Skip to content

Determinants of the Physical Demand for Gold: Evidence from Panel Data (2008)

by on July 7, 2017

This paper offered the first examination of the drivers of demand for physical gold. This is a research area where there has only been a minimal amount of work done as most of the published papers in this area focus on the demand for “Paper” based claims on gold such as Futures contracts and Exchange Traded Funds. The demand for gold in this paper is measured by physical gold imports, which covers gold in bar, coin and jewellery form. Below the findings are summarised and some significant reservations expressed about these.

Findings: The authors argue that the drivers of physical gold demand are significantly different from those that drive paper investment demand, though this area is not specifically researched here. Their paper finds significant differences in the drivers of demand for the different countries examined as might be expected, with China and India having significantly stronger demand than western countries.

Surprisingly the level of income in a developing country (GDP per capita) has no relationship to gold imports,  but this is a very strong driver in developed countries. A 1% increase in GDP per capital translates into between a 2.3% and 1.8% increases in demand. In developing countries gold consumption rises with falling growth in per capita income which the authors argue points a precautionary investment motive. The development of credit markets is concurrent with a decrease in the demand for gold imports for both groups which reinforces this safe haven type view of gold.

When the authors look at jewellery demand and physical investment demand (Coin and Bar) separately the results are very different. Firstly the vast majority of explanatory power for jewellery demand comes from lagged demand, with a coefficient of 0.86 to 0.98 depending. For the coin and bar estimations this falls to 0.16 to 0.35. So the investment portion of physical demand is far less stable.

The results for the drivers of bar and coin demand are unclear based on the 5 models estimated. In the fixed effects model only the level of the stock market is statistically significant and the coefficient is small (a 1% increases in the stock market capitalisation as a share of GDP increases gold coin and bar imports by 0.0035%). This effect is also found in 4 of the 5 models for both jewellery and physical investment demand.

Physical investment demand has a negative relationship with inflation is some models, but jewellery demand is positively related to inflation. Increases in 5 year GDP volatility increase jewellery demand in 3 out of 5 models.

Overall the only consistent explanatory variable for physical gold demand across the models is lagged gold imports. Additionally no indication of the economic significance of the relationships which are found to be statistically significant.

Problems: There are a number of significant problems with the paper which make the above findings questionable in my opinion.

  1. The paper does not discuss unit root or co-integration issues that might be present in the data, despite some of the variables used being commonly seen as non-stationary and this being a normal pre-testing procedure in financial economics papers. It is unclear whether the dependant variable (gold imports per capita) is stationary. If it is not all the estimations are invalid.
  2. The paper includes lagged gold imports per capital as an independent variable. This is done as “gold’s perceived value may depend in part on how highly others seem to value it, so that recent strong demand presently would bolster demand in future periods” (page 422). Following on from the previous point, if the paper had modelled changes in imports this would not have been an issue.
  3. Sample size is 230 for the panel but the number of explanatory variables used is 22 when year dummies are counted in, reducing the power of the tests. The models seem to include too many variables as some such as real interest rates have literally zero explanatory power in 4 of the 5 models, with a coefficient of 0.0001 in the 5th.
  4. Jewellery demand and Bar and Coin demand are treated as being the same for much of the paper. While jewellery is sometimes seen as an investment vehicle, in western countries jewellery is primarily bought for aesthetic purposes due to the high mark ups. Disaggregating these gives very different results.
  5. Year specific dummies are used. However the economic explanation for the inclusion of these variables is unclear. Their inclusion implies that the authors think that each year has events which are not captured but the other 11 regressors. The authors mention geopolitical factors as an example but recessions etc. would be expected to show up in inflation and/or real interest rate data again pointing to over specification.
  6. The division between Developing and Developed countries seem arbitrary, Singapore is listed as a developing country.

Methodology: Panel econometric techniques used: Pooled OLS, Fixed Effects, GMM

Data: Annual gold imports 1992 – 2003 for 21 Countries, GDP per capita, Inflation, Private credit as a share of GDP, Real Interest Rate, change in PM Fixing Gold price, Exchange rate Volatility

Citation: Starr, Martha, and Ky Tran. “Determinants of the physical demand for gold: Evidence from panel data.” The World Economy 31.3 (2008): 416-436.

Abstract: Although the role of gold in the world economy has declined since the gold standard was abandoned, it remains important as a central bank reserve, a hedge against risks, a barometer of geopolitical uncertainty, and an input for jewelry. While portfolio demand for gold has been well studied, determinants of physical demand are less understood. Certain emerging-market countries like China and India import substantial amounts of gold, with several factors that may contribute: low financial development, need for precautionary savings, and/or strong cultural valuation of gold itself. This paper uses panel data on gold imports of 21 countries to examine determinants of physical demand. We find that determinants of physical demand differ from those of portfolio demand, and that they differ between the developed and developing worlds.



From → Empirical, Gold

Leave a Comment

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: