
CP_AG CP_EN CP_IN CP_LI CP_PR EA RIR CP_AG - 0.425 0.799 1.417 2.107 0.000 0.252 CP_EN 0.256 - 1.424 0.085 0.826 7.437*** 1.720 CP_IN 0.865 1.585 - 0.000 0.001 4.000** 0.001 CP_LI 0.086 2.709* 0.224 - 4.439** 0.006 0.296 CP_PR 1.009 2.672 0.475 0.027 - 0.062 0.001
EA 3.360* 1.272 3.828* 0.021 1.232 - 0.181
RIR 3.600* 0.017 2.138 0.070 1.556 0.100 -
Note: Superscripts ***, ** and * represent significance levels at 1%, 5%, and 10%, respectively. Significance means that the column variable Granger causes the row variable.
The T&Y Granger causality tests indicate that energy (CP_EN) as well as industrial metals (CP_IN) prices are Granger caused by China’s economic activity (EA) at the 1%
and 5% level of significance respectively. These findings are in line with the findings of Roache (2012) who reports significant Granger causality running from the industrial production of China to global oil and lead prices over the period from 2000M01 to 2011M09. This finding provides significant evidence that the economic
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
66
activity of China is a determinant of global commodity prices, at least in the group of energy and industrial metals. At the 10% level of significance, the test statistics shows that agricultural (CP_AG) and industrial metals (CP_IN) prices are Granger causing economic activity (EA). This relationship might be due to commodity producers in these sectors who have benefited from increasing prices and thus stimulate economic activity.
As for the monetary policy variable of China, the analysis provides no empirical evidence that the real interest rate of China is Granger causing commodity prices. On the reverse side however, test results indicate that agricultural commodity prices (CP_AG) are Granger causing the real interest rate. This relationship is in line with Awokuse and Yang (2003) who report Granger causality running from commodity price to nominal interest rate and inflation. One explanation for this relationship is that fluctuations in commodity prices may provide signals to policy makers to adjust monetary policy variables such as the interest rate. Another explanation that seems to be more plausible in the case of China with a relatively flat interest rate is that changes in commodity prices affect inflation and thus real interest rates. This is because commodities are raw materials that are used as input factor, in the production of higher order goods. A change in the cost of the input factor has an impact on the higher order goods and thus on inflation and real interest rates.
The test results further indicate spillover effects from precious metals (CP_PR) and energy (CP_EN) to livestock prices (CP_LI). While the effect of precious metals is surprising, the impact of energy prices on livestock prices might be explained by an energy cost spillover that is incorporated in the price of livestock.
The estimated coefficients of the dummy variable included in the lag augmented VAR model are presented in Table 3.14. The coefficients of the dummy variable, capturing
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
67
the effect of the managed floating- against the fixed exchange rate regime, indicates that industrial metals (CP_IN) prices are significantly higher when the managed floating exchange rate system is in place. One possible explanation for this relationship might be that the value of the RMB versus the USD tends to increase during the time of a managed floating exchange rate system. As the USD is the predominant invoicing currency in international commodity transactions, an appreciation of the RMB relative to the USD implies that purchasing power of Chinese commodity consumers and thus commodity demand and prices are pushed up.
Table 3.14: Exchange Rate Regime Dummy Estimates
CP_AG CP_EN CP_IN CP_LI CP_PR EA RIR
Coefficient 0.026 0.018 0.045** 0.012 0.010 -0.017 0.022 Standard Error 0.016 0.025 0.018 0.013 0.015 0.032 0.038 t-statistic 1.601 0.718 2.552 0.942 0.667 -0.538 0.581 Note: Superscripts ***, ** and * represent significance levels at 1%, 5%, and 10%. The respective critical value for the t-test is 2.576(1%), 1.96(5%) and 1.645(10%). Estimates are based on the lag augmented VAR model.
3.4.4. Short-Run Dynamics
The augmented Granger causality test provides information whether there is a long-run relationship between the variables. However, it does not show how each variable reacts to innovations in other variables and whether the effect is time persistent or not.
The analysis of the GIRF provides an approach to investigate this relationship. In the following, the GIRF analysis is applied to the aforementioned VAR(1) model.
The following figure shows the response of the variables included in the VAR(1) model to a generalized one standard deviation innovation in the economic activity (EA) and real interest rate of China (RIR). Monte Carlo simulation procedure with 1000 replications is used to generate error bands. The dashed line shows the ±2 standard deviation error band, representing the 5% level of significance.
‧
Figure 3.5: Commodity Price Responses to Generalized One S.D. Innovations
-.04
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
69
The first column of the GIRF shows the response of the variables to a shock in the economic activity of China (EA). As for the commodity group prices the analysis indicates that energy (CP_EN) and industrial metals (CP_IN) prices respond positively to a positive shock in the economic activity of China (EA). While the response of the energy price (CP_EN) continues to be significant for around 16 months, the feedback of industrial metal prices (CP_IN) is less persistent, already becoming insignificant after around six months.
The positive response of these commodity prices to an unexpected shock in the economic activity of China is consistent with the findings of Roache (2012), where global oil and copper prices displayed positive and persistent responses to innovations in the industrial production of China in the time interval from 2000M01 to 2011M09.
The GIRF analysis further indicates that the real interest rate of China (RIR) responds negatively to an innovation in the economic activity of China (EA). One possible explanation for the negative response to an innovation in the economic activity is that the boost of the economic activity is followed by an increase in the general price level or a cut in interest rates of the monetary authority in an attempt to stimulate the economy.
The second column of figure Figure 3.5 indicates that agricultural commodity prices (CP_AG) respond negatively to a positive shock in the real interest rate of China (RIR) and vice versa. The overshooting behavior of commodity prices in response to a drop in the real interest rate is in line with Frankel’s (1986, 2008) commodity overshooting model as well as with empirical evidence on the relationship between real interest rate of the US and different individual- as well as commodity- group prices (Akram, 2009;
Frankel, 2008).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
70
3.5. Conclusion
This essay examined the long- and short-run dynamics between commodity prices of the agriculture, energy, industrial metals, livestock and precious metals group, economic activity and real interest rate of China over the period from 1998M01 to 2012M12. The time period starts at the point when China accelerated banking sector reforms and officially replaced its credit quota system by a target system and interest rates started to be increasingly determined by market forces.
Results from the T&Y (1995) type Granger causality tests provide significant evidence for a long-run relationship between China’s economic activity and global commodity prices. In particular, energy as well as industrial metals prices are Granger caused by China’s economic activity at the 1% and 5% level of significance respectively. As for the monetary policy of China, the analysis finds no significant evidence of a long-run relationship between the real interest rate and international commodity prices. The coefficient estimates of the exchange rate regime dummy variable, however, indicate at the 5% level of significance that industrial metals prices tend to be higher when the fixed exchange system of China is relaxed.
As for the short-run dynamics, the GIRF analysis indicates at the 5% level of significance that both energy and industrial metals prices respond positively to an innovation in the economic activity of China. As for China’s monetary policy, the GIRF indicates, at the 5% level of significance, that the agricultural commodity prices overshoot in response to a drop in the real interest rate of China.
These results confirm that global commodity markets have become increasingly interrelated with macroeconomic developments in China. The most apparent reason for this is that China, as illustrated in the introduction, has become a dominant player in both financial and physical commodity markets. Shifts in economic and monetary
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
71
policies have an impact on China’s domestic supply and demand patterns which in turn have a spillover effect on global commodity prices. Macroeconomic variables, in particular economic activity as well as real interest rate, may be useful predictors of future commodity price movements. This finding supports not only investors in commodity markets to better determine the extent to which they are exposed to changes in macroeconomic conditions of China, but also policy makers to assess the possible effects of economic and monetary policies on commodity prices.
Policymakers should be aware that abrupt policy interventions might lead to disruptive fluctuations of commodity prices. These sharp movements can have a severe impact on commodity consumers and producers of commodities as well as a destabilizing effect on the economy as a whole.
When formulating economic and monetary policies, decision makers should not only consider the effects on commodity markets in general but also the possible up- and down side effects of commodity price changes on market participants. It is important to note that while commodity producer’s generally benefit from an increase in prices consumers lose. The negative effect of high commodity prices can be particularly strong for consumers in developing countries. In these countries, a large proportion of households typically spend a relatively large share of their income on primary, especially agricultural and energy, commodities. Policies that bolster commodity prices have a disproportionately severe effect on available income and living standards of a large proportion of the population in these countries.
Further research is needed to examine the impact of China’s policies on specific commodities. From the domestic perspective for China, it would be also interesting to examine the effect of economic and monetary policy shifts on domestic commodity prices.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
72
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
73
4. Conclusion
The recent global financial and Eurozone crisis has attracted new attention to the occurrence of asset price booms and busts. Monetary policy has often been identified as a major driver of asset price cycles as well as an instrument to mitigate the possibly severe consequences on economic and financial stability.
This dissertation consists of two essays that examine the dynamics between monetary policy and asset prices in two specific markets. The first essay presented in chapter two looks at the role of the ECB in the formation of real estate bubbles in Greece, Ireland, Portugal and Spain. The reason for looking at this market and region is that these four countries stand in the epicenter of the ongoing financial crisis in Europe and that all of them relied strongly, above EU average, on the construction sector. The essay first examines the extent to which the four countries experienced property bubbles and then investigates the role of the monetary policy of the ECB in the formation of these real estate price bubbles. The findings on the extent of bubble formation indicate that Spain and Ireland experienced the largest positive bubble formation, followed by Portugal and Greece with a slightly increasing bubble. The results of the cointegration tests and impulse response analysis applied to the VAR and VEC models provide evidence for a weak short-run relationship between monetary policy and bubble formation in Portugal, but evidence for both, long- and short-run relationship in the case of Ireland, Greece and Spain. The qualitative analysis in the discussion part points out that the varying extent of the bubble formation and the differing impact of the monetary policy on the bubble across the four countries can be mainly attributed to characteristics in the domestic financial-, fiscal- and macroprudential system as well as to other structural factors. This essay
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
74
also emphasizes the argument of Confrey and Gerald (2010) that the introduction of the European monetary union brought about a dilemma. The single monetary policy framework not only hinders national central banks to cope with precarious developments in domestic housing markets, but also potentially induces booms and busts by artificially lowering the cost of capital and providing access to finance for investments in property markets. However, individual countries within the monetary union still have mechanisms, i.e. financial-, fiscal- and macroprudential as well as structural policies, at their disposal to either soften or strengthen the effects of single monetary policy shifts on domestic real estate markets.
The second essay presented in the third chapter focuses on the relationship between monetary policy and global commodity prices. In light of China’s emergence as world’s second largest economy and dominant player in commodity markets, the study attempts to shed light on the role of China’s policies in global commodity price dynamics. Specifically, the essay looks at the long- and short-run dynamics between global commodity prices of the agriculture, energy, industrial metals, livestock and precious metals sector, economic activity and monetary policy of China. In regard to monetary policy, T&Y (1995) type Granger causality tests provide no significant evidence for a long-run relationship between the real interest rate of China and international commodity prices. The coefficient estimates of the exchange rate regime dummy variable, however, indicate that industrial metals prices tend to be significantly higher when the fixed exchange system of China is relaxed. As for the short-run dynamics, the GIRF analysis indicates that agricultural commodity prices overshoot in response to a drop in the real interest rate of China.
In sum, the two essays show that monetary policy has an impact on asset prices across different regions and markets. The first paper shows that monetary policy can
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
75
significantly affect domestic real estate markets, whereas central bank’s policies are crucial to trigger the boom and burst of property bubbles by manipulating the interest rate and availability of lending for house purchase. The second essay looking at global commodity markets shows that the effect of monetary policy has the potential to affect asset prices that go beyond the domestic market. Specifically, the analysis shows that agricultural commodity prices overshoot in response to a fall in real interest rate of China.
The empirical findings are in line with both the monetarist and Austrian view to the extent that they acknowledge loose monetary policy as a driver of asset prices. It is important to emphasize that fluctuations in asset prices alone can have severe direct effects on market participants. As for real estate markets, a strong upsurge in prices squeeze individual’s purchasing power for housing for consumption purpose in favor of investors betting on higher returns, further fuelling the housing boom. Changes in
The empirical findings are in line with both the monetarist and Austrian view to the extent that they acknowledge loose monetary policy as a driver of asset prices. It is important to emphasize that fluctuations in asset prices alone can have severe direct effects on market participants. As for real estate markets, a strong upsurge in prices squeeze individual’s purchasing power for housing for consumption purpose in favor of investors betting on higher returns, further fuelling the housing boom. Changes in