Professional Documents
Culture Documents
https://doi.org/10.1007/s13563-019-00213-3
ORIGINAL PAPER
Abstract
In the development of mining projects, there is a period in between commitment of finance and production commencement.
Risks are present in various aspects of development such that production start-up could be delayed, thereby affecting the
value of the project. This paper demonstrates the application of convolution in the capital investment problem for projects
with long lead-times as a means of characterising the relationship between capital invested and the materialisation of
capacity for production to begin. The system is functionally similar to that of a causal linear time-invariant system in signal
processing. We find that the filter for capital projects, affecting the rate at which capacity becomes available for use, can be
approximated by a straight-line function with positive gradient on bounded support contained in the positive time domain.
The deterministic version is derived and evaluated, and uncertainty is simulated in the stochastic version, where variance
results from the presence of a random process in the form of a standard Brownian motion.
of continuous-time problems, and the convolution sum for development with a focus on the lead-time. Furthermore, the
discrete-time problems. The use of convolution in capital research presented in this paper establishes the link to signal
investment theory is also found in Jorgenson (1996) and processing, thereby making signal processing techniques
Kamien and Muller (1976). When considering linear time- available to solving variations on the lead-time problem in
invariant systems, the system is fed with an input signal f (t) project investment. The derived formulation is solved and
which, when convolved with the system’s impulse response the results validate the model. In addition, we expand on
h(t), produces an output signal y(t), i.e. the applications of the model and show how uncertainty in
lead-time can be accounted for.
The research presented in this paper is useful and
In situations where the input signal is known and an improvement on previous approaches in that we
the output signal can be measured, it is theoretically look specifically at the lead-time that exists during
possible to calculate the system’s impulse response, thereby the construction stage of the development phase. We
obtaining information about the system’s configuration. The formulate the problem such that techniques from signal
linear time-invariant filter is completely described by the processing, a well-established research area in a number
convolution of engineering disciplines, become available for solving
∞ this investment problem. The formulation could be used
y(t) = h(τ )x(t − τ )dτ . (1) for improving estimates of the onset of production and
−∞ the subsequent initiation of cash flow. This would support
A system that produces a signal only in response to an input accuracy in project valuation. In addition, our approach can
signal is termed a causal system, and in such a case the accommodate uncertainty by the introduction of a random
lower limit of integration in Eq. 1 is zero. The application of variable, thereby making it viable for use in studies of
convolution in signal processing is discussed in texts such project risk.
as Mitra and Kaiser (1993), Haykin and Van Veen (2005), Our research is applicable to “completion risk” in
and Carlson (1992). particular, or the possibility that the production capacity
Thus, for a mining project in particular, one might think for a mining project might not be available for utilisation
of the capital injections as the input signal, the physical as planned. This may be due to delays in construction,
mine development process as the impulse response, and the costs overruns requiring additional funding, and engineering
rate at which capacity comes on-line as the output signal. It flaws that require redesign (Park and Matunhire 2011). It
follows that since both the rate at which capital is injected follows that our model relates to aspects of the design of the
and the rate at which capacity comes on-line are known mine and its construction, as opposed to the mine schedule2 .
or measurable, it becomes possible to gain insight into We begin by discussing literature on capital investment
the project’s characteristics and the underlying mechanisms theories, investment in mining projects, and project risk
governing the project’s development. Moreover, techniques evaluation. This is followed by the derivation of the
developed for solving problems in signal processing are transformation of invested capital into physical production
equally valid for solving the mathematical problem of lead- capacity, thereby showing that this problem can indeed be
time in project investment. structured as a problem in signal processing as outlined
A similar formulation for capital investment appears above. We then show how, by introducing a random
as a term in the objective function of the coal mining variable to simulate the presence of uncertainty during the
optimisation presented in Rademeyer et al. (2019). There development phase, one might observe how this uncertainty
is however no motivation given for the viability of this would affect the rate at which the physical capacity reaches
formulation. completion and becomes available for utilisation. We also
The research presented in this paper contributes to the cover briefly how the formulation could be applied before
knowledge on investment in mining projects by fully deriving ending with concluding remarks.
the characterisation1 of the relationship between capital
invested and eventual production capacity that exists during
the construction phase of a project. This characterisation 2 Mine design versus mine plan or schedule is discussed in Savolainen
presents a novel means for studying investment in projects (2016). The former is the design of the layout of the mine and the
with long development lead-times, such as mine projects. operational flows. The latter is the plan for the extraction of the
To our knowledge, no earlier work discusses project orebody, the sequencing for the removal of individual units (smallest,
mining units, SMUs) by which the deposit is depleted, and the
areas of operational activity. Whereas the mine design is finalised,
and thus fixed, before mine construction begins, the mine plan is
1 The terms “characterisation”, “formulation” and “model” are used revised periodically and can be adjusted to support mine economics in
interchangeably throughout the paper. response to changing market conditions.
A characterisation of the mechanisms transforming capital investment...
input cost uncertainty, concerning construction factor prices It become available to the miner over the years until L,
and government regulations. That is, investment decisions effective capacity online Qt is given by
vary when confronted with costs over which the firm has
Qt = Qt−1 + α0 It + α1 It−1 + · · · + αL−1 It−L+1 , (2)
influence as compared with costs that are external to the
firm. where
In response to the growing importance of risk assessment
L−1
in project development, a couple of prominent risk αt = 1; αi ∈ [0, 1], i = 0, . . . , L − 1.
measurement techniques have emerged. One such method is t=0
by Monte Carlo simulation. Heuberger (2005) demonstrates Expanding Eq. 2 further yields an expression for the
how risk can be accounted for in the widely used net present effective capacity online at time t in terms of initial capacity
value (NPV) or discounted cash flow model for mine Q−1 and all preceding investments, that is,
evaluation by introducing a random variable to simulate
L−1
uncertainty. Another method is the real options valuation Qt = Q−1 + αs (I−s + · · · + It−s )
(ROV) which argues that the decision to invest in a project s=0
can be considered as a financial option instrument. The
so that
superiority of ROV when compared with the NPV approach
is discussed by Foo et al. (2018). ROV takes into account the
L−1
t
L−1 t
ability to delay investment until such a time that the project Qt = Q−1 + αs Iτ −s = Q−1 + αs Iτ −s .
will render the greatest value. That is, it accounts for project s=0 τ =0 s=0 τ =0
Fig. 1 Capital accumulation (blue bars) resulting from investment (red line, rescaled) over time
Fig. 2 Capital accumulation (blue bars) resulting from investment (red line, rescaled) over time and subjected to a constant rate of deterioration.
Capacity unaffected by deterioration is shown in the background (grey bars) for comparison
which, when accounting for age, would after t periods Case II: If 0 ≤ t ≤ L,
effectively be
A t 1 − (1 − rA )t
AQ2 (t) = AQ∗ + + .
k ln(1 − rA ) (ln(1 − rA ))2
L−1
(12)
(1 − rA )t dQt = (1 − rA )t αs It−s ,
s=0 Case III: If t > L,
where rA is the rate at which functionality is lost. Therefore, A −kL ekt − (1 − rA )t
the effective capacity in-place at time t is then AQ3 (t) = AQ2 (L) + Le . (13)
k ln(1 − rA ) − k
Results
A(Qt ) = (1 − rA )t+1 Q−1 + (1 − rA )t dQ0
+ · · · + (1 − rA )dQt−1 + dQt Figure 2 illustrates the accumulation of capital resulting
L−1
t from the investment profile AQ(t) given by Eqs. 11–13.
= (1−rA )t+1 Q−1 + αs (1−rA )t−τ Iτ −s , (9) The capital in-place is also affected by a constant rate
s=0 τ =0 of deterioration rA so that the capital accumulation curve
where A(Qt ) ≤ Qt , i.e. actual capacity in-place is no more exhibits a gentler incline before reaching its maximum as
than apparent capacity invested in. compared with the curve seen in Fig. 1. In addition, the
The continuous version of Eq. 9 is given by capital in-place begins to decline after its initial peak, unlike
the curve in Fig. 1, during the time when investment tapers,
thereby indicating that the tapering level of investment is
t L
inadequate to compensate for capacity loss.6 Additional
AQ(t) = AQ∗ + (1 − rA )t−τ α(s)I (τ − s)ds dτ .
0 0 investment would thus have to be undertaken to maintain a
(10) constant capacity level.
Using the result given for h(t) in Eq. 7 for the inner integral 6 In practice, plant deterioration becomes apparent as decreasing
of Eq. 10, we find that an expression for AQ(t) can be found efficiency. However, since efficiency is the ratio of output to input,
by combining the following three cases: a decrease in efficiency would have to result in a decrease in output
(or production) for a constant input. This would hold true for the
Case I: If t < 0, maximum output (as designed) as well. Then, since capacity is defined
as the maximum output per unit time, a decrease in efficiency per unit
AQ1 (t) = AQ∗ . (11) time results in a decrease in capacity.
A characterisation of the mechanisms transforming capital investment...
Capital accumulation subject to uncertain Rewriting ehs in its Maclaurin series representation, we
project delivery have
t ∞
hn t
The problem of capital accumulation where the progress Ws e ds =
hs
Ws s n ds,
on the delivery of capital invested in is uncertain can be 0 n! 0
n=0
investigated by restating the problem with the delivery
where the integral gives the nth moment of Wt .
determination function α(t) cast as a random process. A
comprehensive treatment of stochastic processes is provided
in text such as Øksendal (2003). Differentiating Eq. 4 with
Applications
respect to t yields the differential equation
L
dQ Models are useful because they give decision-makers an idea
= (α ∗ I )(t) = α(s)I (t − s)ds. about possible outcomes that may materialise. This section
dt 0
outlines applications where the formulations or models pre-
Suppose α(t) consists of a deterministic part r(t) and “white
sented in Sections “Deriving a model of capacity investment
noise”, i.e. αt = rt + aWt where 0 ≤ t ≤ L. Wt is
with lead-time” and “Capital investment and accumulation
a Brownian process satisfying the stochastic differential
with capacity-loss due to deterioration” could be used.
equation
The models would be useful to commodity analysts
dWt = μt dt + σt dBt , for simulating the lag between project investment and
commodity production in market studies. This is seen in
where Bt is a standard Brownian motion, i.e. Bt ∼ N(0, t), Rademeyer et al. (2019) where the model appears as a term
μ : Rn → Rn is the drift field and σ : Rn → Rn×m is the representing production capacity in the objective function
diffusion field. Then of the cash flow optimisation problem. In fact, it is the
L L
right-hand term of Eq. 3 in Section “Deriving a model of
dQt = rs It−s ds dt + a Ws It−s ds dt, capacity investment with lead-time” that is used. Rademeyer
0 0
et al. (2019) solve the cash flow optimisation and this yields
with It = I (t) as given in Eq. 5.
results which are consistent with what might be seen for
The solution to the first term was obtained in the
real-life mining operations.
“Deriving a model of capacity investment with lead-time”
When determining capacity requirements and accounting
section. For the solution to the second integral, there are
for delays, using the model, we can determine capacity to
again three cases to consider:
be invested in at the time an investment decision is made.
Case I: t < 0 The product of capacity invested in, and the price of that
t capacity, yields the investment value at the time of decision-
h1 (t) = [0]It−s ds = 0 making. The same model can be used to address both
−∞
expansion investment and replacement investment.
Case II: 0 ≤ t ≤ L Suppose total project capacity requirement is given by
0 t K, consisting of several distinct aspects such that total
h2 (t) = [0]It−s ds + Ws It−s ds production capacity is given by
−∞ 0
Case III: t > L K= Kj .
0 L t j
h3 (t) = [0]It−s ds + Ws It−s ds + [0]It−s ds
−∞ 0 L
Then when taking into account lead-time from investment
until operational start-up, capacity to be invested in is given
In essence, we are required to evaluate
by
t t t
Ws It−s ds = Ws Aek(t−s) ds = Aekt Ws e−ks ds, I = F (K),
0 0 0
keeping in mind that A > 0 and k < 0. It is noted where F is the function given in Section “Deriving a model
that for our choice of function I (t), the integral is the of capacity investment with lead-time”. Capacity invested in
Laplace transform of Wt . Let −k = h > 0. The presence could be a sum of a series of investments in the lead-time
of the positive exponential results in increasingly unstable period [0 : T ] so that
behaviour for t → ∞, suggesting that variance increases
T
over time. That is, greater volatility in project completion is I= It .
associated with longer lead-times, as expected. t
Maryke C. Rademeyer et al.
The single investment or injection at time t would be given lead-time. The system is functionally similar to that of a
by It . Then, the value of that investment would be given by causal linear time-invariant system in signal processing.
It was found that the filter for capital projects can be
Vt = It ∗ Pt .
approximated by an increasing affine function on bounded
That is, the product of the capacity invested in It and its support contained in the positive time domain. Such a
price at time t as denoted by Pt . function convolved with a decreasing exponential integral
In this way, changes over time in procurement prices, representing capital investment over time provides a feasible
due to market cycles and dynamics, can be accommodated. model for investment and the resulting capital expansion
Economies of scale whereby procurement costs are in projects with long lead-times. The deterministic version
lowered as a result of increasing project size could was first evaluated, with consideration given for plant
also be accommodated here by reformulating Pt as a deterioration due to aging. In addition, a stochastic version
function of total production capacity K. Mine size and of the problem was considered by introducing a random
the heterogeneity of ore material should be accounted process in the form of a standard Brownian motion for
for when determining production capacity requirements.7 simulating uncertainty. The variation introduced was found
Capacity to be invested in and lead-times would vary to increase exponentially with time, resulting in rapidly
depending on whether the investment is for replacement or increasing variability in situations where lead-times are
expansion. Extending the model over the production time large.
horizon would enable further investigation into the trade- Our proposed approach lays a basis upon which layers
off between investing in technology to expand or maintain of complexity seen in mine development can be built.
production capacity, and the cost savings that result. Modifications can be introduced towards establishing a
Risk analysis supports decision-making by indicating more representative model of projects with long lead-times.
what the outcomes from a decision taken might be, and It is expected that the functional form for the lead-time
how likely the outcomes are to deviate from the expected filter will change when accounting for particular aspects
outcome or average. In practice, one might have a model of project development, such as logistics, procurement and
with a random variable as input and one would then politics. Future work should be concerned with determining
observe results as a range of possible outcomes. This has viable candidates for lead-time filters. For instance, it would
been demonstrated analytically by the use of a geometric be useful to know what lead-time filter would render a
Brownian motion as input in the model in Section “Capital step-wise output signal such as would more realistically
accumulation subject to uncertain project delivery”. Monte represent how production capacity becomes operational,
Carlo simulation for modelling project risk is demonstrated i.e. by step-wise increases in production. It would also
by Heuberger (2005). be useful to investigate scenarios with functionality loss
Determining the most appropriate filter, as in Eq. 1 in as determined by the rate of capacity utilisation, as well
Section “Introduction”, for the lead-time in mining project as the form of the filter in the case where investment
development could be done on a trial-and-error basis. That cost is proportional to rate of investment. The implications
is, by using financial investment data from actual projects, for investment in expansion technology versus investment
convolving with a trial filter (as shown in Section “Deriving in replacement technology by means of varying capacity
a model of capacity investment with lead-time”), and then requirements in the model present another area of future
comparing the resulting production capacity with actual research.
outcomes. Once suitable filters or “transformation” profiles We believe that the formulation presented in
are determined, then the model can be used to estimate Section “Capital accumulation subject to uncertain project
capital expenditure under uncertain scenarios. delivery” provides insight into how uncertainty affects the
mechanism at work during the time-to-build lead-time. This
should support the structuring of more cost-effective risk
Concluding remarks mitigation measures. Future work should investigate how
structuring financial injections or the use of appropriate
We have demonstrated the application of convolution in financial instruments could be used to manage this risk and
the capital investment problem for a project with a long to compensate for delayed income resulting from a delayed
start in operations.
7 The determination of the capacity required to exploit the heteroge-
We acknowledge that in reality, industrial production
nous orebody at every stage of mine-life is a very complex process. capacity is either available for use or not, and that the
The required production capacity is non-uniform as well. To deal with
investment in capacity composed of several separate parts, one could
proposed gradual availability of production capacity is not
construct a model as the sum of the models as in Section “Deriving a entirely realistic. In mining for instance, a new mine would
model of capacity investment with lead-time” for each individual part. ramp up production from new capacity over time from when
A characterisation of the mechanisms transforming capital investment...
the capacity first becomes available for use. However, the Jorgenson DW (1963) Capital theory and investment behaviour. The
model is still useful when the production capacity yielded Am Econ Rev 53:247–259
Jorgenson DW (1996) Investment: capital theory and investment
by solving the problem in Section “Deriving a model of
behaviour, vol 1. MIT Press, Cambridge
capacity investment with lead-time” is taken to be level Jorgenson DW, Stephenson JA (1967) The time structure of investment
of completion and production begins only once capacity behaviour in United States manufacturing. Rev Econ Stat 49:16–
reaches a certain level, thereby achieving the binary start-up 27
Kamien MI, Muller E (1976) Optimal control with integral state
seen in mining projects in reality.
equations. Rev Econ Stud 43:469–473
Larsson V, Ericsson M (2014) E&MJ?s Annual Survey of Global
Acknowledgements The authors would like to thank Prof. Bruce A. Metal-mining Investment. Engineering & Mining Journal, January
Watson (The University of the Witwatersrand) for his many valuable 2014, 26–31
suggestions. The authors would like to thank the reviewers for their Mitra S. K., Kaiser J. F. (eds) (1993) Handbook for Digital Signal
comments and suggestions. Processing. Wiley, New York
Nourali H, Osanloo M (2019) Mining capital cost estimation using
support vector regression (SVR). Resources Policy 62:527–
540
References Øksendal B (2003) Stochastic differential equations - an introduction
with applications, 6th edn. Springer, Berlin
Abel AB (1983) Optimal investment under uncertainty. The Am Econ Park S-J, Matunhire II (2011) Investigation of factors influencing
Rev 73:228–233 the determination of discount rate in the economic evaluation of
Bradley PG (1985) Has the ‘Economics of Exhaustible Resources’ mineral development projects. The Journal of The South African
advanced the economics of mining? In: Scott A (ed) Progress Institute of Mining and Metallurgy 111:773–779
in natural resource economics, essays in resource analysis by Pindyck RS (1993) Investments of uncertain cost. J Financ Econ
members of the programme in natural resource economics (PNRE) 34:53–76
at the University of British Columbia. Claredon Press, pp 317–333 Rademeyer MC, Minnitt RCA, Falcon RMS (2018) Revisiting
Carlson GE (1992) Signal and linear system analysis. Houghton operating costs in resource extraction industries. Resources Policy
Mifflin Co., Boston 54:159–164
Costa Lima GA, Suslick SB (2006) Estimating the volatility of Rademeyer MC, Minnitt RCA, Falcon RMS (2019) A mathematical
mining projects considering price and operating cost uncertainties. optimisation approach to modelling the economics of a coal mine.
Resources Policy 31:86–94 Resources Policy 62:561–570
Eklund JE (2013) Theories of investment: a theoretical review Savolainen J (2016) Real options in metal mining project valuation: a
with empirical applications, working papers series from Swedish review of literature. Resources Policy 50:49–65
entrepreneurship forum, working paper no. 22 Teisberg EO (1993) Capital investment strategies under uncertain
Foo N, Bloch H, Ruhul S (2018) The optimisation rule for investment regulation. The RAND J Econ 24:591–604
in mining projects. Resources Policy 55:123–132 Tobin J (1969) A general equilibrium approach to monetary theory,
Gould JP (1968) Adjustment costs in the theory of investment of the journal of money. Credit and Banking 1:15–29
firm. Rev Econ Stud 35:47–55 Zhang K, Olawoyin R, Nieto A, Kleit AN (2017) Risk of commodity
Haykin S, Van Veen B (2005) Signals and systems, 2nd edn. Wiley, price, production cost and time-to-build in resource economics,
New York environment, development and sustainablity, pp 1–24
Heuberger R (2005) Risk analysis in the mining industry, SAIMM
Proceedings of the management of risk in the minerals industry,
February 2005, South African institute of mining and metallurgy, Publisher’s note Springer Nature remains neutral with regard to
pp 75–79 jurisdictional claims in published maps and institutional affiliations.