5. Analysis of the Results
5.4. Vector Auto-Regression Orthogonalized Impulse Responses
primarily due to the prior empirical evidence on the relationship between the effective federal fund rate against construction equipment and general equipment prices which was statistically significant for every model in all three of the one-year differenced regression matrixes.
Theoretically, it also stands in conjunction with what has been thought of to increase capital prices.
With this specific series of models though, it appears to enjoy a much stronger relationship than the others given the coefficients. Again, this could be due to the fact that the variable is a reflection of specific pieces of equipment as opposed to many pieces of machinery all indexed into one figure. Either way, the data obviously indicates that the more specific the figures are with the type of machine used, the less useful it is as a correlate of macroeconomic variables. While this means that it makes it more difficult to predict as a dependent variable, it likely (conversely) becomes less useful as a predictor variable even though these models did not use any of the dependent variables as independent variables.
At any rate, when using the R-squared and adjusted R-squared statistics in comparison from model to model, it seems to be clear in this set of regressions that the effective exchange rate and the ratio of inventory over sales are not the only contributors to price action or price variation in the long-term. However, they do seem to enjoy an outsized impact compared to a number of variables in the model. The same could be said as well for the NASDAQ Composite and data processing CPI.
5.4. Vector Auto-Regression Orthogonalized Impulse Responses
Most of the VAR results were significant with the OLS results. This not only validates OLS method which held time constant, but also reinforces the findings by increasing the robustness of the findings. Similarly to the other results of the OLS models, the VAR results as extracted from the OIRFs appear to also demonstrate that there is a negative relationship between the most
‧
statistically significant variables in the OLS models and the primary dependent variable of CEPs.
This is the case for the VAR with macroeconomic variables, technological variables, and commodity variables. Moreover, variables which were negatively correlated in the OLS model were similarly negative in the VAR OIRFs.
To summarize the findings of the VARs, the NASDAQ Composite was found to inflict a negative impact on CEPs when it was differenced by a year and when it was differenced by a month. The DXY shortly at first inflicted a positive impact on CEPs, but the majority of the impact was negative while diesel and steel prices caused a positive impact. GDP differenced by a month caused a negative impact with industrial production causing a positive impact. All other variables could not be shown to cause a positive or negative impact. And tech CPI caused a negative impact in addition to interest rates differenced by a month. Diesel prices were found to have caused a positive impact. All other variables could not be shown to cause either a positive or negative impact with 90 percent certainty.
With some statistical evidence, it may be suggested that the DXY and diesel prices play a role in price fluctuations according to the VAR models. It also should be recognized that the DXY and diesel prices are in many ways inversions of one another. Without a doubt, there is a strong negative correlation between petroleum prices and the DXY. Oil prices and diesel prices share a near 100 percent correlation, asserting that the relationship between the DXY.
In spite of the likely cointegration between the different equipment price measurements, GDP is consistently statistically significant with CEPs, but not with ECDPs. As can be seen by the scatterplot between these variables, the linear and LOWESS fit for CEPs is much more consistent with one another whereas the LOWESS fit for ECDPs for much of the scatterplot is actually positive as opposed to the negative coefficient which can be seen throughout the OLS regressions.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
At any rate, this suggests that CEPs are much more connected to economic growth than ECDPs.
The reason for this may be due to the fact that CEPs measure a much broader range of equipment prices whereas ECDPs only measures three specific types of equipment prices. Since this is the case, ECDPs are less a reflection of economic growth than CEPs since they measure a much smaller segment of the overall marketplace.
Figure 5-19 CEPs, ECDPs, and GDP
It is also important to keep in mind that comparing the totality of the dataset between CEPs and ECDPs is somewhat like comparing apples and oranges. It was pointed out in length the degree in which relationships change when limiting the time series for CEPs. Correlations strengthened dramatically across the board. This is most evident in the OLS models as was previously discussed.
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
Moreover, tests conducted with longer time series appear to show more relationships as is evident in the OLS results in conjunction with the VAR OIRF results and the Granger causality results.
Apparently, larger datasets are more generous with statistical significance while smaller datasets are more generous towards the strength of these relationships. What remains more evident though is the notion that in spite of these differences, a number of relationships remain the same regardless of the degree to which the time series is available or not. This is clear in the relationship between these two variables and the NASDAQ Composite differenced by one-year.
Figure 5-20 GDP and Industrial Production of Business Equipment One-Year Differenced
‧
國立 政 治 大 學
‧
N a
tio na
l C h engchi U ni ve rs it y
At any rate, the relationship between GDP, industrial production, and equipment prices is also key. GDP and industrial production of business equipment enjoys a very strong positive relationship as can be seen by Figure 5-20. This is also demonstrated in this correlation coefficient as can be seen by Table 3-3. If then prices of equipment and GDP is increased by industrial production, why then is GDP and equipment prices negatively correlated? This is likely due to the way in which the data was differenced. Since equipment prices lag GDP, The true nature of the relationship between GDP and equipment prices is concealed by the progression of time and the rapidity in which GDP can change its momentum.