Googling Gold And Mining Bad News
Kuehne Logistics University; Financial Research Network (FIRN)
University of Tuebingen – Department of Statistics and Econometrics
January 7, 2016
This paper studies investor’s attention to gold price movements by analyzing the relationship between internet search queries for gold and the price of gold. We find a long-run positive co-movement of gold price changes and search queries for gold and a short-run negative co-movement driven by negative gold price changes. The estimation results show that negative gold price changes attract more attention of investors than positive price changes indicating a preference to mine (google) bad news rather than good news. An analysis of silver, palladium and platinum demonstrates that the findings for gold are unique.
Googling Gold And Mining Bad News – Introduction
The price of gold and Google search queries for gold display a positive co-movement over the 12-year period from 2004 to 2015 as presented in Figure 1. Upon closer inspection and a focus on large changes of the price of gold and search queries, one can also identify a negative relationship for positive spikes in search queries (SQ) and negative spikes in the price of gold. This suggests that negative gold price movements may have a significant effect on SQs. Similar relations can also be detected for silver, palladium, and platinum. Following Da, Engelberg, and Gao (2011) we interpret google search queries as a measure for retail investors’ attention and assume that large asset price changes induce an elevated interest of retail investors. The focus of this paper lies on precious metals and in particular on gold. Since gold has a long tradition as a currency, a medium of exchange and a store of value, it is possible that investors behave differently with respect to gold than to silver, palladium, and platinum. Gold’s high value density as well as its bright and shiny reflection may further influence investors’ behaviour and attention. Hence, we aim to analyse the question if retail investors react to price changes by using Google and thus by mining information about the asset. Furthermore, we are interested in the question whether information generation (e.g. through googling) and possible subsequent trading of noise traders increases the volatility of the related asset as suggested by, amongst others, Black (1986) or Shleifer and Summers (1990).
There is a growing literature on internet search queries, investor’s attention, and sentiment. In their ground-breaking article, Choi and Varian (2012) use Google search queries to predict different economic indicators. Da et al. (2011) show that the volume of Google search queries serves as a direct measure of retail investor attention towards individual stocks. Stocks which receive high attention in the current week outperform other stocks by approximately 30 basis points in the following two weeks. Similarly, Drake, Roulstone, and Thornock (2012) use Google SQs as a way to approximate demand for (public) information around earnings announcements. Building on these findings, Vlastakis and Markellos (2012) and Dimpfl and Jank (2016, forthcoming) investigate the relationship between stock market volatility and Google search queries. They find that incorporating SQs in the prediction significantly improves the forecasting accuracy, a result which is well in line with the theoretical information demand model of Andrei and Hasler (2014, forthcoming). Recently, Da, Engelberg, and Gao (2015) proposed an aggregate index of Google search queries called “FEARS” index to measure investor sentiment. The authors show that this index is able to explain contemporary mispricing of stocks, volatility of stock prices, and the transfer of wealth from mutual funds to bond funds. Their results are fully consistent with the theory about investor sentiment in financial markets (see De Long, Shleifer, Summers, and Waldmann, 1990).
Investor attention in a gold context is discussed in Baur (2015) who looks at “attention-grabbing” in the context of different day and overnight returns. The main focus of his paper, however, is on claims that central banks influence and manipulate the price of gold. This gold manipulation claim may also be relevant for this study as there is often a rather harsh tone in internet blogs and comments. If discussions about gold are generally more heated than discussions about other precious metals, the relationship of gold prices and related search queries may also behave differently.
This paper contributes to the growing literature on internet search queries and investor attention with an analysis of precious metals. Using a new filter provided by Google, we are able to concentrate the search queries to searches related to the metals only, leaving aside unrelated searches which might involve the search term, but not be related to the asset price.
See full PDF below.