Evaluación de nanocompuestos de RA con S. aureus

Using the threshold point for USD/AUD real exchange rate volatility found in Section 7.5, this study creates a dummy variable and an interaction variable in order to analyse the threshold effect of USD/AUD real exchange rate volatility on real labour productivity in Australia. The dummy variable (DUM) is an indicator that triggers the threshold effect when the USD/AUD real exchange rate exceeds or is equivalent to the threshold value of -6.0629 estimated in the preceding section. The interaction term captures the effect of the higher USD/AUD real exchange rate volatility on real labour productivity in Australia. The results from the OLS estimation of the threshold trade model are summarised in Table

7.6.⁵⁶The real labour productivity threshold model estimated in this section follows equation (5.19) in Chapter 5 as follows:

t t t

t t

t

t LKD LTO LR LV DUM INTERACTION

LLP =α₂ +β₁ +β₂ +β₃ +β₄² +γ₁ +γ₂ +ε₂ (5.18)

Table 7.6: Results for the Labour Productivity Threshold Model Variables

C 0.263070

(0.312116) [0.842860]

LKD 0.674805*

(0.049604) [13.60375]

LTO 0.269241*

(0.023471) [11.47110]

LR -0.072942*

(0.015281) [-4.773249]

LV 0.079995*

(0.011086) [7.215716]

DUM -0.454982*

(0.132560) [-3.432267]

INTERACTION -0.075760*

(0.022142) [-3.421614]

R-square 0.988136

Notes: Heteroskedasticity-Consistent Standard errors in ( ) Heteroskedasticity-Consistent T-statistics in [ ] * Significant at 5 per cent level.

56Significant autocorrelation and heteroskedasticity are found in the basic and threshold productivity models.

Hence, autocorrelation/heteroskedsaticity consistent standard errors and t-statistics are generated using the White heteroskedasticity consistent covariance matrix option in EViews (White, 1980).It should however be noted that the use of the heteroskedasticity consistent t-statistics does not appear to have affected the significance of the variables greatly as all variables retain their significance before and after the corrections were made.

As seen in Table 7.6, all the variables in the system are statistically significant at the 5 per cent level.⁵⁷ The estimated coefficients for LKD, LTO, LR and LV have the same signs as the basic productivity model. In addition, the variables have highly similar magnitudes to the ones in the basic productivity model. Hence, when DUM=0 (when there is no threshold effect), similar economic interpretations on these variables from the basic productivity model can be applied to the threshold productivity model as well. Overall, the explanatory power of the threshold productivity model is relatively high as well, with an R-squared of approximately 98.8 per cent.

From Table 7.6, the results can be summarised by equations (7.1) and (7.2) illustrating the scenario when there is no threshold effect and when there is a threshold effect respectively.

LL)P_t =0.26+0.67LKD_t +0.27LTO_t −0.07LR_t +0.08LV_t

(7.1)

when τ <−6.0629

LL)P_t =−0.20+0.67LKD_t +0.27LTO_t −0.07LR_t +0.0042LV_t

(7.2)

when τ ≥−6.0629

It can also been seen from Table 7.6 that the interaction factor is negative and significant.

This suggests that when USD/AUD real exchange rate volatility exceeds the threshold value of -6.0629, the individual effect of USD/AUD real exchange rate volatility on the Australian real labour productivity declines from 0.08 per cent to 0.0042 per cent, remaining positive.

In addition, when the USD/AUD real exchange rate enters this excessively volatile period (when DUM=1), the significant coefficient of -0.45 on DUM indicates that the exogenous

57 The t-distribution critical value at the 5 per cent level of significance is 1.6669.

value of Australian real labour productivity is negatively affected. In fact, when the USD/AUD real exchange rate is excessively volatile, the impact on the exogenous value of real labour productivity in Australia switches from a positive 0.26 per cent to a negative 0.20 per cent.

If only the USD/AUD real exchange rate volatility and its threshold effects are considered in isolation, then we see that in periods where the USD/AUD real exchange rate exceeds the threshold value of -6.0629, the overall impact of the excessive real exchange rate volatility is negative. This suggests that when the threshold effect of USD/AUD real exchange rate volatility is accounted for, Harris (2002)’s hypothesis that an increase in real exchange rate volatility negatively impacts the real labour productivity can be observed.

7.7 Tests for Cointegration

To ensure that the basic and threshold productivity models have “super-consistent”

estimates, it is important to make sure that a long run equilibrium relationship (cointegration) does exist among the variables in the system. In addition, from earlier discussions, we know that the inclusion of dummy variables should not affect the other coefficients in the cointegrating relationships to be estimated in this section.

In this section, we employ the same methods utilised in the empirical analysis on trade in Chapter 6. The appropriate lag length criteria considered in this chapter are: the sequential modified LR test statistic (LR), Final prediction error (FPE), Akaike information criterion (AIC), Schwarz information criterion (SC) and Hannan-Quinn information criterion (HQ).

As in Chapter 6, the chosen lag length should be minimised in more of the information criteria than other lag lengths. Therefore, with this objective in mind, this study selects a lag length of 6 for the unrestricted VAR model. However, it should be noted that, in the event that there are competing choices of lag length to choose from, the one that gives us no serial correlation will be chosen. As such, this results in a lag length of 5 for the VECM to be estimated in the following section. Similarly, a lag length of 6 for the unrestricted VAR model is found to be suitable for the threshold productivity model.

Table 7.7: Johansen Cointegration Test for the Basic Productivity Model Notes: Critical values at 5 per cent level of significance in parentheses.

* Denotes non rejection of the null hypothesis at the 5% significance level

As seen in Table 7.7 above, conflicting results are found. The λ_trace test selects Model 3 with rank 1 and theλ_maxtest selects Model 2 with rank 1. Due to conflicting results, diagnostic tests are conducted on the two plausible models. Model 2 with rank 1 is selected as it performs better in the diagnostic tests, giving us a better interpretation of the cointegrating relationship.

Since the cointegrating vector is not unique, a normalisation procedure is used, normalising the first r series (LLP) in the vector to an identity matrix, so that estimates of the cointegrating vector can be obtained. The normalised cointegrating relationship, assuming the cointegration rank =1, may be expressed in the equation (7.3) below with the standard errors in parentheses:

LV 0.004265 LR

0.260007

-LTO 0.495611 LKD

0.246191

-5.092567 _t+ _{t t} + _t

t = P L L)

(7.3) (0.87863) (0.16504) (0.05644) (0.04250) (0.02226)

As can be seen from above, real capital deepening (LKD) was found to have a counter-intuitive negative effect although not surprisingly this coefficient is statistically insignificant at the 5 per cent level.⁵⁸ In addition, LV is found to be statistically insignificant as well. All the variables have signs that are similar to what has been obtained from the OLS estimates discussed in the earlier sections. It can be suggested that the insignificance as well as ‘wrong’ sign of LKD may be due to real capital deepening not being a significant factor in affecting real labour productivity in Australia. In particular, human capital instead of physical capital may be a better contributing factor to real labour productivity in Australia. As noted by Dowrick (2003), if the average years of schooling of young people in Australia increases by 1 year, real GDP would increase by up to 8 per cent over an approximate time span of 40 years. Hence, this contributes to an increase in real labour productivity in Australia.

On the other hand, results from the cointegrating equation supports the competitiveness view of negative (positive) effects of real exchange rate appreciation (depreciation) on real labour productivity. As such, this implies that the competitiveness approach is valid in the longer term as well, which opposes the view that exchange rate depreciations will have a negative impact on labour productivity in the longer term.

Overall, a long term equilibrium relationship is found between LLP, LTO and LR. The insignificant LV term is consistent with the possibility that the threshold effect of LV needs to be taken into account of before we observe a significant impact of LV on LLP.

The Johansen (1992) cointegration test has also been conducted on the threshold productivity model to ensure that the economic interpretations made on the model

58 The t-distribution critical value at 5 per cent level of significance is 1.6663.

estimated in Section 7.6 are plausible ones.⁵⁹ This study finds cointegration in the threshold productivity model, with a possibility of the existence of 6 cointegrating relations using Model 3. The Johansen cointegration test for the productivity threshold model is summarised in Table 9.7 in Chapter 9.

7.8 VECM Estimation

The estimation of the VECM for the basic productivity model is presented in this section.

Due to time constraints, the VECM for the threshold productivity model will not be covered in this thesis and is left as an area for further research. Residuals and correlograms plots which can be found in Figures 9.13 and 9.14 in Chapter 9 appear to be generally that of white noise processes as well. This indicates that the estimated VECM is adequate and stable. Further confirmation can be found in the LM autocorrelation test conducted for the basic productivity model can be found in Table 9.11 in Chapter 9.

Table 7.8: Error Correction Terms from Basic Productivity VECM

D(LLP) D(LKD) D(LTO) D(LR) D(LV)

Error Correction Term (ECT)

0.043181 (0.05217) [ 0.82778]

0.105904 (0.07323) [ 1.44628]

-0.343405 (0.18399) [-1.86648]

0.104488 (0.35057) [ 0.29805]

3.373591 (0.57261) [ 5.89157]

Notes: Standard errors in ( ) T-statistics in [ ]

As seen in Table 7.8, the speed of adjustment coefficients for LTO and LV are statistically significant at the 5 per cent level.⁶⁰ Therefore, when there is a positive disequilibrium between LLP, LKD, LTO, LR and LV in this quarter, LTO will decrease and LV will increase in the next quarter to bring the system back into long run equilibrium. As we can see, such an adjustment in LTO is logical as a decrease in trade openness will decrease trade volumes and hence real GDP. This will lead to a decline in real labour productivity in Australia, bringing the system back into equilibrium. On the other hand, the positive and

59 The interaction term for the threshold productivity model has been pre-tested for stationarity. It is found to be I (1). In addition, as noted in Chapter 5, although dummy variables are neither considered as I (0) nor I (1) variables, they can be incorporated into cointegrating equations.

60 The t-distribution critical value (relevant for this section) at the 5% level of significance is 1.6663.

significant adjustment coefficient on LV suggests that an increase in LV will bring labour productivity back to long run equilibrium which is plausible if Harris (2002)’s approach is adopted. Although the coefficient on D (LR) is insignificant, the sign on it is logical. By application of the competitiveness approach in the short run, a positive disequilibrium in LLP should lead to an appreciation of the real exchange rate (an increase in real exchange rate) in order to reduce the real labour productivity.

7.9 Granger Non Causality Tests

Granger causality tests are essential for the empirical analysis on real labour productivity in Australia in order to determine if a bi-causality effect runs between USD/AUD real exchange rate movements/volatility and real labour productivity in Australia.

Table 7.9: Granger Causality Results based on Basic Productivity VECM

H0 Chi-sq test statistic p-value LR does not granger cause

LLP

1.786721 0.8778

LV does not granger cause LLP

2.102409 0.8348

LLP does not granger cause LR

11.11623 0.0491*

* Statistically significant at the 5% level.

The granger causality tests conducted and summarised in Table 7.9 above indicates that the USD/AUD real exchange rate and its volatility does not appear to contribute to real labour productivity in any manner on an aggregate basis. However, this study finds uni-causality running from LLP to LR. This is predicted by the Balassa-Samuelson hypothesis that states that trends in real exchange rates are accruable to differential trends in relative price of non-traded goods. In turn, these differential trends in relative prices are driven by differences in national productivity growth rates.

While bi-causality effects do not appear to run in the aggregate economy, it is plausible to find bi-causality effects in the traded/non-traded sectors and individual sectors of the economy. This thesis has only conducted the study based on the aggregate economy. As such, it appears instructive to conduct similar tests on different sectors of the Australian economy.

7.10 Innovation Accounting Impulse Response Functions

The impulse response functions (IRFs) are illustrated in Figure 7.3, with the tabulated impulse response functions found in Table 7.10. The impulse response functions presented in Figure 7.3 map out the 24-quarters (6 years) responses of a one standard deviation shock to the set of innovations on current and future values of the endogenous variables: LLP, LKD, LTO, LR, LV. As seen from Figure 7.3, a one standard deviation positive shock to LLP and LKD seems to affect LLP and LKD significantly and permanently, vice versa for a period longer than 24 quarters.

Interestingly, the tabulated impulse response functions as seen in Table 7.10 tells us that with a one standard deviation positive shock in LR, the response of LLP is negative for the first 9 quarters and positive thereafter. For brevity sake, only the first 12 quarters of the tabulated impulse response functions will be shown in Table 7.10. The impulse responses appear to be generally consistent with Harris (2001)’s hypothesis in the longer run which states that real exchange rate depreciations (appreciations) will have a negative (positive) impact on real labour productivity. This result is also consistent with the empirical analysis Harris (2001) who finds that recent exchange rate depreciations (appreciations) actually increases (decreases) labour productivity. However, longer term deviations from the purchasing power parity (as a misalignment measure) worsen it (improves it), both by relatively small amounts.

- .02 Response of LLP to LLP

- .02 Response of LLP to LKD

- .02 Response of LLP to LT O

- .02 Response of LLP to LR

- .02 Response of LLP to LV

- .04 Response of LKD to LLP

- .04 Response of LKD to LKD

- .04 Response of LKD to LT O

- .04 Response of LKD to LR

- .04 Response of LKD to LV

- .02 Response of LT O to LLP

- .02 Response of LT O to LKD

- .02 Response of LR to LLP

- .06 Response of LR to LKD

- .06 Response of LR to LR

- .06 Response of LR to LV

- .6 Response of LV to LLP

- .6 Response of LV to LKD

- .6 Response of LV to LR

- .6 Response of LV to LV Response to Generalized One S.D. Innov ations

Figure 7.3: Impulse Response Functions – Basic Productivity VECM

Table 7.10: Tabulated Impulse Response Functions Response of LLP

Period LLP LKD LTO LR LV

1 0.005568 0.003478 0.000494 -0.000473 0.001406

2 0.007464 0.003675 -3.98E-05 -0.000883 0.001667

3 0.007799 0.004043 -0.000549 -0.001245 0.001983

4 0.007816 0.004042 0.000275 -0.001183 0.001966

5 0.012092 0.008051 0.001352 -0.002410 0.002909

6 0.012959 0.008777 0.002348 -0.001996 0.002497

7 0.012932 0.008871 0.001397 -0.002032 0.002888

8 0.012978 0.008887 0.001539 -0.001235 0.002438

9 0.016888 0.012396 0.001955 -0.000969 0.002465

10 0.018004 0.013435 0.002563 0.000375 0.001544

11 0.018600 0.014150 0.001454 0.001103 0.001725

12 0.019346 0.014517 0.001109 0.002512 0.001447

Variance Decompositions

Similar to Chapter 6’s analysis, variance decompositions for the empirical analysis on real labour productivity are analysed as well. Inspection of Figure 7.4 shows that the results of the variance decompositions are mostly consistent with the granger causality tests conducted earlier on. For instance, the proportion of variance in LTO is due its own innovations and variations in LR (real exchange rate), which the granger causality test conducted in this chapter has found that LR does granger cause LTO.

On the other hand, the proportion of variance in LLP is largely due to its own innovations, with a small proportion accounted for by variations in LR (real exchange rate) in the later quarters. While granger causality tests conducted earlier on are unable to find a link flowing from LR to LLP, the variance decompositions seem to suggest that real exchange rate does significantly influence the variance in LLP to a small extent in the later quarters.

0 Percent LLP variance due to LLP

0 Percent LLP variance due to LKD

0 Percent LLP variance due to LT O

0 Percent LLP variance due to LR

0 Percent LLP variance due to LV

0 Percent LKD variance due to LLP

0 Percent LKD variance due to LKD

0 Percent LKD variance due to LT O

0 Percent LKD variance due to LR

0 Percent LKD variance due to LV

0 Percent LT O variance due to LLP

0 Percent LT O variance due to LKD

0 Percent LT O variance due to LT O

0 Percent LT O variance due to LR

0 Percent LT O variance due to LV

0 Percent LR variance due to LLP

0 Percent LR variance due to LKD

0 Percent LR variance due to LT O

0 Percent LR variance due to LR

0 Percent LR variance due to LV

0 Percent LV variance due to LLP

0 Percent LV variance due to LKD

0 Percent LV variance due to LT O

0 Percent LV variance due to LR

0 Percent LV variance due to LV Variance D ecomposition

Figure 7.4: Variance Decompositions – US basic productivity VECM

Table 7.11: Tabulated Variance Decompositions – US basic productivity VECM Variance

Decomposition of LLP

Period S.E. LLP LKD LTO LR LV

1 0.005568 100.0000 0.000000 0.000000 0.000000 0.000000 2 0.009457 96.94719 1.786495 0.382378 0.059784 0.824152 3 0.012385 96.17671 1.775458 1.082520 0.185744 0.779569 4 0.014708 96.43685 1.794153 0.814108 0.263190 0.691696 5 0.019105 97.22168 1.174271 0.495699 0.694596 0.413755 6 0.023157 97.48936 0.941367 0.572689 0.667854 0.328730 7 0.026559 97.81790 0.861758 0.438639 0.627254 0.254454 8 0.029584 98.08122 0.808467 0.363358 0.507359 0.239598 9 0.034168 97.95838 1.084933 0.276888 0.406776 0.273027 10 0.038827 97.36277 1.361146 0.246852 0.572543 0.456690

7.11 Concluding Remarks

In this chapter, the relationship between the USD/AUD real exchange rate/volatility and real labour productivity in the Australian aggregate economy has been analysed by using a basic productivity model and a threshold productivity model.

Using the basic productivity model, this study finds that an appreciation (depreciation) of the USD/AUD real exchange rate has a negative (positive) impact on real labour productivity in Australia, consistent with the ‘competitiveness’ approach. Similarly, an increase (decrease) in the USD/AUD real exchange rate volatility has a positive (negative) influence on real labour productivity in Australia.

According to the cointegration test conducted, there is a significant long run equilibrium relationship between the variables in the basic productivity model. However, real exchange rate volatility (LV) and real capital deepening (LKD) are found to be insignificant in explaining real labour productivity in Australia. On the other hand, an appreciation of the real exchange rate has a negative impact on real labour productivity in Australia, which is consistent with the OLS estimates obtained in the basic productivity model. Hence, the

implication of this has been that Harris (2001)’s proposition which has been argued for the case of Canada does not seem applicable to the Australian aggregate economy. Instead, the

‘competitiveness’ approach that states that a real exchange rate appreciation (depreciation) will impact the real labour productivity in a negative (positive) manner seems applicable.

The Granger causality test conducted in this chapter also finds no evidence of the USD/AUD real exchange rate/volatility influencing real labour productivity in Australia.

Instead, the results from the Granger causality test appear to provide evidence in support of the Balassa-Samuelson hypothesis in Australia. This means that trends in real exchange rates are accruable to differential trends in relative price of non-traded goods. In turn, these differential trends in relative prices are driven by differences in national productivity growth rates.

In addition, the threshold effect of USD/AUD real exchange rate volatility has also been incorporated and estimated. Strong evidence points to the existence of a significant threshold effect of USD/AUD real exchange rate volatility. The results from the real labour productivity threshold model estimated in this study can be summarised as follows. If only the USD/AUD real exchange rate volatility and its threshold effects are considered in isolation, then we see that in periods where the USD/AUD real exchange rate exceeds the threshold value of -6.0629, the overall impact of the excessive real exchange rate volatility is negative. In turn, this implies that Harris (2002)’s hypothesis appears to be only applicable when the USD/AUD real exchange rate is excessively volatile.

With the decline in real labour productivity in Australia in recent years, further research on the link between real exchange rate movements/volatility and real labour productivity in the Australian aggregate economy and the various individual sectors is warranted. While previous micro-economic reforms have helped raise real labour productivity levels in Australia, the results from this study has shown that the effects of excessive real exchange rate volatility on real labour productivity can be negative and future reforms could possibly account for that.

CHAPTER 8: CONCLUSIONS AND DIRECTIONS FOR

In document CENTRO DE INVESTIGACIÓN EN QUÍMICA APLICADA (página 79-0)