Int. J.

Production Economics 183 (2017) 14–20

Contents lists available at ScienceDirect

Int. J. Production Economics

journal homepage: www.elsevier.com/locate/ijpe

On the cost of capital in inventory models with deterministic demand

a,b a,c,⁎ d,e
crossmark
Alejandro Serrano , Rogelio Oliva , Santiago Kraiselburd
a
MIT-Zaragoza International Logistics Program, Zaragoza Logistics Center, C/ Bari 55, Edificio Nayade 5 (PLAZA) 50197, Zaragoza, Spain
b
IESE Business School, Avenida Pearson 21, 08034 Barcelona, Spain
c
Mays Business School, Texas A & M University, College Station, TX 77843-4217, USA
d
INCAE Business School, Campus Walter Kissling Gam, La Garita, Alajuela, Costa Rica, 960-4050
e
Escuela de Negocios, Universidad Torcuato Di Tella, 1010 Calle Juan Pablo Senz Valiente, Buenos Aires, Argentina, C1428BIJ

A R T I C L E I N F O A BS T RAC T

Keywords: In the operations management literature, the financial risk in an inventory model is usually assumed to be
Supply chain management captured by the (constant) weighted average cost of capital (WACC) of the firm. This assumption is, at best, an
Risk approximation, since this cost depends on the risk of the cash flows, which, in turn, depends on the inventory
Operations management-finance link policy. We investigate what the right cost of capital should be in an inventory model with deterministic demand.
Cost of capital
To do so, we study an inventory model with a generic inventory cost function where risk depends on the
inventory decision made. Additive and multiplicative financial noise functions are included to assess the impact
of these on both the cost of capital of the firm and the optimal inventory policy. We find that, in contrast to
previous models, risk is not in general a monotone function of inventory. Also, a rate close to the risk-free rate,
which typically deviates significantly from the WACC, should be used to value inventory-related investments
when the inventory cost function is dominated by holding cost for large order quantities, even if investments are
subject to other sources of financial variability.

1. Introduction shareholders, i.e., maximize the sum of dividends discounted at the

According to a survey conducted by the Association for Financial
Professionals, 53 per cent of firms use the weighted average cost of
capital (WACC) to evaluate all projects and investments (Association
for Financial Professionals, 2011). Similarly, in contrast to the finance
literature (e.g., Brealey and Myers, 2003), the operations management
literature has traditionally assumed that financial risk in inventory
models can be captured by the (constant) WACC of the firm irrespec-
tive of the resulting cash flow risk (e.g., Chopra and Meindl, 2007,
Chap. 10.6), typically computed by the accounting department of a firm
(Nahmias, 2005, p. 190). This assumption is, at best, an approxima-
tion, since this cost depends on the risk of the cash flows, which, in
turn, is a function of the operational policies (Birge, 2015). More
specifically, inventory models tend to ignore all sources of variability
but the ones related to the operational realm of the firm under study,
such as demand, price, or cost variability. Furthermore, these models
typically ignore the effect of operational decisions on the financial risks
faced by shareholders. While this may be a plausible approach for
scenarios where investors are risk-neutral, that is not necessarily
appropriate for their risk-averse counterparts.
In broad terms, a firm's objective is to maximize value for its
shareholders' required cost of capital. If the shareholders' cost of
capital was indeed separate from operational decisions, then the
separation of responsibilities would be consistent with the firm's
objective: finance would concern itself with minimizing the cost of
capital and operations, using the cost of capital as a fixed parameter,
would attempt to maximize the profits for the firm. However, the cost
of capital depends on the firm's operational decisions. For example,
holding larger amounts of inventory entails larger working capital to be
financed, either from debt or equity, which may increase or decrease
risk. Thus, reducing costs is neither a necessary nor sufficient condition
for maximizing a firm's value: it may lead to increases in cash flows for
shareholders (which increase value) and risk (which decreases value).
As a consequence, reducing costs can cause risk to increase and could
potentially destroy shareholders' value.
It is true that if operational changes are relatively small or the
context of the firm is relatively insensitive to those changes (for
instance, if inventory and financial leverage are small) then the impact
on the firm's value may be negligible. However, even a small change in
the firm cost of capital due to operational changes may have a relatively
large impact on the firm value, due to the high sensitivity of value to
risk.

⁎
Corresponding author at: Mays Business School, Texas A & M University, College Station, TX 77843-4217, USA.
E-mail address: roliva@tamu.edu (R. Oliva).

http://dx.doi.org/10.1016/j.ijpe.2016.10.007
Received 9 March 2016; Received in revised form 4 August 2016; Accepted 4 October 2016
Available online 06 October 2016
0925-5273/ © 2016 Elsevier B.V. All rights reserved.
A. Serrano et al.
This paper addresses the question of what the impact of inventory
decisions on the firm's cost of capital is and, consequently, how optimal
inventories policies should change given that the cost of capital is
endogenous to the firm's operational decisions. The motivation for
these questions comes from an engagement, where a multinational firm
in the pharmaceutical industry asked one of the authors for advice on
how to compute the appropriate production batch size for one product
family, given that its demand was fairly stable. The firm's managers
agreed that a plausible model to use would be the economic order
quantity (EOQ) model, although were reluctant to accept that the
financial portion of the inventory holding cost in the denominator of
the EOQ formula should be the risk-free rate—as there is no risk in the
EOQ model, since the state of the practice in the company was to use
the firm's weighted average cost of capital (WACC).
In order to address these practical questions, we derived a model
where a firm seeks to maximize value for its shareholders by deciding
on the amount of inventory to hold. To set the inventory problem in a
business context, the cash flows obtained by the firm are subject to two
sources of financial noise, one additive and one multiplicative. One
portion of the additive noise function results from cash flows linked to
other investments undertaken by the firm, which are not related to the
product under consideration (as in Eeckhoudt et al., 1995). The other
portion of the additive noise function together with the multiplicative
noise function can be thought of as the convolution of financial noise
functions affecting the individual model parameters. Overall, the model
considers all potential cash flows and three sources of variability,
namely, market returns, additive, and multiplicative noise functions.
The model also includes a generic inventory cost function, which makes
the model versatile to be applicable in various settings. We derive an
expression for the firm's cost of capital as a function of inventory, and
find that it depends on all three sources of uncertainty. We also show
that risk is not monotone in the amount of inventory, which is in
contrast with previous results in the literature (Singhal, 1988).
Additionally, we characterize the optimal solution for inventory, which
depends on the multiplicative noise variability, but not on the additive
one.
The rest of the paper is organized as follows: After the literature
review in the next section, we describe the model and the relationships
between cost of capital and order quantity in Section 3. In Section 4, we
derive an expression for the firm's cost of capital and characterize the
solution for the optimal inventory problem. Section 5 illustrates the
usefulness of our findings in terms of value creation for one business
application. Finally, in Section 6 we draw our conclusions.
2. Literature review
The literature linking operations and finance has been growing in
the last two decades looking for more holistic solutions for firms'
problems (for a recent review of this literature, see Zhao and
Huchzermeier, 2015). Several authors in operations management have
attempted to integrate operational and financial decisions by addres-
sing wealth maximization problems (McDaniel, 1986; Porteus, 1986;
Park and Son, 1989; Zipkin, 2000; Sun and Queyranne, 2002; Soni
et al., 2006). These authors assume that the cost of capital is exogenous
to the problems posed. In contrast, we do acknowledge that the cost of
capital is endogenous to the operational problem.
A second research avenue includes papers that consider an en-
dogenous cost of capital for shareholders. Among them, Anvari (1987)
analyzes, using the Capital Asset Pricing Model—CAPM (Sharpe, 1964;
Lintner, 1965), an all-equity firm facing a newsvendor problem with
limited investment capacity. Singhal (1988) uses the CAPM to derive a
specific relation between the firm's cost of capital and inventory. Birge
and Zhang (1999) make use of risk-neutral option methods to
introduce risk into inventory models. Xu and Birge study a newsvendor
model subject to financial constraints and bankruptcy cost (2004) and
present a multistage stochastic program where decisions on inventory
15
Int. J. Production Economics 183 (2017) 14–20
levels, amount of loans, and dividend payouts are made simultaneously
(2006). Finally, Li and Arreola-Risa (2016) address the impact of
stochastic capacity on inventory policies and the firm's value. These
papers use risk-neutral equivalent approaches to discount dividends at
the risk-free rate, either by adjusting cash flows (Anvari, 1987; Singhal,
1988; Birge and Zhang, 1999; Li and Arreola-Risa, 2016), or by using a
risk-neutral equivalent demand function (Xu and Birge, 2004, 2006).
Our approach is more intuitive—although equally correct—in that we
discount dividends at the cost of capital for shareholders, keeping track
of the effects of the inventory decisions not only on the value of the
firm, but also on the cost of capital for shareholders. We can identify
and quantify the effects of the parameters that have an impact on the
cost of capital.
As far as the inventory policy is concerned, of these papers, Singhal
(1988) and Birge and Zhang (1999) are the closest to ours as we use
CAPM to quantify risk. Our work, however, departs from these papers
in two main ways. First, our model includes a generic inventory cost
function, financial leverage, as well as investments not related to
inventory. Second, in our formulation, risk comes from either exogen-
ous or endogenous financial shocks, represented by two financial noise
functions, one additive and one multiplicative. We are able to
determine under which conditions these sources of variability are
relevant for the inventory solution.
Finally, for the particular case of deterministic demand, some
papers include discounting (e.g., Trippi and Lewin, 1974; Klein
Haneveld and Teunter, 1998), and even endogenous discount rates
(Thorstenson, 1988), but they do not address the impact of both noise
functions. In these papers, if demand is deterministic, the discount rate
for cash flows becomes the risk-free rate. This is not necessarily the
case in our model, since a firm, even one facing deterministic demand,
is usually subject to random disruptions and, therefore, its risk-averse
rational stockholders may require an expected return above the risk-
free rate for their investment.
3. Model
Consider a profit-making firm that faces constant demand, d, buys a
product at cost c and sells it at price p. We focus on deterministic
demand to ensure we understand the effects of other sources of
variability. Practical examples include demand of items with fairly
stable consumption, such as, say, brake pedals for constant-production
car manufacturers, or demand in a market with a severe capacity
constraint, so that effective demand as seen by producers can be
considered as constant over time. A portion θ of the initial investment
used to buy the product comes from debt holders (e.g., a bank) at cost
rd, and the remaining portion, 1 − θ , from the firm's shareholders.
Additional funds required, other than for purchasing, may also come
from shareholders or debt holders (respectively, Eε and Dε ). The latter
funds may or may not be related to the product under consideration.
Specifically, we let Eεp (Eεp′) be the portion of Eε that is (is not) related
to the product. Dεp and Dεp′ are defined similarly. These additional
funds are included to study the impact they may have on risk and the
inventory policy.
There are three stationary sources of uncertainty: (1) a moderate
multiplicative financial noise function, ν, with mean μν = 1 and
standard deviation σν ; (2) an additive financial noise function, ε, with
mean με = 0 and standard deviation σε ; (3) the market returns function,
rm, with mean μm and standard deviation σm. The financial noise
functions are included to acknowledge the fact that the risk of a
realistic project cannot be zero, and these functions capture variability
from other model parameters apart from demand. Covariances are
denoted by σ with two subscripts. For instance, σνm (σεm ) represents the
covariance of the multiplicative (additive) noise function with the
market returns function. A portion εp of the additive noise function ε is
related to the product, whilst the remaining portion, εp′, is not, i.e., it
can be thought of as the convolution of the cash flows coming from all
A. Serrano et al.
other investments undertaken by the firm (Eeckhoudt et al., 1995). The
financial noise captures the variability that some of the model's
parameters may have in practice.
A generic strictly convex inventory cost function of class C2, l(q),
drives other expenses in addition to purchasing and loan costs and
taxes. This function includes (1) the linear non-financial portion of the
inventory holding cost, which dominates other costs within l(q) if
inventory is large and (2) other possible costs such as the ordering cost.
The firm maximizes the value of expected discounted cash flows for
its risk-averse shareholders,  V (q ), by optimally choosing the order
quantity q. Discrete discount is used, the discount rate being calculated
according to the CAPM model. Although this model is typically used in
operations management in models with stochastic demand, it can also
be applied to other environments. Birge (2015) discusses the use of
CAPM as a way to capture systematic risk in a number of settings,
including those with production rates uncorrelated to the market.
Note that, when discrete discount is used, the usual saw-tooth
shape of the inventory function of maximum height q is equivalent, as
far as the inventory holding cost is concerned, to a constant inventory
function of height q/2 , as changes in inventory level occur only within
inventory cycles. Using this equivalence, the initial outlay can be
written as
q order quantity, according to the expression
φ0 (q ) = (1 − θ ) c + Eε
2 (1)
and the future flows at the end of each cycle as
⎡ ⎛ q ⎞ ⎤
φt (q ) = ⎢ ( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd ⎥ ν + ε
⎣ ⎝ 2 ⎠ ⎦ (2)
including revenue, the generic inventory cost function, and interest
expenses.
First, note that Eq. (1) refers only to material acquisition and that
the generic inventory cost function (including, e.g., set-up costs) is
ignored. We ignore the generic inventory cost function at t=0 as a
matter of mathematical convenience, but including it does not change
the nature of our qualitative results and insights. Second, since sources
of funding affect the cost of capital, our model formulation allows for
diverse funding options, including a mix of equity and outstanding
debt, and so debt is used to finance purchases in every period. The
results of the non-levered model can be easily derived by making θ = 0 .
Finally, note that the multiplicative noise function affects the revenue,
inventory cost function and the interest expenses on the outstanding
debt. This noise function can be thought of as a convolution of
individual noise functions impacting any of the parameters in Eq.
(2)—except demand. We chose this representation as it makes the
model more parsimonious without affecting the insights derived from
the model.1 The objective function for the infinite-horizon model
implied by Eqs. (1) and (2) can be written as
⎡∞ φt (q ) ⎤
max  [V(q )] = −φ0 (q ) +  ⎢∑ ⎥ faster with q than the denominator, as the concavity of G(q) dominates
q ≥0 ⎢⎣ t =1 (1 + rê (q ))t ⎥⎦
where rê (q ) is the equity cost of capital.
Since the expected cash flows φt (q ) define a perpetuity, we can write
 [φt (q )] =  [φ (q )] for all t, and the equivalent two-period objective
function can be written as
1
max  [V(q )] = −φ0 (q ) +  [φ (q )]
q ≥0 rê (q )
where, according to the CAPM,
rê (q ) = r f + λcov(re (q ), rm ) re (q ) = φ (q )/ φ0 (q ) λ ≐ Pm / σm2 (3)
where rf is the risk-free rate and Pm the market risk premium.
1
We performed a number of numerical analysis as in Section 4.2.1 using noise
functions for individual parameters and found no or negligible differences.
16
Int. J. Production Economics 183 (2017) 14–20
Maximizing value in any period yields an identical solution, thus
adding up cash flows in period 2 yields
(P) max  [V(q )] = −φ0 (q ) rê (q ) +  [φ (q )]
q ≥0
This expression together with Eqs (1) to (3) define the maximization
problem.
4. Solution
Once the problem has been defined, we now focus on finding (1) the
firm's cost of capital as a function of the decision made, and (2) the
optimal inventory policy.
4.1. Cost of capital
As stated in problem (P), the firm's cost of capital depends on the
inventory decision made. The rational is that a change in the inventory
policy will necessarily change the interaction between shareholders'
and market returns, which has an impact on the risk perception of risk-
averse investors. The next proposition formalizes this relationship.
Proposition 1. The required equity cost of capital is a function of the
λ (G (q ) σνm + σεm )
rê (q ) = r f + q
(1 − θ ) c + Eε
2 (4)
where
⎛ q ⎞
G (q ) = ( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd
⎝ 2 ⎠
Proofs are shown in the appendix.
From this expression, it is apparent that the cost of capital is
endogenous to the inventory policy through G(q). The risk premium is
proportional to a weighted sum of the covariances of the market
returns with the multiplicative and additive noise functions, with
respective weights G(q) and unity. As expected, only the non-diversifi-
able risk is relevant. Inspection of G(q) reveals that the equity cost of
capital increases with the margin obtained. As the multiplicative noise
function ν has an impact on the cash flows, these are subject to higher
variability as they grow, resulting in a larger equity cost of capital.
Proposition 1 drives a notable property of the cost of capital.
Corollary 1. The cost of capital rê (q ) decreases for q sufficiently large.
One might expect risk, and hence cost of capital, to be an increasing
monotone function of inventory (e.g. Singhal, 1988). However, this is
not the case even if demand is deterministic, as the cost of capital
decreases for relatively large values of q. When the order quantity is
small, the numerator of the second term of rê (q ) in Proposition 1 grows
the linear function in the denominator. The intuition is that even
though the “amount of risk” (i.e., the numerator) may monotonically
increase with q, as q increases, risk may get “diluted” over a larger
initial investment.
Fig. 1 shows the cost of capital as a function of the order quantity
for different values of the noise functions variabilities, σν and σϵ . Even
at times when the equity cost is mathematically monotone, under
realistic parameter values, the cost of capital always decreases for
sufficiently large values of q.
4.2. Inventory policy
We turn now our attention to finding the optimal inventory policy.
Proposition 2. The optimal order quantity satisfies the first-order
condition
A. Serrano et al. Int. J. Production Economics 183 (2017) 14–20

Fig. 1. Equity cost of capital as a function of the order quantity for different values of σν and σϵ . Model parameters were set as: c=4, p=10, d=1,200, θ = 0.3, rd = r f = 0.03, Pm=0.06,
σm = 0.15, Eε = 500 , Dεp = 0 , ρνm = 0.2 , ρε, m = 0.4 , l (q ) = 600, 000/q + 0.2q . The choice of values for Pm an σm follows, respectively, from Fernández (2009) and Singhal (1988).

c⎡ (1 − θ ) r f ⎤ variation of ν does not exceed the unity, thus σν should be confined to

l′(q ) + ⎢θrd + ⎥=0
2⎣ 1 − λσνm ⎦ the interval [0, 1]; and ρνm will indeed belong to the interval [-1,1].
Now, for usual values of hm, rd, and rf, say 10%, 5% and 2%
If l(q) is invertible, respectively, the term in brackets for extreme cases (such as
ρνm = 1 − θ = 1) will only change by roughly 10% at most when varying
⎛ c⎡ (1 − θ ) r f ⎤ ⎞
q* = l′ −1 ⎜ − ⎢θrd + σν from 0 to 1. In contrast, for milder values (say, ρνm = 1 − θ = 0.5) the
⎥⎟
⎝ 2 ⎣ 1 − λσνm ⎦ ⎠ change in the right-hand side will not typically exceed a 1%.
In sum, neither the additive nor multiplicative noise functions have
a significant impact on the inventory policy.
Proposition 2 reveals that the order quantity decreases with σνm and
rd. This is so because the term in squared brackets in Proposition 2 is 4.2.1. Example: Economic order quantity
the financial portion of the marginal inventory holding cost. Other As noted, all results derived so far are valid for a generic inventory
things being equal, increasing σνm or rd leads to decreasing q so as to cost function. In order to gain further intuition and insights of the
reduce the average inventory. . results and explore the applicability of our results in practical settings,
this section addresses, as an example, the case of the EOQ model,
Remark 1. The optimal order quantity is independent of the additive Kd q
where l (q ) = q + c 2 h m (K is the ordering cost per batch). Specifically,
noise function.
This result can be explained in light of the definition of problem (P). from Proposition 2, the inventory policy for the EOQ model changes to
The additive noise function and the inventory in the objective function Corollary 2. If l (q ) =
Kd q
+ c 2 h m , the optimal order quantity that
q
only interact through addition, thus the former is irrelevant for the
solves (P) is
optimization problem. The intuition is that the additive noise function
impacts cash flows which correspond or are equivalent to investments 2Kd ∼ (1 − θ ) r f
q* = ∼ , where h f = θrd +
exogenous to the decision under scrutiny. Those cash flows are c (h m + h f ) 1 − λσνm
irrelevant as far as the inventory policy is concerned.

Remark 2. The impact of the multiplicative noise function, ν, on the

inventory policy is negligible for usual parameter values.Note that, The solution has the structure of the EOQ formula, but the financial
∼
for l(q) functions where the linear holding cost dominates for large portion of the inventory holding cost, h f , depends on the multiplicative
values of q, l(q) can be written as noise function. Fig. 2 shows the optimal order quantity as σνm varies.
q Note that the optimal inventory policy hardly changes as the multi-
l (q ) = ls (q ) + c h m plicative noise becomes more volatile. These results are consistent with
2
the observation made right after Remark 2.
where ls(q) is the generic function that dominates when q is small and This observation can shed some light on the state of the practice in
hm is the non-financial portion of inventory holding cost. operations management, where the constant firm's cost of capital
The first-order condition in Proposition 2 becomes (usually the WACC) is often used even when demand is deterministic.
The usual argument is that the EOQ model is normally applied in
c⎡ (1 − θ ) r f ⎤
situations where drivers other than demand and relevant costs are not
−ls′(q*) = ⎢h m + θrd + ⎥
2⎣ 1 − λσνm ⎦ deterministic, thus the risk-free rate should not be used. At the
where the right-hand side is the marginal cost of holding one more unit theoretical level, our model seems to support this argument in that
of inventory. the multiplicative noise function makes the financial portion of the
Consider first the denominator of the last term, 1 − λσνm , which inventory holding cost larger than the risk-free rate. However, at the
modifies the impact of the risk-free rate on q*. For typical market practical level, our experiments suggest that the impact of the noise
values for Pm and σm, say 0.06 and 0.15 respectively, and given the functions is negligible thus
definition of σνm , 2Kd
q* ≃
P P c (h m + r f + θ (rd − r f ))
1 − λσνm = 1 − m2 σν σm ρνm = 1 − m σν ρνm = 1 − 0.4·σν ρνm
σm σm
Furthermore, if rd=rf or θ = 0 , then the financial portion of the
where ρνm is the coefficient of correlation between the multiplicative inventory holding cost is almost exactly the risk-free rate.
financial noise function and the market returns. Likewise, if the noise functions are absent, there is no risk in the
Considering moderate values for σν , such that the coefficient of model thus ideally rd=rf, and

17
A. Serrano et al. Int. J. Production Economics 183 (2017) 14–20

Fig. 2. Optimal order quantity as a function of σν for different values of rf and ρνm . Model parameters were set as: θ = 0.3, rd=rf, hm=0.09, Pm=0.06, σm = 0.15, K=500, d=1,200, c=4.

Kd q
Remark 3. If l (q ) = q + c 2 h m and the noise functions are absent,
then the optimal order quantity that solves (P) is
2Kd
q* =
c (h m + r f )

This result may be somewhat surprising because of two reasons.

First, the financial portion of the inventory holding cost is the risk-free
rate, not the firm's WACC. But this should be the case, as this particular
version of the model exhibits no risk. Second, the solution of the
maximization problem exactly yields the well-known EOQ formula.
The operations management literature states that the EOQ formula is a ∼
good approximation of the solution of the corresponding value max- Fig. 3. Percentage error in order quantities as a function of WACC less h f . Model
parameters were set as: θ = 0.3, rd = r f = 0.03, hm=0.09, Pm=0.06, σm = 0.15, ρνm = 0.2 ,
imization problem (e.g. Zipkin, 2000; Beullens and Janssens, 2014).
σν = 0.25.
Note that our work does not contradict this statement, as we used
discrete discount rather than continuous discount. This observation
leads us to conclude that . the well-known robustness of the EOQ formula, the relative difference
is too large to be ignored. This result suggests that firms are grossly
Remark 4. The standard cost-minimization EOQ problem is understocking by overestimating the risk associated to holding cycle
equivalent to the corresponding value-maximization problem as far inventory.
as the inventory policy is concerned.
This finding reinforces the robustness of the standard inventory
6. Conclusions
problems in operations management, which are sometimes criticized
for not describing reality with more precision.
This work addresses the question of how cost of capital and
inventory decisions change given that the cost of capital is endogenous
5. Application
to the firm's decisions. Even if the issue has been addressed in the
operations management literature, we contribute by considering the
In this section we present one application of the analytical results
impact of noise functions, either additive or multiplicative, levered
derived in the previous sections, which illustrate that accurately
investments, as well as a generic cost function.
capturing risk may lead to relevant increases in value.
Instead of considering the firm's cost of capital for shareholders as a
Recall the example described in the introduction to motivate the
fixed constant, in our model it depends on the sources of variability,
problem. It refers to a division of a leading manufacturer in the
specifically on the covariance of these with the market returns, since all
pharmaceutical industry with sales of approximately $50b worldwide.
sources of variability may have a significant impact on the cost of
In the early steps of the manufacturing phase, a key point is defining
capital, and ignoring any of them may lead to significant value
the appropriate production lot size. Set-up costs are very large,
distortion.
typically measured in hundreds of thousands of dollars per changeover.
Interestingly, risk is found not to be in general a monotone function
Since demand for some product lines is fairly stable, the state of the
of inventory. As inventory increases, the amount of risk also does, but
practice in the firm is to use the EOQ formula to calculate the adequate
gets “diluted” over a larger amount of investment, thus the cost of
lot size. To compute EOQ, the financial part of inventory holding cost is
capital may either increase or decrease.
assumed to be the WACC of the firm (around 12%), a common practice
As for the inventory policy, we find that the additive noise function
in business. The percentage difference in order quantities when using
∼ is irrelevant to compute the optimal inventory level. Furthermore, even
h f as the financial portion of inventory holding cost versus the WACC
though the solution of the inventory problem depends on the multi-
can be measured as plicative noise function, the impact of this function on the solution is
q h∼f − qw h m + WACC very limited for usual values of the parameters.
Error(%) = = ∼ −1 Even though we make some more realistic assumptions about the
qw hm + hf
cost of capital and its mutual influence on ordering quantities/
∼
Fig. 3 shows this error as a function of the difference WACC less h f . inventory than the classical EOQ model, we are able to reach a solution
According to the previous section, the financial portion of the inventory that is as simple as the EOQ and can be applied without a big loss of
holding cost can be approximated by the risk-free rate. Given this, if the accuracy in most practical cases. Despite the simplicity of the approx-
WACC is 12%, then the percentage error is approximately 32%. Despite imate solution we found—it essentially involves just changing one

18
A. Serrano et al. Int. J. Production Economics 183 (2017) 14–20

parameter in the EOQ formula, namely the WACC for the risk free rate, deterministic/very low variability SKUs, allowing significantly more
our model suggests that companies should think differently about stock than the extant approach suggests.

Appendix A. Proofs of results

Proposition 1. According to the CAPM,

rê (q ) = r f + λcov(re (q ), rm )

where
⎛ ⎛⎡ ⎛ ⎞ ⎤ ⎞ ⎞
⎜ ⎜⎜ ⎢ ( p − c ) d − l (q ) − ⎜θc q + Dεp⎟ rd ⎥ ν + ε⎟⎟ ⎟
⎜ ⎝⎣ ⎝ 2 ⎠ ⎦ ⎠ ⎟ 1 ⎡ ⎛ q ⎞ ⎞ ⎤
cov(re (q ), rm ) = cov ⎜ q , rm⎟ = q ⎢ ( p − c ) d − l (q ) + ⎜θc + Dεp⎟ rd ⎟ σνm + σεm ⎥
⎜ (1 − θ ) c + Eε ⎢
⎟ (1 − θ ) c + Eε ⎣ ⎝ 2 ⎠ ⎠ ⎥⎦
⎜ 2 ⎟ 2
⎝ ⎠
q
Letting G (q ) = ( p − c ) d − (l (q ) + (θc 2 + Dεp ) rd ), the result follows.□
Corollary 1. The derivative of the required equity cost of capital is
⎛ ⎞
drê (q ) λ ⎜ ⎟
=− q ⎜ ( l ′( q ) + θcrd /2) σνm [(1 − θ ) cq /2 + Eε ]+( G ( q ) σνm + σεm )(1 − θ ) c /2⎟
dq [(1 − θ ) c + Eε]2 ⎜ ⎟
2 ⎝ ⎠

The function l(q) is concave thus l′(q ) is negative, and therefore the derivative is unrestricted in sign.□
Proposition 2. The objective function can be written as

⎡ ⎤ ⎡⎛⎡ ⎛ q ⎞ ⎤ ⎞⎤
q
 [V(q )] = −⎢ (1 − θ ) c + Eε ⎥ (r f + λcov(re (q ), rm )) +  ⎢ ⎜⎜ ⎢ ( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd ⎥ ν + ε⎟⎟ ⎥
⎣ 2 ⎦ ⎢⎣ ⎝ ⎣ ⎝ 2 ⎠ ⎦ ⎠ ⎦⎥

Developing the covariance term as in Proposition 1 yields

⎛ ⎛ ⎛ ⎛ q ⎞ ⎞ ⎞⎞
⎜ λcov ⎜⎜ ( p − c ) d − ⎜l (q ) + ⎜θc + Dεp⎟ rd ⎟ σνm + σεm⎟⎟ ⎟
⎡ ⎤⎜ ⎝ ⎝ ⎝ 2 ⎠ ⎠ ⎠⎟ ⎡⎛⎡ ⎛ q ⎞ ⎤ ⎞⎤
q
 [V(q )] = −⎢ (1 − θ ) c + Eε ⎥ ⎜r f + q ⎟ +  ⎢ ⎜⎜ ⎢ ( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd ⎥ ν + ε⎟⎟ ⎥
⎣ 2 ⎦⎜ (1 − θ ) c + Eε ⎟ ⎢⎣ ⎝ ⎣ ⎝ 2 ⎠ ⎦ ⎠ ⎥⎦
⎜ 2 ⎟
⎝ ⎠
⎞
⎟
⎡ q ⎤ ⎡⎛ ⎛ q ⎞ ⎞ ⎤ ⎛ q ⎞ ⎟
= −⎢ (1 − θ ) c + Eε ⎥ r f − λ ⎢ ⎜ ( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd ⎟ σνm + σεm ⎥ +( p − c ) d − l (q ) − ⎜θc + Dεp⎟ rd ⎟
⎣ 2 ⎦ ⎢⎣ ⎝ ⎝ 2 ⎠ ⎠ ⎥⎦ ⎝ 2 ⎠ ⎟
⎟⎟
⎠

where the latter equality uses the fact that the expected values of ν and ε are, respectively, 1 and 0.
The first-order condition is
−(1 − θ ) crf /2 + λσνm (l′(q*) + θcrd /2) − l′(q*) − θcrd /2 = 0

Working out,

c⎡ (1 − θ ) r f ⎤
l′(q ) + ⎢θrd + ⎥=0
2⎣ 1 − λσνm ⎦

The second-order derivative is clearly negative.□

References Hill, Boston.

Anvari, M., 1987. Optimality criteria and risk in inventory models: the case of the
newsboy problem. J. Oper. Res. Soc. 38 (7), 625–632.
Association for Financial Professionals, 2011. Current trends in estimating and applying
the cost of capital. Report.
Beullens, P., Janssens, G.K., 2014. Adapting inventory models for handling various
payment structures using net present value equivalence analysis. Int. J. Prod. Econ.
157, 190–200.
Birge, J., Zhang, R., 1999. Risk-neutral option pricing methods for adjusting constrained
cash flows. Eng. Econ. 44 (1), 36–49.
Birge, J.R., 2015. Operations and finance interactions. Manuf. Serv. Oper. Manag. 17 (1),
4–15.
Brealey, R.A., Myers, S.C., 2003. Principles of Corporate Finance 7th edition. Mc Graw
Chopra, S., Meindl, P., 2007. Supply Chain Management: strategy, Planning, and
Operation 3rd edition. Pearson, Upper Saddle River, N.J..
Eeckhoudt, L., Gollier, C., Schlesinger, H., 1995. The risk-averse (and prudent) newsboy.
Manag. Sci. 41 (5), 786–794.
Fernández, P., 2009. The equity premium in 150 textbooks. SSRN eLibrary, [〈http://
ssrn.com/paper=1148373〉.].
Klein Haneveld, W.K., Teunter, R.H., 1998. Effects of discounting and demand rate
variability on the EOQ. Int. J. Prod. Econ. 54 (2), 173–192.
Li, B., Arreola-Risa, A., 2016. Financial risk, inventory decision and process
improvement for a firm with random capacity, working paper, Mays Business School,
Texas A & M University.
Lintner, J., 1965. The valuation of risk assets and the selection of risky investments in
stock portfolios and capital budgets. Rev. Econ. Stat. 47 (1), 13–37.
McDaniel, W.R., 1986. The economic ordering quantity problem and wealth

19
A. Serrano et al. Int. J. Production Economics 183 (2017) 14–20

maximization. Financ. Rev. 21 (4), 527–536.
Nahmias, S., 2005. Production and Operations Analysis 5th edition. McGraw-Hill, New
York.
Park, C.S., Son, Y.K., 1989. The effect of discounting on inventory lot sizing models. Eng.
Costs Prod. Econ. 16 (1), 35–48.
Porteus, E., 1986. Investing in new parameter values in the discounted EOQ model. Nav.
Res. Logist. Q. 33 (1), 39–48.
Sharpe, W.F., 1964. Capital asset prices: a theory of market equilibrium under conditions
of risk. J. Financ. 19 (3), 425–442.
Singhal, V.R., 1988. Inventories, risk and the value of the firm. J. Manuf. Oper. Manag. 1,
4–43.
Soni, H., Gor, A.S., Shah, N.H., 2006. An EOQ model for progressive payment scheme
under DCF approach. Asia-Pac. J. Oper. Res. 23 (4), 509–524.
Sun, D., Queyranne, M., 2002. Production and inventory model using net present value.
Oper. Res. 50 (3), 528–537.
Thorstenson, A., 1988. Capital Costs in Inventory Models - A Discounted Cash Flow
Approach. Profil, Linköping, Sweden.
Trippi, R.R., Lewin, D.E., 1974. A present value formulation of the classical EOQ
problem. Decis. Sci. 5, 30–35.
Xu, X., Birge, J.R., 2004. Joint production and financing decisions: modeling and
analysis. SSRN eLibrary, [〈http://ssrn.com/paper=652562〉].
Xu, X., Birge, J.R., 2006. Equity valuation, production, and financial planning: a
stochastic programming approach. Nav. Res. Logist. 53 (7), 641–655.
Zhao, L., Huchzermeier, A., 2015. Operations-finance interface models: a literature
review and framework. Eur. J. Oper. Res. 244 (3), 905–917.
Zipkin, P.H., 2000. Foundations of Inventory Management. McGraw-Hill, Boston.

20