4. Empirical Results
4.2. ECDP VAR Models
4.2.2. ECDP Model 2—DXY, Interest Rates, and GDP
4.2.2. ECDP Model 2—DXY, Interest Rates, and GDP
The next set of variables used in conjunction with ECDPs were macroeconomic indicators.
These variables included one-year differenced DXY, one-month differenced GDP, and one-month differenced effective federal fund rates. Dickey Fuller unit root tests were already conducted on all the variables in this VAR model. A total of 13 lags were used in the VAR. The justification for the use of 13 lags stems from the lag order selection criteria results which found the AIC, FPE, and LR test statistics to be statistically significant with six lags. Specifically, this model uses two variables which are quite connected to each other: the DXY and the effective federal funds rate.
Clearly, GDP is connected with both of these variables as well, but theoretically currencies and interest rates are more immediately impacting one another. Moreover, capital prices are also quite connected to the exchange rate and interest rates as well from a theoretical perspective. When the interest rate increases, the cost of capital increases and the strength of the DXY increases as well since the demand for holding that currency increases. In addition to this, when the dollar becomes stronger, it causes the demand for domestic equipment to decrease since the competitiveness of international equipment becomes stronger. However, capital is inelastic in the short-term due to the fact that investment in capital is permanent in a sense (a factory for construction equipment cannot be easily transformed into a factory for automobiles). Therefore, the impact of the interest rate and the DXY are two different forces pulling on equipment prices in different directions during different durations of time. Because of this, the ramifications on the relationship between these variables is likely imperative. As a technical note, one-month differenced effective federal funds rate had to be used as opposed to one-year differenced because the one-year differenced data failed to reject the null hypothesis of the presence of a unit root. Therefore, this likely may impact the VAR results.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Table 4-24 Lag Order Selection Criteria Test Results for ECDPs, One-Year Differenced DXY, One-Month Differenced Interest Rates, and One-Month Differenced GDP
After the lag order selection criteria provided a lag order to fit the model, the equation of the VAR was conducted which found that all variables were statistically significant (see Table 4-24). It is important to keep in mind that the sample size for the dependent variable is much smaller than that of CEPs (only 212 observations). Therefore, this likely increased the statistical significance and strength of the relationship between the variables as had been the case for CEPs.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Table 4-25 VAR Equation Results for ECDPs, One-Year Differenced DXY, One-Month Differenced Interest Rates, and One-Month Differenced GDP
As can be seen by the above results, the R-squared statistic was quite high for ECDPs in addition to the DXY which were both one-year differenced. Similarly, one-month differenced GDP also enjoyed a very high R-squared statistic. However, one-month differenced effective federal funds rates suffered from a relatively low R-squared statistic. In spite of this, the results of the OIRFs show that it is easier to predict the movement of equipment prices from interest rates than other variables.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Figure 4-26 OIRF for the DXY Impulse and ECDPs Response
The first OIRF conducted was the DXY as the impulse variable with equipment prices as the response variable (see Figure 4-26). The results of the DXY were very similar to that of the DXY’s impact on CEPs. Essentially, what is likely occurring is that a stronger dollar causes the price to naturally increase since firms purchasing domestic equipment must pay more for the same machinery in their own currency. Eventually, exports decrease and imports increase because domestic firms lose out on market share from foreign firms since the stronger dollar can purchase more from abroad. Another dynamic that could be occurring is that the increase in prices from the aforementioned trend adjusts back to equilibrium. To do so, prices must essentially decrease in the later steps to return to normal equilibrium after prices increased in the initial steps from the ‘shock’.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Another factor to take into consideration is the fact that the US construction equipment industry effectively acts as an oligopoly through not increasing the quantity to the point of equilibrium which can be amounted to reducing quantity over time in order to keep prices elevated. In that instance,
In spite of these trends, it should also be noted that the 90 percent confidence bands are both only consistently negative or positive for any number of steps is in the negative side of the horizontal zero reference line. Because of this, there is only a 90 percent confidence that a shock from the DXY has a negative impact at some point in time on ECDPs. Since this is the case, it can be surmised that the only impact on equipment prices is that they are lowered from the trade weighted trade increasing meaning that the latter explanation previously discussed of the relationship between these two variables is more statistically valid than the former. Simply put, the DXY increasing lowers equipment prices because it increasing international competition.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Figure 4-27 OIRF for One-Month Differenced Interest Rates Impulse and ECDPs Response
The impact of interest rates on equipment prices is much clearer, although with the latter steps it becomes less so (see Figure 4-27). Clearly, interest rates increase the cost of capital since it is the cost of capital. This can be seen as a statistically significant relationship as well since both 90 percent confidence bands are positive for nearly all of the first ten steps after the initial ‘shock’.
What is less clear is the last five steps of the OIRF which asserts at a 90 percent confidence level that prices will decrease. A few explanations could be made. The first is that prices are merely returning to equilibrium after a shock. Eventually, higher interest rates do not have as much of an impact on investment and the cost of capital since construction still occurs during times of higher
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
interest rates. Moreover though, the model takes into consideration the impact of the dollar which also is negative during the last five steps of the OIRF. Because of this, it may be the case that a lower DXY is the result of exogenous currencies and exogenous interest rates. Residuals from this model also may factor in other international macroeconomic variants.
Figure 4-28 OIRF for One-Month Differenced GDP Impulse and ECDPs Response
This final OIRF looks at the relationship between GDP and equipment prices. As can be seen by Figure 4-29, the moving average is nearly entirely negative, but the 90 percent confidence bands are only both negative during the first step after the initial ‘shock’. While it can be asserted that this clearly shows a negative relationship between these two variables, it appears more
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
accurate to suggest that the relationship between short-term GDP and long-term equipment prices is only certainly negative in the term. Longer-term impacts on equipment prices from short-term GDP changes is less strong even though the moving average is nearly entirely negative.
Table 4-26 Lagrange-Multiplier Test for ECDPs VAR Model 2
After the OIRFs were conducted, a Lagrange-multiplier test was performed. As can be seen by the above results, only the fifth, sixth, and twelfth lags rejected the null hypothesis of the test.
Therefore, it can be surmised that the overall model is valid in terms of serial correlation.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Table 4-27 Granger Causality Test Results for ECDPs, One-Year Differenced DXY, One-Month Differenced Interest Rates, and One-Month Differenced GDP
The next step of this model is to use the Granger causality test. The results of the Granger causality tests are clear in the continued assertion that equipment prices are more accurately a response variable as opposed to an explanatory variable (see Table 4-27). Only the effective federal funds rate Granger causes equipment prices, but equipment prices do not Granger cause any variables. Other relationships also exist within the model. For instance, interest rates and GDP Granger cause the DXY while interest rates also Granger cause GDP. Clearly then, it can be asserted that interest rates are the most useful explanatory variable in the model which is visually represented below in Figure 4-29.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Figure 4-29 Granger Causality Map for ECDPs, One-Year Differenced DXY, One-Month Differenced Interest Rates, and One-Month Differenced GDP
After the Granger causality tests were conducted, dynamic forecasts were implemented.
As can be seen, equipment price movements are certain given the model’s variable and lag structure. However, the movements of the explanatory variables themselves are less certain. Only GDP differenced by a month is certain to be negative and it is only negative for a certain number of steps after the first initial steps in which the assertion is that the one-month differenced rates are negative.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Figure 4-30 Dynamic Forecasts for ECDPs, Year Differenced DXY,
One-Month Differenced Interest Rates, and One-One-Month Differenced GDP
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
4.2.3. ECDP Model 3—Industrial Production, Iron and Steel Prices, and Diesel Prices