The objective of this thesis is to answer the question, “how do the characteristics of gold deposit transactions affect their price”? Four hypotheses are posed which address fundamental variables that affect the price of gold deposits. Despite the obvious and significant nature of these variables, their influence on gold deposit prices is poorly described. As well as advancing the scientific knowledge base, this research has direct commercial relevance as it uses public domain data and methodologies that can be readily adopted by mining industry professionals. Even small improvements in the understanding of gold deposit transactions have a significant monetary value, given that during the five-year period between 2008 and 2012 the global gold deposit transactions tallied to (US)$75 billion (Wright 2014).
This thesis demonstrates that there is no single ‘going rate’ for an ounce of in-situ gold in a deposit. The data investigated shows that the size and grade of a gold deposit strongly affects both its price, as well as its price behaviour. It is shown that level of ownership acquired in a gold deposit affects the $/oz Au unit price, and that less than 100% ownership can lead to substantially higher $/oz Au prices being paid during risk-averse market conditions. However, it is shown that in risk-tolerant markets that less than 100% ownership can lead to relatively modest discounts. A reversal in the market behaviour is also observed in the research in country risk, with heavy discounts being applied for less favourable country exposures during risk-averse market conditions. It is also shown that the rate at which a deposit’s price changes is disproportionate to the prevailing gold metal price, and that increasing the confidence of a mineral estimate usually leads to an increase in the deposit’s price. However, it is observed that the market is strongly stratified with small deposits either expressing different rates of price change relative to larger deposits, or different markedly different behaviour. For example, small deposits tend to achieve lower prices when their mineral estimates become highly certain. These behaviours are not shown in the literature nor are they correctly accounted for in industry practice, which means that the research outcomes have direct commercial relevance as well as academic value.
Through a review of the academic literature, it was identified that the field of mineral asset pricing is a poorly researched field, which falls between a number of well-documented fields such as gold deposit valuation, real estate pricing and security pricing. Often, the term price (what you pay) is used interchangeably with value (what you get). Price is heavily influenced by exogenous aspects (e.g. negotiating with a willing buyer), whereas value is oriented more toward endogenous characteristics (e.g. quantity, quality, depth, morphology). Within the pricing field, there is often an assumption that the price of a security is a direct measure of the price of the underlying asset. However, the process of securitising an asset fundamentally changes its ownership structure and market liquidity, and in doing so, will affect the price (Yiu et al. 2006). The confusion between asset value, asset price, and different ownership structures (e.g. securitisation) is because of “a general lack of mineral property valuation [pricing] understanding” (Lilford 2004). In part, this is due to the need for an interdisciplinary understanding of gold deposit price drivers that requires competence in the technical aspects of the mining industry; an understanding of gold deposit valuation; and how, along with other project-specific and macroeconomic variables, the interaction leads to the determination of the price for a gold deposit.
The Lilford (2004) thesis is a key document in describing the price behaviour of a gold deposit given its size, grade, depth, location and mineral estimate confidence. Building on that research, this thesis hypothesises that:
Ownership risk – the price of a gold deposit on a $/oz basis does not necessarily increase with increasing ownership.
Commodity price risk – the price of a gold deposit changes disproportionally to the prevailing market price for gold metal at the time of the transaction.
Certainty risk – the price of a gold deposit increases disproportionally to increases in the certainty of a deposit’s quantity and quality estimate.
Country risk – the prevailing risk tolerance of the market influences the impact of a jurisdiction’s systematic risk on the price of a gold deposit.
As the knowledge gaps being tested by the hypotheses are not described in the existing literature, they are investigated in isolation of each other. Future research may expand on the finding by undertaking simultaneous methodologies.
This thesis uses price, something that is directly observable in the market. Consequently, to address the hypotheses an empirical dataset was created from public-domain data concerning mineral asset transactions. These transactions are largely based on gold deposits, as gold mining does not rely heavily on support infrastructure, gold projects are frequently traded relative to other commodity types, and they require little downstream/offsite refining before sale into a terminal market. Identified transactions are manually entered into a database before categorisation, extraction and manipulation in each of the models. Information that is external to the transaction but relevant to the research is sourced from the public domain (e.g. risk indexes, commodity prices).
Market data are inherently noisy, and it is often necessary to simultaneously account for a number of variables to identify patterns. Sight is an important sense and plays a role in pattern recognition by being able to account for four variables [spatial location X-Y-Z, and magnitude (e.g. colour)]. By plotting the price of a mineral asset in an artificial space defined by its quantity (X), quality (Y), risk (Z) and magnitude ($), it can be visually interrogated and described using spatial statistics (geostatistics). To analyse a spatial dataset, semi-variograms are used to identify the nature of the relationship between the data when expressed in three-dimensional spaces. The information from the semi-variograms defines the shape and orientation of the search ellipse. The ellipse is used to apply weights of importance to data points that fall within its space, by considering the direction from which it comes as well as its distance from the centroid. The resulting value for the centroid is then assigned a volume for the cuboid (block). Unlike a simple matrix, (e.g. polygional modelling) ordinary kriging uses soft boundaries where information that falls outside a discrete block influences the estimation of the value contained within the block. The ordinary kriging process acknowledges that the relationships between data points are continuous . A block model is produced when numerous search ellipses are run to define contiguous or related blocks.
The quantity (tonnes, X-axis) and quality (grade, Y-axis) dimensions of the dataset distribute log-normally due to the natural underpinnings of the geological inputs. To enhance visual validation, the X and Y points are transformed prior to estimation. The transformation avoids the resulting model having an excessively planar shape (e.g. like a sheet of paper). No back-transformations are performed, alleviating the difficulty and risk associated with such operations. The block model’s results are also to the mean and median of the data in each block. The patterns within the block models are then used to describe the behaviour of the market variables being assessed. It is important to emphasise that the trends within the dataset are important in this research, not the magnitude of the estimated price. These trends are based on coarse 3x3x3 divisions (i.e. low, medium, high) that minimise the impact of noisy data and estimation nuances.
In this thesis, the Z-axis represents the dependent variable that underpins each of the hypotheses (e.g. commodity price, ownership, certainty and country risk). The ability to quantify the Z-axis variable determines its reliability in the block model method. While it is possible to determine a precise and reliable scale for the Z-axis in some applications, others are difficult to quantify and express. Consequently, the hypotheses are sub-divided into two main groupings that reflect the ability to measure their basis:
While it is possible to create a block model using any four variables, it does not mean that the mathematically correct outcome has a connection with reality. As a result, this research is conducted collaboratively with Specialists in geostatistical estimation. These Specialists were supplied with the relevant datasets, advised on how to distribute and transform the inputs and tasked with undertaking semivariogram analysis and the population of the block models. Different geostatistical experts were used in each hypothesis for the purpose accessing a breadth of knowledge, exposure to a wide range of geostatistical experience, as well as peer validation. The resulting block models were compared with polygonal estimates.
It is important to emphasise that this thesis is not about geostatistics, but is about observing price behaviour. The geostatistical methodology used is well established and does not add to the literature per se. Instead, the geostatistical method is used as a means of dynamically modelling statistical significance of any relationship between the variable. This geostatistical aspect serves to satisfy the statistical significance requirement of the hedonic pricing method. This leaves the focus of the thesis on
This research describes previously unidentified market behaviours, notably that small deposit transactions appear to have a different market behaviour to their larger equivalents; and that deep deposits are more expensive than their near surface equivalents which may be due to a perception of upside optionality. Furthermore, the empirical research yields observations that are not consistent with theories obtained from other fields of knowledge and applied to mineral assets (e.g. securities) or falsifies practices described in the informal literature. As there is little existing literature on mineral asset pricing, the maiden descriptions in this thesis are intended only to demonstrate what relationships exist. Specifically, the underlying data are inherently noisy and erratically populated and as such not suited to specific point-estimation. However, having established the nature of the relationships, this interdisciplinary research provides a building block for further expansion and refinement in the field of mineral asset pricing, which is surprisingly poorly described and has significant and direct commercial importance.