Capital resource budgeting is the way managers of publicly held companies fulfill their fiduciary Silver price vs. time, 1950-1995. responsibility to maximize the wealth of stockholders. Managers maximize shareholder wealth by allocating investment capital to the most productive of competing alter- native projects. Managers do this in a manner that maximizes the net present value of the firm's future cash flow. The techniques of cash flow analysis are the primary tools available for selecting a corporate portfolio of projects that meets or exceeds the rate of return expected by shareholders. The reliability of cash flow analysis hinges on the ability of managers to forecast the future values of individual components of cash flow. The accuracy and reliability of forecasting determine the soundness of capital investment decisions and, ultimately, determine the value of the firm. In other countries, tax rates may be less predictable and may change dramatically with political instability. Tax Components of cash flow policy also enters the cash flow equation when noncash depreciation and depletion are allowable expenses. Net cash flow equals revenue less the cash costs of In the United States, the deductibility of interest production, taxes, administration and investment. Al- paid on borrowed funds influences the overall capital though the concept of cash flow is simple, in practice it structure of most corporations by reducing the relative can become complicated. Modeling, or forecasting, fu- cost of debt vs. equity capital. The use of tax-deductible ture net cash flows, to compare alternative investments, debt as a vehicle for leveraged buyouts of firms with ex- requires the development of independent projections of cess borrowing capacity has resulted in some of the future value of the various components that make history's most spectacular mergers and acquisitions. up net cash flow. Individual components are interrelated. But they also vary independently with Capital budget production quantity, time, technology, market structure and other exogenous variables. as the business plan Production revenue is the quantity of units of pro- duction times the average price of an individual unit. The success of any business is Total revenue depends on unit price (marginal revenue) determined by the firm's ability to and selection of the optimal production scale for a firm. develop and execute a realistic busi- Daniel K. Dysinger For most mineral commodities, profits are maximized ness plan. As the primary compo- is director, The where the marginal cost of producing one more unit just nent of the business plan, the Center for Advanced equals the marginal revenue, or price. Revenue maximi- capital budget summarizes the expectations of the firm for revenue, Mineral Processing, zation is rarely consistent with maximizing profits. The firm's production cost function varies in a non- costs, taxes, growth and market Montana Tech of the linear manner with production quantity. Therefore, de- share and the required capital University of velopment of accurate models of the various production investment to attain the planned Montana, Butte, MT cost functions is important for generating reliable cash business activity. The expected flow estimates. An accurate cost function analysis is profitability of the firm is reflected in 59101-0991. also fundamental for determining the optimum output the capital budget. And the level of of the firm. The amount and types of taxes are a understanding that a firm has of its function of governmental philosophy. In the United business is reflected in the accuracy States, taxes are relatively stable and predictable and and execution of the capital budget. are based on profitability
MINING ENGINEERING SEPTEMBER 1997 35
The firm's cost of capital also depends on the accu- The product, or output, developed in this process racy and execution of the business plan and capital bud- is pure information. The informational value is deter- get. Each lender and investor to a firm has an expecta- mined by the future correlation of the model tion of return on the funds they risk. Firms that deliver predictions with the operating and financial results of predictable and stable returns to investors pay less for the final physical asset. On the strength of this capital. Firms with unpredictable returns pay more be- argument, the market value of the firm is ultimately cause investors demand a higher risk premium. equal to the value of the information developed in the Therefore, the discount rate, or weighted average capital budgeting process. cost of capital, is related to the firm's ability to forecast Accuracy and precision in capital budgeting is one and manage its capital budget. Failure of the capital of the most important functions of management. Ulti- budget through sloppy or inaccurate forecasting will re- mately, the net present value of the company is deter- sult in a real cost to the firm in the form of higher costs mined by the value of the information contained in the of capital or bankruptcy. discounted cash flows of the capital budgeting process. Management can increase the value of the company by Capital budgeting process refining techniques to anticipate the market price structure, operating costs, general and administrative Capital budgeting in the resource business involves costs, taxes and other components that determine a predictable and systematic approach. The process be- future cash flows. gins with an estimate of the inventory of units available Decay in the value of information with time for production based on exploration, whether a new ore Investment of funds in mineral production is long term body or an existing mine. and risky. Capital requirements for greenfield in- The next phase is based on developing an engi- vestments can exceed several hundred million dollars. neered system for extraction and processing of the valu- The investment horizons can extend from 10 to 30 able mineral and disposal of the waste. In the engineer- years. Capital budgeting for mineral investment, ing phase, a conceptual strategy for developing the prop- therefore, requires forecasting the components of cash erty is determined. It integrates the best practices of the flow far into the future. day with the physical realities of the site location, sup- The reliability of any forecast declines with time. port infrastructure, transportation, work force, etc. Capital markets reflect the increasing uncertainty over Alternative extraction and processing strategies are time by demanding a higher rate of return to assume screened. These identify fatal flaws and determine a long-term risk. The yield on government bonds, working universe of potential technologies and operating society's proxy for a risk-free asset, increases with time. plans. Metallurgical treatment schemes, conceptual mine This reflects the risk premium demanded by investors plans and order-of-magnitude capital budgets are for assuming the time risk. The risks associated with prepared. These further refine the information required uncertain future states of the world are incorporated to make the ultimate decision: accept, reject or defer. into capital budgets through the market-determined At each stage in the process, many assumptions, discount rate. rules of thumb and estimates are required to develop a Designers, engineers and financial analysts are mathematically based picture of what an operating mine aware that forecast accuracy varies between the indi- will look like. These parameters are based on vendor vidual components of cash flow. Equipment prices, la- quotes, historical and experiential data, scale factors and bor, and environmental costs and metal prices are less other engineering methods to develop the most robust certain next year or five years from now. engineering model possible. The accuracy of the model In addition, the accuracy of forecasting individual depends on the talent and knowledge of the estimating components is not necessarily equal. For instance, oper- team and the inherent bias of the organization. ating cost estimates may be more accurate than com modity price estimates. Because the FIG. 2 internal uncertainty in capital bud- geting models is determined by the Linear regression of inflation at T vs. Inflation at T-1. cumulative uncertainty of the indi- vidual components, each component must be scrutinized equally. Forecasting techniques Theoretical concepts can explain the importance of minimizing the forecast variance. But they do not provide the how. No one technique has ever been shown to exactly predict the future. The best that can be expected of financial forecasting is to quantify the probability of occurrence of potential future states of the economy and the impact on the future cash flow of the firm. The most useful forecasts are unbiased and quantitatively repre-
2 SEPTEMBER 1997 ... MINING ENGINEERING
sent the downside investment risk. FIG. 3 Returns above budgeted, due to un expectedly strong market Markov inflation model vs. actual data. conditions or other random events, are a pleas~uit surprise that may increase the value of the firm. The impact of below- expected returns can potentially result in bankruptcy or other financial reversal that limits the firm's ability to operate. For this reason, forecasting techniques that quantify the probability of downside risk, while positioning General price equilibrium of markets Many market price structures can be viewed from the aspect of a continuum. The price oscillates around a general, market-clearing equilibrium price for significant periods of time. Daily or weekly hiccups are not really predictable. the short run, they can be described FIG. 4 by a log-normal distribution of re Comparison of the multivariate model fit with actual silver prices vs. time. turns. This is the basis for valuing call and put options where the option premium is priced to cover the probability that the market price will exceed the strike price before the option expires. Over longer, or mid-run, periods, prices tend to follow identifiable trends that are not randomly distributed. Mid-run commodity price trends are linked to specific market circumstances. These in- clude overproduction, the onset of booms or recessions, listing as a toxically challenged material and acceleration or deceleration in the inflation rate. Price changes are not randomly distributed. But they are related to the elasticity and shifts of the supply and demand function. Understanding and planning for price changes during the mid-run is the short run. As time increases, fewer and fewer of the greater capability. internal and external variables that influence the market price are fixed. Four separable trend periods are identified in Fig. 1. In the long run, price prediction becomes even less It presents the yearly average silver price ($US/oz) from dependable. Mid-term trends reverse, new technology is 1950 through 1995. As a first approximation, the trends assimilated, governments and laws change radically, are 1950 to 1966, a time of relative price stability; 1967 wars occur and generations change. The probability of to 1978 - a period of steadily increasing silver price; correctly predicting prices and costs declines as time in- 1979 to 1980 - a period of speculation; and 1980 to 1995 creases. - a period of steadily declining prices. Silver has markets as both an industrial metal and as Silver market 1950-1995 a monetary substitute during inflationary periods. The nominal silver price has been found to be statistically The silver market provides an opportunity to test correlated with both the consumer price index (CPI) and some forecasting techniques. This analysis is not meant the rate of inflation. to be exhaustive. Many more refinements can be made Simulation models can be derived. They take advan- and several other techniques applied that may provide MINING ENGINEERING SEPTEMBER 1997 37 FIG. 5 generated inflationary paths, based on the statistical distribution of his- Simulated inflation and silver price. torical data, can be used to evaluate the sensitivity of capital budgeting T% " outcomes. For example, the simple 1950 to 1995 Markov-Monte Carlo simulation would have indicated that the high inflation rate and silver price of the late 1970s was not sustainable. Firms that invested money in silver production in the late 1970s and early 1980s, based on a market price of 32 cents/g ($10/oz), have likely lost a lot of money. Combining models
to simulate the future
A simple, multivariate model was constructed to correlate silver prices with several economic indicators. These included the rate of inflation, the nominal level of the CPI and a in the future under randomly generated states of the d 1970s and early 1980s. i bl t model A real-world dj would t f th also economy. incorporate variables such as the elasticities of supply A statistically significant linear correlation exists and demand, inventory levels, consumption between the silver price and the CPI. However, the price expenditures and other variables that have known data is poorly represented by a simple linear correlation. correlations with the silver price. The multivariate Other correlated factors must be included that explain analysis of historical data from 1915 to 1995 resulted in the obvious divergence of the data from the linear $/oz a functional model capable of predicting silver price with vs. CPI trend line. The addition of other factors is re- an adjusted RZ of 96 % for the period. The fit of the ferred to as multivariate modeling. It provides a better model is shown on Fig. 4, which compares the actual overall fit with the real world data. It also enhances the price data with the calculated data. ability to forecast the impact of future states of the The Markov-Monte Carlo inflation model and the world on the price of silver. multivariate model can be combined to simulate likely Multivariate regression analysis indicates that the future states of the economy and the potential effect on price of silver at any point in time also depends on the the future price of silver. The top line of Fig. 5 rate of change of the CPI or the level of inflation in the represents one potential scenario of inflation for the economy. The future cash flow from a silver producing next 25 years. It is derived from the Markov-Monte Carlo operation depends on the future price of silver. So fore- inflation model seeded with 1995 inflation data. The casts for capital budgeting must take potential inflation bottom line of Fig. 5 presents the projected silver price into account. based on the above potential inflation projections. Unfortunately, forecasting future inflation or defla- If the actual future values of inflation and price are the tion is difficult. Inflation is caused by several economic same as predicted, it will be accidental. However, the mechanisms, or decisions, that impact the management simulations represent unbiased projections of potential of growth of the money supply. From deficit spending states of the future. These are consistent with historical during times of war or to ease the effect of oil supply experience and conditions of market equilibrium. shocks, inflation can be traced to monetary factors. Multiple simulations can be generated to provide cases However, natural economic equilibrium constraints for sensitivity analysis, to identify fatal scenarios and function in the money markets, as in other markets. quantify the potential downside risk for capital bud- This limits the duration and magnitude of sustainable geting decision making. Worst- and best-case scenarios inflationary episodes. can be identified. Alternative planning for various situ- Figure 2 compares historical inflation data from ations can be stimulated. Unbiased estimates of future 1950 to 1995 (points) with a correlation generated from cash flow also can be provided. a Markovian process model (solid line) in which inflation All economic entities use financial forecasts to raise in period, t, is linearly correlated with inflation in capital, plan operations and make investments. Ulti- period, t-1. The Markov correlation can be combined mately, the value of the company is determined by the with a Monte Carlo procedure to develop conditional, value of the budget information and its correlation with random estimates of the future path of inflation. the actual state of the economy in the future. Figure 3 compares the actual inflation data and a Simulation can provide glimpses of the most likely Markov-Monte Carlo simulated inflation, seeded with future and quantify the probability of the occurrence of the 1949 inflation rate, from 1950 to 1995. The data individual states and the range of potential states. demonstrate that exact predictions of inflation are not Management must interpret the results, plan for possible. However, multiple simulations of randomly contingencies and maximize the net present value for the owners of the firm.