Twenty-First Century Economics of the Gold Price

Abstract

The twenty-first century has seen two profound changes occur in the gold market. The first is that the gold price has surged in unexpected ways and secondly, the price appears to have become more volatile over time. This chapter considered the various factors that have shaped gold spot prices over the last two decades and examined the long term gold prices over the last two decades. This is important as two questions, for quite different reasons, arise and are of vital importance to Africa and its economic future. Leaving aside the obvious macro-economic factors associated with the rising or falling price of a dominant export, where Africa was in 2018 the world’s largest supplier of gold, there is an important microeconomic consideration. Without a rising or at least a high price of gold on the world market, the expansion of production of both artisanal small scale mining (ASM) and large scale mining (LSM) would not have occurred. Perhaps just as significantly, the question of volatility stems from the declining industrial demand for gold and the increasing demand for gold as a financial instrument. The traditional industrial role of gold for the production of jewellery is in relative decline. This raises a profound question as to whether the down-stream processing of gold can act as a basis for African industrialisation as it is so commonly assumed in so many of the high level policy pronouncements on the matter.

Appendices

Annex 1: Structural Changes of the Twenty-First Century and their Impact on the Gold Price

1.1 Methodology

The purpose of this analysis was to examine the various factors that have shaped the surge of gold spot prices over the last two decades using quarterly data. The period under review in this study is Q2, 2000 to Q3, 2017. The period was chosen as it coincides with the turn/beginning of the twenty-first century and because it also covers the full trade cycle for gold, which saw gold prices rise exponentially³¹ and then eased but not to pre-boom levels. The data used was collected from various sources that include: the World Gold Council (WGC),³² the Federal Reserve Bank of St. Louis³³ and the London Bullion Market Association LBMA.³⁴

Various methods have been applied in analysing price determinants, but as Borensztein and Reinhart (1994) asserted, a majority of these methods are single equation frameworks with the Ordinary Least Squares (OLS) being of preference. The analysis therefore applied a multiple linear regression to observe what empirical impact if any, the select twenty-first century variables have had on the gold price. Other tests include pre estimation tests and post estimation diagnostic tests to ensure the validity of the results. Therefore, the model applied in this analysis is:

$$ lnG{P}_t={\beta}_0+{\beta}_1 lnCG{D}_t+{\beta}_2 lnVIX+{\beta}_3 lnQ{E}_t+{\beta}_4 lnEX{C}_t+{\xi}_t $$

(2.1)

Where:

• GP - Gold price

• CGD - China gold demand

• VIX - Volatility index

• QE - Quantitative easing

• EXC - Exchange rate

• ξ - Error term

It is worth noting that while the literature underlines inflation as an important factor in determining the gold price, during the period of interest, global inflation was low and on a downward trend. For these reasons, Inflation was not considered as a significant factor in explaining the gold price surge over the past two decades.

1.2 Estimation Results and Discussion

After the pre estimation tests were satisfied, the models were estimated. As cointegration was found, the Error Correction Model (ECM) was estimated. The results are presented in the Table 2.1 below.

Table 2.1 ECM estimation

The Error Correction Term shows and reiterates a long-run relationship between the variables as it has a negative sign and is significant (p value < 0.05) at 0.0268. The coefficient shows the rate at which the disequilibrium of the system of the previous period will be corrected in one year; being at a rate of 14.07%. The model was found to be significant with the F-Statistic at 7.894445, and not spurious as the R squared (0.252802) is less than the Durbin Watson statistic (2.145239). From the results, it was observed that the R squared is low and there is only one insignificant t ratio of importance, implying that multicollinearity may be present but may not be an issue. This is all the more important given the individual significance of the variables, and the joint significance of the model. The Durbin Watson statistic (2.145239) shows that there is no autocorrelation which was confirmed by the Breusch-Godfrey Serial Correlation LM Test as F test statistic was in excess of 0.05, at 0.4665.

The other diagnostic tests included a normality check and a Heteroskedasticity Test. From the normality test, it was found that the normality assumption of the residual term had been supported as the P value was greater than 10%, 5% and 1% (at 0.192685), with the skewness statistic less than zero (at 0.001730), though the kurtosis was leptokurtic in shape. Furthermore, there was also no evidence of the presence of heteroscedasticity. Both the F and Chi-square test statistics suggested the same conclusion, that there is no evidence for the presence of heteroscedasticity, since the p-values were in excess of 0.05 respectively (0.7820 and 0.7711).

The estimation of the short run model, revealed that only consumer perception and the exchange rate are of significance. With consumer perceptions having a positive relationship with gold price. As expected, the exchange rate has a negative influence on gold price, implying that an appreciation of the exchange rate will be detrimental to gold price. Whereas a depreciation will not as Beckmann, Czudaj and Pilbeam (2015) assert. Furthermore, the Error correction term was found to be significant with the correct sign. However, as has been alluded to in the objective of the study, the long-run structural shift that has occurred is of principal interest. Therefore the long-run estimation is of importance to observe what variables, of the macroeconomic and structural factors are responsible for the price rise. The long-run results are presented in the Table 2.2 below.

Table 2.2 Long –run regression estimates

The initial estimation suggested that all the variables were significant in affecting the price of gold. With an R² of 0.964450 (approximately 96%), the goodness of fit is relatively high. Furthermore, it was observed that the F-statistic is 474.7618 which is greater than the Prob (F-statistic) = 0.0000 showing that the overall fit of the regression is good. However, with the Durbin-Watson statistic at 0.704427, the issue of autocorrelation is present, which warranted post estimation tests.

1.3 Post Estimation Analysis

The residual diagnostic tests for normality, autocorrelation and heteroscedasticity were used. From the normality test, it was found that the normality assumption of the residual term was supported as the P value was greater than 10%, 5% and 1% (at 0.156951). Implying the null hypothesis could not be rejected of a normal residual, with the skewness statistic fairly close to zero though the kurtosis was platykurtic in shape.

However, it was found that the model suffered from autocorrelation as evidenced from the D-Watson Statistic in Table 2.2 and as per the Breusch-Godfrey Serial Correlation LM Test since the F test statistic was less than 0.05, at 0.000. There was, however, no evidence of the presence of heteroscedasticity as both the F and Chi-square test statistics produced the same conclusion, that there is no evidence for the presence of heteroscedasticity, since the p-values were greater than 0.05 respectively (0.0714 and 0.0726). To address the problem of autocorrelation, the study used the Newey-West standard errors that are robust to autocorrelation.

Upon using the Newey-West standard errors, the model estimated provided significant variables (though the coefficients changed slightly) and unchanged variable signs just as before (See Table 2.3). However, the R² decreased to 0.830551 from 0.964450. Furthermore, upon testing for normality, it was observed that the normality assumption of the residual term had been supported as the P value was greater than 10%, 5% and 1% which is the same as the case of the long–run post regression analysis above.

Table 2.3 Long –run regression estimation with Newey-West standard errors

The estimated results suggest that all the variables except the composite exchange rate of the USD against major currencies have a positive relationship with gold price changes. The inverse relationship between the composite exchange rate and the gold price is similar to that found by Sjaastad (2008). This indicated that an increase in risk/perceived risk, Chinese gold demand and quantitative easing will cause the gold price to rise: whereas, a decline in the exchange rate index will cause an increase in the gold price. Conversely, from the empirical results, the positive relationship between the variables (except the exchange rate) with changes in the gold price implies that decreases in risk/perceived risk, China gold demand and quantitative easing would lead to a decrease in the gold price. Of further interest is the ‘ranking’ of the significance of the variables under study in terms of the long-run ‘Prob.’ Values. This being the exchange rate first, followed by Quantitative easing, China gold demand, then risk.

Individually and collectively, there is no doubt that these issues contributed significantly to the gold price surge during this period of interest. It is worth noting that among other variables considered that may have affected the gold price surge, was geo political risk. However, as it was found insignificant and its exclusion did not affect the study’s results it was eventually excluded.

Annex 2: Gold Jewellery Demand and Gold Price Volatility: A Global Perspective

The purpose of this analysis was to examine the impact on gold price volatility of the decline in gold jewellery demand and the concomitant increase in investment demand for gold. The study used quarterly time series data from Q1, 1998 to Q3, 2017. The period was chosen not only because of data considerations but because it also covers the full trade cycle for gold including the nine year bull run from 2002–2011 as well as the long subsequent decline in gold prices.

Gold price volatility function:

The study adopted the gold price volatility function of Hashim et al. (2017) which is stated as:

$$ {\displaystyle \begin{array}{c} lnGPVo{l}_t={\beta}_0+{\beta}_1 lnJD{D}_t+{\beta}_2 lnQE+{\beta}_3 lnIN{V}_t\\ {}+{\beta}_4 lnPSIL+{\beta}_5 lnDJ{I}_t+{\beta}_6 lnCD{D}_t+{\xi}_t\end{array}} $$

(2.2)

Where:

• GPVol - Gold price volatility

• JDD - Total world jewellery demand

• QE - Quantitative easing

• INV - Gold for investment use

• PSIL - Silver price

• DJI - Dow Jones Index

• CDD - China gold demand

• ξ - Error term

Prior to the estimation of the model, the volatility of the gold price was investigated. It was found from the sum of ARCH and GARCH coefficients (α+β) which were around unity (1.07), which indicated that the gold price was volatile and persistent.

1.1 Empirical Results and Discussion

After deriving the volatility series, the model in eq. 2.2 was estimated to examine the determinants of gold price volatility. The results are reported in Table 2.4. From the general statistics it can be observed that the value of the F-statistics is 117.28. As the Prob (F-statistic) is greater than the F-statistic (at 0.0000), this shows that the overall fit of the regression is good. Moreover it indicates that the model is significant at 1% significance level, thus, we can proceed with analysis as the parameters are jointly statistically significantly different from zero. Table 2.4 further reports that the goodness of fit is relatively high, as given by the R² of about 93%. A majority of the variables are also significant at 5% significance level.

Table 2.4 Long-run regression estimates

The estimated results report that there is an inverse relationship between gold price volatility and jewellery demand. A 1% increase in total jewellery demand will lead to a fall in gold price volatility by about 1.32% ceteris paribus. This indicates that a decline in jewellery demand will exacerbate gold price volatility while the increases in the investment use of gold will also increase volatility. A current decline in global jewellery demand is therefore among the factors that exacerbates gold price volatility. Whereas, a positive relationship was reported from quantitative easing, the Dow-Jones index, the price of silver, demand for gold by China and gold investment demand. The results of the modelling suggest that neither the effects of QE nor the impact of Dow-Jones as a proxy for alternative financial instruments are statistically significant in explaining long term gold price volatility.

The analysis does not, like many previous studies focus on short term price volatility but that which stems from a long term structural change in the composition of demand. The statistically most significant factors in explaining long term volatility are China demand, global investment demand and the demand for gold for jewellery consumption. India is not included due to the unreliability of Indian gold and jewellery data which is in turn driven by round tripping.

Post estimation, the residual diagnostic tests for normality, autocorrelation and heteroscedasticity were applied. Upon which it was found that the normality assumption of the residual term was supported. However, the model suffered from autocorrelation as evidenced from the d-statistics in Table 2.4 and therefore the Breusch-Godfrey Serial Correlation LM Test was applied to address the problem of autocorrelation. There was however no evidence of the presence of heteroscedasticity.

1.2 Conclusion and Implications

The results of the analysis confirm that the long term decline of gold jewellery demand over the last twenty years has a statistically significant and adverse effect on gold price volatility. The move away from gold jewellery by the current generation of Millennial consumers in a range of countries will mean that the demand for gold will become increasingly dependent upon investment demand and this, as has been shown will lead to an increase in volatility of gold prices. The two most statistically significant factors in explaining gold price volatility are global jewellery demand and investment demand. The emphasis which gold industry stakeholders such as the World Gold Council and the mining industry in general have traditionally placed on promoting gold jewellery consumption is therefore entirely justifiable but increasingly difficult in light of the changing patterns of luxury good consumption. Clearly an enhanced marketing strategy by key players lead by the World Gold Council in the gold jewellery sector is now essential to avoid a worsening of gold market instability.