You are on page 1of 15

FORECASTING AND PERFORMANCE:

CONCEPTUALIZING FORECASTING MANAGEMENT


COMPETENCE AS A HIGHER-ORDER CONSTRUCT
TORSTEN DOERING AND NALLAN C. SURESH
State University of New York - Buffalo

This note empirically validates forecasting management competence


(FMC) as a higher-order construct based on four underlying sets of prac-
tices: internal integration, forecasting process quality, effective use of
advanced systems, and evaluation of forecasting. The results from a partial
least squares structural equation model support FMC as a significant dri-
ver for three outcomes: forecast accuracy, cost reduction, and delivery per-
formance, thereby establishing predictive validity. The study provides
insights to the relative effectiveness of the four sets of forecasting practices
constituting FMC. A mediator–moderator analysis shows that forecast
accuracy is not a mediator or a moderator between FMC and the other
two outcome variables of cost reduction and delivery performance. This
indicates that in addition to effective demand forecasting practices, other
processes such as sales and operations planning (S&OP) are also impor-
tant for translating the benefits of effective forecasting to realize favorable
business outcomes. A multigroup analysis shows that the impact of the
four practice elements on the three performance outcomes also differs
based on specific business contexts.

Keywords: forecasting management; human judgment and decision making; supply


chain performance; partial least squares; scale development; survey methods

INTRODUCTION Despite such valuable insights, the managerial pro-


Many organizations are still striving to make cesses required to organize the forecasting process for
demand forecasting more effective, despite years of eliciting information from various external and inter-
effort on the part of researchers and practitioners. It is nal sources, communicating them among various
widely known that demand forecasting drives strategic business functions, using the right forecasting meth-
and operational decisions such as identification of ods for the right contexts, and translating them via
market opportunities, planning of financial require- the sales and operations planning process into
ments, production and capacity planning, and inven- bottom-line advantages for the firm are yet to be
tory management (Fildes, Nikolopoulos, Crone & understood fully.
Syntetos, 2008; Makridakis, Wheelwright & Hyndman, More recently, forecasting has received attention at
1998; Moon, Mentzer & Smith, 2003). Given this the supply chain level, highlighting the benefits of
importance, forecasting methods have been studied sharing forecasts among partners via collaborative
extensively (Dalrymple, 1987; Kahn, 2002; Klassen & planning, forecasting, and replenishment (CPFR), and
Flores, 2001; Sanders & Manrodt, 1994), yielding vendor managed inventory (VMI). This broader per-
many useful insights. For instance, it has been found spective is seen in the framework shown in Table 1,
that simple forecasting methods such as exponential adapted from the work of Smith (2001). This frame-
smoothing can outperform more sophisticated ones; work distinguishes between forecasting techniques
the relative ranking of the performance of different and managerial processes on the one hand and fore-
methods depends on the accuracy measure being casting as an intrafirm or interfirm (supply chain)
used; and the accuracy of combined methods is process on the other. But even as forecasting is now
greater than that of individual methods on average being addressed at the supply chain level, forecasting
(Armstrong, 2001; Makridakis & Hibon, 2000). processes within firms continue to be problematic.

Xxxx 2016 1
Journal of Supply Chain Management

TABLE 1
Forecasting Methods and Management

II IV
Management

Intra-organizational integration, Inter-organizational integration. collaboration,


communication, collaboration, coordination, coordination, alignment, and forecast
internal information flow, employee information sharing, directly or via CPFR, VMI,
accountability, broad-based (team-based) extranet connectivity, etc. and other supply
approach and accountability chain initiatives
I III
Proper use of qualitative and quantitative Models for assessing impact of information
Methods

forecasting methods developed flow on supply chain (multi-echelon)


inventories, operations, transportation costs,
bullwhip effects, etc.
Individual Organization Supply Chain

Note: Adapted from Smith (2001).

Thus, the focus of our study is on quadrant II in process quality; (3) systems employed for forecasting
Table 1. (e.g., hardware and software); and (4) performance
Several contributions are made by this study. First, measurement (i.e., evaluation of forecast accuracy and
we advance the notion of forecasting management as updating). Based on a substantial number of case
a competence, in the form of a higher-order construct. studies, Mentzer and Moon (2005) also categorized
In past studies, four underlying groups of forecasting the level of sophistication on each of these four prac-
practices have been prescribed as essential compo- tices into four stages. Several authors have pointed
nents of forecasting management (Mentzer & Moon, out the need for empirical support of these practices
2005; Moon et al., 2003). We empirically assess the (Fildes et al., 2003). Based on these four sets of prac-
construct validity of these prescribed practices as a tices, we propose a second-order FMC construct
reflective-formative hierarchical component model (Henceforth, all construct names are in capitals for
(HCM) (Becker, Klein & Wetzels, 2012). We draw ease of exposition). Table 2 summarizes these four
from resource advantage (R-A) theory to substantiate practice elements, along with the three dependent
the proposed forecasting management competence variables investigated in the current study.
(FMC). A survey-based investigation is conducted to The first element, Internal Integration, denotes the
establish the construct and predictive validity of FMC. extent of communication (i.e., forecast information
Given many intervening organizational processes sharing), collaboration, and coordination among the
between forecasting and operational outcomes, we internal functions, along with information exchange
then investigate if forecasting accuracy acts as a media- with external sources such as customers and suppliers.
tor or moderator between FMC and cost reduction Integration of the forecasting process, internal and
and delivery performance. In addition, we explore external collaboration, top management support, and
contingency effects through multigroup analyses to employee accountability are important elements of
better understand the relationships between forecast- forecasting management (Winklhofer, Diaman-
ing practices and performance outcomes. topoulos & Witt, 1996). A second aspect of Internal
Integration refers to the organizational processes
THEORETICAL DEVELOPMENT adopted for forecasting. It has been argued that
broader participation, accountability, and individual
Underlying Practices of Forecasting Processes contributions to forecasting across various functions
Four groups of practices have emerged from past will ensure greater acceptance and commitment to act
studies as being essential for successful forecasting effectively in order to realize operational advantages
management (Mentzer & Moon, 2005; Moon et al., from better forecasting. It has also been found that
2003). These four elements are as follows: (1) internal team-based approaches to forecasting can improve
integration among marketing, production, and other industry-level and long-term forecasts (Kahn & Ment-
relevant functions; (2) forecasting methods and zer, 1994). Thus, we hypothesize that organizational

2 Volume 0, Number 0
Forecasting Management Competence

TABLE 2
Forecasting Management Constructs

Construct Description
Internal integration Collaboration, communication, coordination for forecasting within firm, forecast
information flow within firm, broader participation, team-based approaches to
forecasting, accountability for forecasts
Forecasting process Proper selection and use of forecasting methods, use of top-down and
quality bottom-up approaches and their reconciliation, choice of relevant variables,
time horizons, level of sophistication of techniques used, training,
documentation of forecasting process, segmentation of products for differential
pursuit of forecast accuracy
Effective use of Shared and up-to-date information systems and technologies for forecasting,
advanced systems adequacy of hardware, and software platform for forecast information sharing
within the organization
Evaluation of Monitoring accuracy, relative impact of forecast errors for various product
forecasting categories, utilizing these for improvements; use of confidence intervals,
demand variabilities in addition to single-point forecasts
Accuracy Accuracy of forecasts, subjective, and objective measure
Cost reduction Impact of forecasting on operational costs such as Inventory, transportation, and
production costs
Delivery performance On-time deliveries, meeting delivery promises, and reactions to changing
customer demand

integration for forecasting is a key element of forecast- across organizational functions. Thus, we posit that the
ing management. This construct of Internal Integra- Effective Use of Advanced Systems serves as another
tion has been widely investigated in supply chain construct that is part of forecasting management.
management research (Leuschner, Rogers & Charvet, Fourth, the systematic Evaluation of Forecasting Per-
2013). All measurement items utilized in this study formance is another critical factor which may con-
are presented in Table S1. tribute to FMC. It denotes the extent to which a firm
The second practice construct pertains to the assesses not only the accuracy of its forecasts, but also
“methodological approach” (Mentzer & Moon, 2005). the relative performance impact of forecast errors as a
It denotes proper selection and use of forecasting basis to update the forecasting process.
methods given the wide range of techniques currently
available for forecasting, the use of top-down and bot- Forecasting Management Competence
tom-up forecasting processes and their reconciliation. Based on the four underlying sets of practices dis-
This also involves assigning different levels of impor- cussed above, we define forecasting management as a
tance for forecasting demand for various product second-order competence construct. Similar
classes, the level of sophistication of the techniques approaches have been previously adopted, for instance
used, and the level of training and documentation of in the development of a purchasing competence con-
forecasting procedures adopted. Emerging advanced struct based on a set of defining practices, as in Nara-
forecasting methods, data mining, and predictive ana- simhan, Jayaram and Carter (2001) and in a low-cost
lytics have served to introduce even greater sophistica- country sourcing competence by Kusaba, Moser and
tion and complexity in forecasting, adding to the Rodrigues (2011).
necessity for statistical and technical training of fore- The conceptualization of forecasting management as
casting personnel. We refer to this set of practices, a competence can be viewed through the lens of the
another first-order construct in our model, as Forecast- resource-based view (RBV) of the firm, with forecasting
ing Process Quality (FPQ). competence as a resource that is valuable, not widely
Third, the Effective Use of Advanced Systems can held, and hard to imitate or substitute (Barney, 1991).
enhance the accuracy and effectiveness of forecasts and More recently, R-A theory (Hunt, 2000; Hunt & Davis,
improve forecast information sharing, eventually lead- 2008; Hunt & Morgan, 1995) has been advanced as an
ing to better organizational performance. The use of extension to the RBV (Leuschner et al., 2013). R-A the-
computers and software for forecasting has increased ory has a downstream, market-oriented perspective,
over the years. The effective use of information systems stating that organizations pursue two distinct kinds of
may also facilitate better forecast information flow advantages: advantages in resources as well as

Xxxx 2016 3
Journal of Supply Chain Management

positional advantages in the marketplace. In R-A the- analyzed independently instead of conjointly in a
ory, resources are defined as tangible and intangible nomological network. The four forecasting practice
entities available to an organization that enable it to elements cover different aspects of forecasting man-
produce efficiently and/or effectively a market offering agement and are therefore formative (as opposed to
that has value for some market segment (Hunt & Davis, being reflective) in nature. In a formative model, the
2008). Entities available to a firm are classified as “re- directional arrows point from indicator variables to
sources” only when they enable a firm to produce a the construct, or from a construct to a higher-order
market offering that has value as determined by the construct. Each of the four practice elements repre-
market. A theoretical exposition of the defining ele- sents a distinct aspect of FMC. A high level of Internal
ments of R-A theory, in relation to the RBV, can be Integration, for example, does not necessarily mean
found in the recent works of Barney (2012) and Hunt that a high level of FPQ must be present. The four
and Davis (2012). As Hunt and Davis (2012) articu- sets of practices affect the level of forecasting manage-
late, with R-A theory, there is no fundamental depar- ment but are not conceptually interchangeable, which
ture from a RBV other than also requiring market- also indicates that FMC should be formative. In con-
driven determination of what constitutes a resource. A trast to the reflective mode, in which a construct
demand-side orientation in theories relevant to supply “causes” the indicators, the four forecasting practices
chain management is also discussed at length in Priem do not need to covary. In the ensuing analysis, we
and Swink (2012). attempt to study the individual impact of each fore-
Resource advantage theory rests on nine founda- casting practice element through an HCM. Thus,
tional premises, as articulated in Hunt (2000). Compe-
H1: FMC is a second-order construct which consists
tences are viewed as “higher-order resources” that are
of the four forecasting management practices.
“socially complex, interconnected combinations of
tangible basic resources (e.g., specific machinery), and
intangible basic resources (e.g., specific organizational Hypothesized Model Structure
policies and procedures, skills and knowledge of The overall hypothesized model structure relates
employees) that fit coherently together in a synergistic FMC, along with its underlying practices, to three per-
manner” (Hunt, 2000). In this light, forecasting man- formance variables: Accuracy, Cost Reduction, and
agement can be viewed as a competence, based on the Delivery Performance (Figure 1). The traditional out-
tenets of R-A theory. come variable in a forecasting context has been fore-
Higher-order constructs like FMC are increasingly cast accuracy with the common assumption that
used to reduce model complexity, to express and eval- higher accuracy will lead to better operational perfor-
uate multidimensional concepts, to match the level of mance. However, Danese and Kalchschmidt (2011a,
abstraction between predictor and criterion variables 2011b) could not find a direct connection between
and to increase the breadth of concepts and their gen- sophistication of forecasting techniques and accuracy
eralization (Edwards, 2001; Wetzels, Odekerken- and suggested other links.
Schroder & Van Oppen, 2009). In past work, the In addition to accuracy, operational cost and deliv-
influence of specific forecasting practices has been ery performance have also been used as proxies to

FIGURE 1
Reflective-Formative Hierarchical Component Model

Internal Accuracy
Q1 (6 items) Q5_1, Q6_1
Integration R2=0.376
-0.061
0.613***
Process
Q2 (4 items)
Quality 0.481***
Cost
FMC
0.713*** Reduction Q7 (4 items)
(HCM)
R2=0.508
Advanced 0.292**
Q3 (4 items)
Systems
0.584***
0.414***
Delivery
Q4 (4 items) Evaluation Performance Q8 (3 items)
R2=0.341

*p < .10; **p < .05; ***p < .01

4 Volume 0, Number 0
Forecasting Management Competence

measure forecasting performance (Davis & Mentzer, This occurs when there are volume discounts offered
2007). The outcome measure of Cost Reduction cap- by suppliers, through supply uncertainties, long and
tures the impact of forecasting on inventory, trans- erratic supplier lead times, increased material ordering
portation, and production costs, which can ultimately when material prices are on the rise, transportation
impact customer satisfaction. Operational costs should economies of scale, speculative advance ordering to
be reduced through better forecasting practices and ensure guaranteed material availability, etc. There are
collaboration (McCarthy & Golicic, 2002). The fore- thus numerous factors subsequent to forecasting that
casting exercise also provides opportunities for better may erode the favorable outcomes expected through
understanding of markets and customer behavior, improved forecast accuracy. Hence, a high level of
thus possibly directly improving costs and delivery forecast accuracy does not automatically guarantee
performance independent of improved forecast accu- superior operational performance.
racy. Therefore, we hypothesize that: Forecast Accuracy should therefore be explored as a
mediator between FMC and Cost Reduction, and
H2: FMC is positively related to Accuracy, Cost
between FMC and Delivery Performance. Smith and
Reduction, and Delivery Performance.
Mentzer (2010) found evidence of a significant rela-
In addition, the possible relationships between tionship between forecast accuracy and logistics per-
Accuracy and the other two dependent variables of formance. Zotteri and Kalchschmidt (2007) found
Cost Reduction and Delivery Performance should be that forecasting had an impact on company perfor-
explored. It is often assumed that a high level of fore- mance. However, Danese and Kalchschmidt (2011a,
cast accuracy will automatically result in favorable 2011b) could not confirm forecast accuracy as a medi-
business outcomes such as lower costs and improved ator between forecasting variables and operational
delivery performance. However, the outcomes of fore- performance. The absence of a mediation effect was
casting also depend on other processes and interfunc- explained by the fact that similar levels of forecast
tional integration as part of sales and operations accuracy can have different impacts on operational
planning (S&OP). To culminate in favorable business performance, as the internal environment is difficult
outcomes, forecasting has to overcome the varied to control (Kalchschmidt, 2012). Thus, given limited
interests, differing priorities, and incentives of func- past research on mediation effects, and based on con-
tions in matching demand and supply (Oliva & Wat- flicting findings in past research, we put forth:
son, 2009, 2011). The sales function’s main goal is
H3a: Forecast Accuracy mediates the relationship
typically linked to targets which are determined inter-
between FMC and Cost Reduction, and
nally as well as at the organizational level. The finance
between FMC and Delivery Performance.
function strives to meet revenue targets and adequate
returns on investment besides managing cash flow Among the four sets of FMC practices, Internal Inte-
and financial viability. As part of the budgeting pro- gration and Evaluation of Forecasting can have direct
cess, forecasts may end up becoming sales and rev- effects on the outcomes of Cost Reduction and Deliv-
enue targets. The operations function manages the ery Performance without necessarily depending on
complexities of the production and supply network Forecast Accuracy. These practices may directly
striving to meet demand and delivery dates and improve the efficacy of the S&OP function, thereby
ensures adequate quality levels while keeping costs resulting in positive operational outcomes. A struc-
and inventories under control. It is also common that tured forecasting process may also induce better cus-
the operations function may be limited in its deci- tomer orientation, and more effective planning and
sion-making capabilities due to a cognitive burden, implementation. Thus, given possible direct effects on
resulting in nonoptimal replenishment decisions the outcomes of Cost Reduction and Delivery Perfor-
(Sterman, 1989). In the translation of demand fore- mance, it is conceivable that Forecast Accuracy may
casts into viable production plans, the S&OP process have a moderating effect, causing a differential impact
often involves production smoothing which results in of FMC on Cost Reduction and Delivery Performance.
buffer inventories. In addition, data pertaining to fore- Hence, we posit:
casts may be inaccurate, incomplete, or unavailable.
H3b: Forecast Accuracy moderates the relationship
Given these disconnects, the operations function often
between FMC and Cost Reduction, and
generates its own forecasts. Marketing may also sec-
between FMC and Delivery Performance.
ond guess sales forecasts because they can be distorted
by promotions and incentive targets (Oliva & Watson,
2009, 2011). A significant part of inventory costs for Contingency Analysis
raw materials and outsourced parts depends on the Past research in operations management has shown
ordering and replenishment actions of purchasing that repeatedly that processes and practices may not be
may procure materials in excess of immediate needs. universally applicable and that they depend on the

Xxxx 2016 5
Journal of Supply Chain Management

specific context (Sousa & Voss, 2008). With respect to Invitations to participate in the study were sent based
the external context, demand volatility can be on job profiles in three waves via electronic messages
expected to have a bearing on the impact of FMC on in accordance with the methods prescribed by Dill-
performance. Higher demand volatility may increase man, Smyth and Christian (2009). Manufacturing
the impact of forecasting, necessitating contextual companies from sectors 31 to 33 of the 2012 NAICS
analysis, as pointed out by Kerkk€anen (2010). Internal classification were included in the sample. Out of the
conditions such as the manufacturing process can be 556 qualified individuals contacted, we received 102
classified into intermittent (job shop and batch) pro- complete surveys, corresponding to an effective
duction systems on the one hand and continuous response rate of 18 percent. The sample was checked
(repetitive and flow) process situations on the other. for obvious response patterns such as “straight-lining,”
It can be argued that the relative importance and numerical inconsistencies and for survey “speeders.”
impact of forecasting may be different for these con- Out of the 102 cases and the 32 initial Likert-scale
trasting environments. Another aspect meriting further items used for the structural model, only one value
investigation is the size of the firm. It may be argued was missing which was imputed via mean value
that smaller firms, more mindful of survival, may be replacement. A statistical power analysis was conducted
more market sensitive and may be superior on fore- to confirm the adequacy of the sample and effect
cast management practices. On the other hand, larger sizes.
firms may have better and more dedicated resources Firms ranged from small enterprises to well-known
by way of systems and people for forecast manage- public companies. The average number of employees
ment. Thus, there may be differential effects with in the data set was 16,865 (5 percent trimmed
respect to size as well. Lastly, the influence of forecast- mean = 10,580), with a standard deviation of 52,600
ing can be different depending on the manufacturing and a median of 2,000. A total of 38 percent of
context. Make-to-stock (MTS) systems are known to respondents were at the CEO, VP, or Director level,
be “push” systems driven primarily by forecasts, as with the remainder having relevant functions at the
opposed to the “pull systems,” that is, engineer-, managerial level. Some representative titles at the
make-, and assemble-to-order (ETO/MTO/ATO) sys- managerial level were Forecasting Manager, Sr.
tems. It can be argued that forecasting may be more Demand Planner, Sr. Business Analyst, and Demand
important for forecast-driven push systems as opposed and Supply Planning Manager. The respondents had
to pull systems. Danese and Kalchschmidt (2011b) been working for an average of 8 years in their cur-
suggest that the effect of forecast accuracy under ETO/ rent functions (Table 3).
MTO/ATO regimes may not be as important to To assess nonresponse bias, we first compared the
improve delivery performance. It also has been noted percentage of respondents in each NAICS two-digit
that managers in just-in-time settings place more group with the number of companies in North Amer-
emphasis on accurate forecasts (Wisner & Stanley, ica in each of these three categories (Census, 2010).
1994). This leads us to hypothesize that: There was no significant difference compared to the
sample distribution according to a goodness of fit test
H4: The impact of FMC on the three performance
(v2(2) = .7, p = .65). Nonresponse bias was also
outcomes is contingent on the contextual char-
acteristics of (1) demand volatility, (2) process
type and size, and (3) manufacturing strategy. TABLE 3
Sample Demographics

RESEARCH METHODS Number of NAICS


We employed a survey methodology to collect the Employees (2-digit) Position
data used to test the study’s hypotheses. The survey
Mean 16,865 31 18% CEO, 3%
design began with a thorough literature review to
President
identify scales used in past research. Newly created
Median 2,000 32 21% VP, 34%
scales were based on the theoretical foundations pre-
Director
sented earlier, following Churchill (1979). The survey
Min 20 33 61% Demand 63%
was conducted at the business unit level using a seven
Manager,
point Likert-type scale (Table S1). Sample data were
Analyst
collected from across the United States from con-
(On avg.
tributing members of relevant LinkedIn groups
8 years in
(APICS, IBF, IIF, S&OP/CPFR) with a focus on
function)
demand planning to specifically target senior level
Max 425,000
managers who were directly involved in forecasting.

6 Volume 0, Number 0
Forecasting Management Competence

tested using the procedure of Armstrong and Overton are above .5 for all the constructs, as shown in Table 4.
(1977). A series of two-sample t-tests of the first and Five measurement items marked with an asterisk in the
last 30 cases (on tenure, strategic focus, manufacturing Table S1 had loadings under .7 and were removed after
process, and strategy) did not show significant differ- considering the impact on content validity, resulting in
ences (p > .05). improved AVE scores.
Common method bias was tested using Harman’s Discriminant validity was tested with a matrix of
one-factor test. Principal component analysis without cross-loadings, and no violation was found. A more
rotation produced seven measurement items with conservative approach to assess discriminant validity
eigenvalues greater than one. As normality tests can is the Fornell–Larker criterion, comparing the con-
only provide limited guidance for the application of struct correlations with the square root of each con-
the nonparametric PLS-SEM method, skewness and struct’s AVE (Fornell & Larcker, 1981). As seen from
kurtosis of the responses were tested (Hair, Hult, Rin- the AVE values in Table 4, each construct shares more
gle & Sarstedt, 2013). The majority of the responses variance with its own measurement items than with
were within the recommended 1 to +1 range for other constructs. Recently, the heterotrait-monotrait
skewness and kurtosis. ratio of correlations (HTMT) was suggested as a more
stringent alternative to test for discriminant validity
(Henseler, Ringle & Sarstedt, 2015). All HTMT ratios
RESULTS were clearly under the .9 limit for all seven constructs
Partial Least Squares Structural Equation Modeling (not shown). Only one ratio, between FPQ and Evalu-
(PLS-SEM) was used to estimate the reflective-forma- ation of Forecasting, had a value of .875 and did not
tive HCM. PLS-SEM has been frequently used in oper- meet the more conservative .85 limit. All upper boot-
ations management research (Peng & Lai, 2012), and strapping confidence intervals were below one, how-
it offers several unique advantages (Chin & Newsted, ever, providing evidence of discriminant validity.
1999).
Structural Model
Measurement Model Following the recommendations of Becker et al.
We adopted the procedure of Hair et al. (2013) and (2012), we estimated the hierarchical model with
employed confirmatory tetrad analysis (CTA-PLS) as a repeated indicators in formative mode to achieve real-
supplemental empirical test to confirm the measure- istic parameter recovery. The path estimates of the
ment model specification. The reflective specification of forecasting practice constructs using this approach
the first-order constructs could be confirmed for all highlight their individual contributions to FMC and
four-first-order constructs. All common reliability mea- not their correlations.
sures are displayed in Table S1. The lowest composite The weights between the four forecasting practice
reliability (qc) is .844, and the highest value (for “oper- constructs and FMC are reported on the left-hand side
ational cost”) is .958, confirming the internal consis- of Figure 1. The path coefficient between Internal
tency of the constructs (Nunnally & Bernstein, 1994). Integration and FMC is not significant. However, the
Convergent validity was assessed through item loadings Internal Integration construct was not removed
and average variance extracted (AVE). The AVE values because: (1) formative constructs do typically

TABLE 4
Discriminant Validity (Fornell–Larcker Criterion)

Delivery Internal Cost Advanced Process


Accuracy Performance Evaluation Integration Reduction Systems Quality
Accuracy .856
Delivery .482 .834
performance
Evaluation .501 .490 .812
Internal .387 .360 .701 .809
Integration
Cost reduction .502 .586 .657 .553 .923
Advanced systems .477 .460 .551 .517 .433 .784
Process quality .554 .451 .747 .678 .652 .503 .842
Square root of average variance extracted values are printed in bold; nondiagonal elements are the correlations.

Xxxx 2016 7
Journal of Supply Chain Management

TABLE 5
Predictive Power of Hierarchical Component Model

Stone-
R2 (Full Geisser Q2 R2 Direct R2 HCM
Model) (Full Model) (per DV) (per DV)
Accuracy .376 .228 .384 .422
Cost reduction .508 .413 .501 .536
Delivery performance .341 .209 .309 .354

contribute with distinct content; (2) removing a for- without the intervening construct of FMC, are shown
mative construct may alter the meaning of a higher- in Table 6. The path coefficients of FPQ are significant
order construct; and (3) the relationship may prove at the .01 level for Accuracy and Cost Reduction. But
significant in a specific context. Internal Integration it is not found to be significant for Delivery Perfor-
has a loading of .705 (not shown) in regard to FMC mance, which can be explained by the fact that Deliv-
and is therefore important in an absolute sense (Hair ery Performance is dependent on the performance of
et al., 2013, p. 129). In addition, dropping this con- several other functions such as order processing and
struct did not increase the R2 values of the dependent S&OP systems. Ritzman and King (1993) also showed
constructs or their path coefficients. that Delivery Performance is influenced by order pro-
The second and fourth constructs, FPQ and Evalua- cessing systems, manufacturing flexibility, presence of
tion of Forecasting, have the strongest overall impact excess capacity and safety stocks, and distribution sys-
on FMC, with path coefficients of .481 (t = 3.60) and tems. These issues are expanded upon in the Discus-
.414 (t = 2.81), respectively. Thus, three of the four sion section.
constituent path coefficients, including Effective Use
of Advanced Systems with a coefficient of .292 Mediation and Moderation of Forecasting
(t = 2.43), are moderate to strong and significant. This Accuracy
indicates that the four sets of forecasting practices We adopted the procedure of Hair et al. (2013)
contribute to the overall notion of FMC, suggesting based on the earlier work by Preacher and Hayes
partial support for H1. Collinearity was not a problem (2004) to test for mediation effects by meeting three
with our formative HCM because all four variance conditions (Figure 2). The direct effect between FMC
inflation factors were below a value of 3.0. and Cost Reduction without the mediator is signifi-
The R2 values of all three dependent constructs cant, with a path coefficient of .737 (not shown). But
range from .341 to .508, indicating sufficient explana- the relationship between the possible mediator Accu-
tory power (Figure 1). Their path coefficients, with racy and Cost Reduction, a necessary but not suffi-
values between .584 and .713, are strong and signifi- cient condition, is not significant with a path
cant, indicating that a higher level of FMC is posi- coefficient of .118. Hence no mediation effect exists.
tively related to improved Accuracy, Cost Reduction, Next we tested the possible mediation effect of Accu-
and Delivery Performance, thereby supporting H2. racy on the FMC—Delivery Performance path. The
Table 5 shows the assessment for the predictive
power of the HCM. We utilized the cross-validated
TABLE 6
redundancy measure Q2 to assess predictive validity
via blindfolding (Geisser, 1975; Stone, 1974). Values Direct Effects (Path Coefficients) without
above zero confirm predictive relevance for all three Higher-Order Construct (FMC)
dependent constructs. Johnson, Rosen, Chang, Djurd-
jevic and Taing (2012) recommend substantiating the Cost Delivery
conceptual usefulness of higher-order constructs by Accuracy Reduction Performance
assessing their comparative predictive power. As seen Internal .102 .071 .086
in Table 5, the higher-order FMC construct improves integration
predictive power through higher R2 values in contrast Process .395*** .325*** .149
to the direct paths without the higher-order construct. quality
When higher-order constructs such as FMC are used, Advanced .263** .057 .257***
systems
the direct impact of constituent factors on the out-
Evaluation .141 .338** .311**
comes may not be visible. Thus, the direct effects of
the four constituent factors on the three outcomes, **p < .05; *** .01.

8 Volume 0, Number 0
Forecasting Management Competence

FIGURE 2
Mediation Analysis

Cost
Internal
Reduction
Integration
R2=0.518

-0.053
0.644***
Process 0.118
Quality 0.469***
FMC 0.594*** Accuracy
(HCM) R2=0.352
Advanced 0.295**
Systems
0.452*** 0.221*

0.415***
Delivery
Evaluation Performance
R2=0.372

*p < .10; **p < .05; ***p < .01

relationship between Accuracy and Delivery Perfor- tested simultaneously to model multiple goals, as
mance is marginally significant, with a path coefficient depicted in Figure 2. Testing mediation with Delivery
of .221. The size of the indirect effect is fairly small Performance or Cost Reduction separately had no sta-
(.131, .594 9 .221). The bootstrapping result for this tistically significant impact on these results. Thus, H3a
product is not significant and does not indicate medi- was not supported.
ation, particularly in combination with the low vari- The moderation effect of Accuracy was tested using
ance accounted for of 22 percent (Hair et al., 2013, p. the two-stage approach of Hair et al. (2013, p. 263)
224). It may be noted that the mediation effect was and Henseler and Chin (2010). This approach must

FIGURE 3
Moderation Analysis

Cost
Internal
Reduction
Integration
R2=0.552

-0.000
Process -0.088 0.125
0.643***
Quality 0.401***
FMC
Accuracy
(HCM)
Advanced 0.230*
Systems 0.442*** -0.063 0.237**

0.490***
Delivery
Evaluation Performance
R2=0.387

*p < .10; **p < .05; ***p < .01

Xxxx 2016 9
Journal of Supply Chain Management

be used if formative measurement models are invariance in Step 1 is given as we test an identical
involved. This model tests if the relationship between model for each of the group comparisons with the
FMC and Cost Reduction, and FMC and Delivery Per- same algorithm settings. In Step 2, the equality of
formance, depends on the level of the continuous composite scores (construct level) between groups can
variable Accuracy. The test does not confirm Accuracy be confirmed through a significance test of their corre-
as a moderator in our model. None of the interaction lation values. If invariance in Step 1 and 2 can be
effects are significant (Figure 3). Thus, H3b was not confirmed, we can assume partial measurement invari-
supported. The absence of mediation and moderation ance and are able to continue with the MGA. The
effects was consistent with the results of prior studies results show that correlations between composites
such as Ritzman and King (1993), as well as Danese across multigroup analyses I-III in Step 2 are invari-
and Kalchschmidt (2011a, 2011b) who could not ant, allowing us to proceed. We had to exclude a
confirm forecast accuracy as a mediator between fore- comparison of manufacturing strategies (MGA IV), as
casting variables and operational performance. partial invariance could not be established in Step 2.
Henseler’s robust nonparametric bias-corrected boot-
Heterogeneity and Multigroup Analysis strap technique for MGA was employed to identify sig-
We utilized multigroup analysis (MGA) to assess nificant differences in path coefficients. The results of
possible contingency effects. The split of the first this analysis are summarized in Table 7. The difference
multigroup analysis (MGA I) was related to high and in path coefficients between the models and their sig-
low demand volatility, based on responses to ques- nificance is reported; for example, a path coefficient of
tions IP38 and IP39 (Table S1). The cutoff point was .200 in the first model and a path coefficient of +.400
set at a maximum difference in month-to-month in the second model are reported as a change of +.600.
demand changes of 80 percent, which coincides with The overall results support the MGA approach and
a natural break in the visual distribution of the vari- present interesting findings. MGA I compares firms
able and is near the median. MGA II compares com- based on high and low demand volatility. The average
panies that have repetitive assembly or continuous increase in R2 value of 24.1 percent across all depen-
flow processes with companies with job shop or batch dent constructs compared to the main model is quite
processes (IP13, Table S1). MGA III contrasts compa- high. Splitting the data into two groups and running
nies with 2,000 or more employees (median) to smal- two separate models provides increased explanatory
ler companies. Finally, MGA IV was intended to power. It may be noted that the construct Internal
analyze predominant MTS companies to companies in Integration has now become significant. A significant
an ETO/MTO/ATO environment. difference in path coefficients of +.672 (p = .025)
A new three-step procedure for PLS-SEM models to between models was estimated, going from high to
test measurement invariance of composite models was low demand volatility. The impact of FMC on delivery
applied to each MGA as proposed by Henseler, Ringle performance is lower and significant in the context of
and Sarstedt (forthcoming). The configurational lower demand variability (Table 7).

TABLE 7
Multigroup Analysis Results

MGA I MGA II MGA III


Demand Volatility: Process: Repetitive => Size: Number of
Diff = Peak-Trough Intermittent Employees
Multigroup Diff:High (50) => Diff: Rep./Flow (48) => High (41) =>
Analysis Low (44)a Job/Batch (53) Low (39)
Int. Integer. => FMC +.672 (.025/.036)b
Proc. Quality => FMC
Adv. Systems => FMC .362 (.034/.079) .507 (.019/.042) .666 (.005/.008)
Evaluation => FMC
FMC => Accuracy
FMC => Cost Reduction
FMC => Delivery Perf. .208 (.032/.174) .211 (.004/.061)
Avg. R2 increase 24.1% 13.3% 29.5%
a
Sample size.
b
Difference in path coefficients, p-value (Henseler’s MGA/Welch-Satterthwaite).

10 Volume 0, Number 0
Forecasting Management Competence

MGA II contrasts firms producing with a repetitive Reduction. But there was no significant direct effect
and continuous flow process versus firms operating in on Delivery Performance, which may be influenced
an intermittent environment (job shop, batch). The by other factors such as the presence of adequate
increase in R2 value (13 percent) is modest, but signif- inventory and capacity buffers, and the efficiency of
icant. A significant difference in the Effective Use of order processing and logistics systems, as discussed
Advanced Systems–FMC path coefficients of .507 in Ritzman and King (1993).
(p = .019) between models was estimated, going from The second most important element of FMC was
repetitive and continuous flow to intermittent envi- Evaluation of Forecasting which includes monitoring
ronment. Thus, firms operating as intermittent systems the accuracy of forecasts, understanding the relative
do not show the same level of impact of Advanced operational impact of forecast errors for various pro-
Systems on FMC as firms operating in repetitive envi- duct categories, prioritizing them and utilizing this
ronments. information for improving future forecasts. The shar-
MGA III compares the effects between large and ing and evaluation of forecast performance on the
small firms. The average R2 increase of the three part of several departments is another element in this
dependent constructs is almost 30 percent, the highest category. The effort invested in developing forecast
among the three multigroup comparisons. There is performance measures is wasted if it is not utilized to
also a significant difference in the Advanced Systems– improve forecasting capability and forecasting climate
FMC path coefficients of .666 (p = .005), moving (Davis & Mentzer, 2007). But despite the obvious util-
from large to small companies. Hence, it may be con- ity of performance measurement, there is often resis-
cluded that the impact of the level of Advanced Sys- tance to using performance measures because they can
tems on FMC is greater in firms with more than 2,000 be used to assign blame rather than to address prob-
employees. The effect of FMC on Delivery Perfor- lems (Davis & Mentzer, 2007; Moon et al., 2003; Sha-
mance is also stronger for larger firms. piro, 1977). The direct effects in Table 6 show that
Evaluation of Forecasting has a significant association
with operational Cost Reduction and Delivery Perfor-
DISCUSSION mance but not with Accuracy. Given this strong and
Managers face numerous options in terms of the direct link between the evaluation of forecasts and
wide range of forecasting techniques that are now operational performance, organizations must find an
available, coupled with a wide variety of associated acceptable way to implement forecasting performance
management practices. This study provides support evaluations.
for grouping them parsimoniously into four action- The third important construct was the Effective Use
able sets of practices which constitute a broader con- of Advanced Systems, which refers to the use of up-to-
ceptualization of FMC. It was shown that this date information systems and technologies for inter-
competence directly affects the outcomes of Forecast nal and external sharing of forecast information. It
Accuracy, Cost Reduction, and Delivery Performance. includes ensuring adequate hardware and software
Among the four sets of practices, FPQ emerged as platforms for forecasting and for supporting the criti-
the most impactful dimension of FMC, indicating cal S&OP processes and interfunctional integration.
that managers should pay particular attention to the The direct effects between Advanced Systems and
constituent elements of this construct. Using a struc- Forecast Accuracy as well as Delivery Performance are
tured well-defined forecasting process and guidelines significant (Table 6). The Effective Use of Advanced
within the organization and using the right forecast- Systems does not seem to impact Cost Reduction
ing techniques for each category of items are partic- directly. The impact of Effective Use of Advanced Sys-
ularly important. Given the wide range of available tems on FMC was also found to depend on the con-
forecasting techniques, selecting proper forecasting text. This link was less strong in job shops and in
techniques for each class of items has been a chal- batch production environments, possibly due to lim-
lenge for practitioners. This requires proper training ited capital investments by smaller companies. In lar-
of forecasting personnel on statistical aspects and ger firms, a higher level of Effective Use of Advanced
correct use of forecasting methods. Other aspects Systems based on a generally higher level of invest-
pertaining to this set of practices can be seen in ments in information technologies appeared to pay
Table 2. The effective use of both top-down meth- off through a significant improvement in Delivery Per-
ods of disaggregation and bottom-up approaches formance (Table 7). Larger organizations may also
and their reconciliation, as suggested by Mentzer have to rely more on advanced systems due to
and Moon (2005), could not be confirmed as part increased complexity. With respect to demand volatil-
of FPQ, perhaps due to different approaches fol- ity, the MGA I showed that investments via Effective
lowed by various industries in this study. Table 6 Use of Advanced Systems are more impactful for FMC
shows that FPQ directly affects Accuracy and Cost when demand volatility is high.

Xxxx 2016 11
Journal of Supply Chain Management

Internal Integration did not have a significant Accuracy pathway. These results somewhat extend the
impact on FMC, unlike the other three constructs. results of past research. For instance, the early work of
But the importance of Internal Integration became Lee and Adam (1986) found that higher forecast
clearer in the contingency analysis. All survey items errors may not necessarily result in higher total cost.
loaded well on this construct with high reliability, as These and other similar findings were also empirically
shown in Table S1, except for forecast sharing with supported by the works of Danese and Kalchschmidt
suppliers. This factor requires that all related depart- (2011a, 2011b).
ments jointly arrive at forecasts, sharing forecast-rele- Thus, managers need to be cognizant of the fact
vant information among internal departments, along that improvements in forecast accuracy do not auto-
with a shared accountability for forecasts. This ele- matically translate into cost reduction, improvements
ment also includes what Fildes, Goodwin, Lawrence in delivery performance, and other operational
and Nikolopoulos (2009) call a forecasting meeting, advantages. Such outcomes are also dependent on
where marketing, production, and sales personnel the efficacy of related nonforecasting processes.
come together to agree on a final forecast or make These nonforecasting processes may in turn be rein-
judgment-based adjustments. Table 6 shows that the forced and improved through several elements of
Internal Integration construct does not have signifi- FMC.
cant direct effects on any of the three performance
outcomes. But the significance of Internal Integration
becomes evident in MGA I (Table 7), comparing LIMITATIONS AND FUTURE RESEARCH
firms facing high and low demand volatility. Internal Forecasting processes in many firms continue to
Integration has a stronger and more significant exhibit a distinct lack of structure and inadequate
impact on FMC when demand volatility is low. This mechanisms for translating forecasting methods and
is an interesting finding that merits further investiga- practices into operational advantages. This study per-
tion in future studies taking into account the integra- tains to the foundational issues of forecasting manage-
tion issues pertaining to related processes such as ment within a single firm. Such improvements in
S&OP, and in specific contexts such as lean produc- within-firm processes may also serve to improve fore-
tion environments, as shown in Wisner and Stanley casting processes at the supply chain level, aiding
(1994). information sharing, and collaborative forecasting
All four of the above sets of practices thus coher- among firms in a supply chain via initiatives such as
ently constitute the notion of an FMC. This forecast- CPFR and VMI.
ing competence was also shown to affect the Some of the limitations of this study pertain to the
outcomes of Forecast Accuracy, Cost Reduction, and survey methods employed. The results are based on
Delivery Performance significantly. The highest impact the information provided by key informants and one
of this FMC was on Cost Reduction, followed by survey instrument, which is common with this type of
Accuracy and Delivery Performance. survey research. Tests for heterogeneity such as Finite
Examining the relationships further among FMC Mixture Partial Least Squares (FIMIX-PLS) and MGA
and the three outcome variables, it was shown that ideally require larger sample sizes to estimate all pos-
Forecast Accuracy did not mediate or moderate the sible aspects.
relationship between FMC and Cost Reduction, and More studies are needed to further validate the
between FMC and Delivery Performance. It was notion of the FMC construct among a wider range of
shown that FMC directly affected Cost Reduction and manufacturing and service firms. The multigroup anal-
Delivery Performance without necessarily acting via yses indicated differences between companies in MTS
Forecast Accuracy. The presence of direct effects and make-to-order environments, yet we could not
between FMC and the outcome variables may seem interpret these results based on the measurement
perplexing at first, but it may be noted that FMC was invariance test. This may indicate that firms with dif-
defined as a broad competence including elements ferent production types treat forecasting management
which can have direct effects. Internal Integration may very differently, an aspect that needs more attention.
have direct effects on Cost Reduction and Delivery Among the outcome measures, in addition to forecast-
Performance independent of the Forecast Accuracy ing deviation, forecasting bias measures also need to
level, as it addresses the challenges of cross-functional be studied which may have a stronger impact on
collaboration and integration, contributing to the effi- operational performance (Ritzman & King, 1993).
cacy of the S&OP process. This may serve to augment Further research is warranted to investigate the com-
operational outcomes directly. Another constituent plementarity of S&OP processes with forecasting in
element of FMC, the Effective Use of Advanced Sys- order to translate the benefits of effective forecasting
tems, may also result in favorable direct effects on into favorable business outcomes. The adoption of a
outcome measures without necessarily acting on the configurational perspective for FMC, the identification

12 Volume 0, Number 0
Forecasting Management Competence

of gestalts, based on internal coherence among a set tailored design method. (3rd ed.). Hoboken, NJ:
of attributes, use of clustering techniques to identify Wiley.
and develop taxonomies, and tests of fit as profile Edwards, J. R. (2001). Multidimensional constructs in
deviation from a theoretical or empirical reference organizational behavior research: An integrative
model are other useful avenues for future research. analytical framework. Organizational Research
Methods, 4, 144–192.
Several aspects in forecast management are also sub-
Fildes, R., Bretschneider, S., Collopy, F., Lawrence, M.,
ject to behavioral dynamics relating to goal setting,
Stewart, D., Winklhofer, H., & Moon, M. A.
budgeting, reward structures, and accountability- (2003). Researching sales forecasting practice:
related issues, and these require further investigation. Commentaries and authors’ response on “Con-
ducting a Sales Forecasting Audit” by M. A. Moon,
J. T. Mentzer & C. D. Smith. International Journal
of Forecasting, 19 (1), 27–42.
REFERENCES Fildes, R., Goodwin, P., Lawrence, M., & Nikolopou-
Armstrong, J. S. (Ed.). (2001). Principles of forecasting: los, K. (2009). Effective forecasting and judg-
A handbook for researchers and practitioners (1st mental adjustments: An empirical evaluation
ed.). Dordrecht: Kluwer Academic Publishers. and strategies for improvement in supply-chain
Armstrong, J., & Overton, T. (1977). Estimating non- planning. International Journal of Forecasting, 25
response bias in mail surveys. Journal of Marketing (1), 3–23.
Research, 14, 396–402. Fildes, R., Nikolopoulos, K., Crone, S. F., & Syntetos, A.
Barney, J. (1991). Firm resources and sustained com- A. (2008). Forecasting and operational research: A
petitive advantage. Journal of Management, 17 (1), review. Journal of the Operational Research Society, 59
99–120. (9), 1150–1172.
Barney, J. (2012). Purchasing, supply chain manage- Fornell, C., & Larcker, D. F. (1981). Evaluating struc-
ment and sustained competitive advantage: The tural equation models with unobservable vari-
relevance of resource-based theory. Journal of Sup- ables and measurement error. Journal of Marketing
ply Chain Management, 48 (2), 3–6. Research, 18 (1), 39–50.
Becker, J.-M., Klein, K., & Wetzels, M. (2012). Hierar- Geisser, S. (1975). The predictive sample reuse
chical latent variable models in PLS-SEM: Guideli- method with applications. Journal of the American
nes for using reflective-formative type models. Statistical Association, 70 (350), 320–328.
Long Range Planning, 45 (5), 359–394. Hair, Jr., J. F., Hult, G. T. M., Ringle, C., & Sarstedt,
Census. (2010). Statistics of U.S. Businesses Main M. (2013). A primer on partial least squares struc-
Page. Retrieved May 1, 2014, from http://www. tural equation modeling (PLS-SEM). Thousand
census.gov/econ/susb/ Oaks, CA: Sage Publications.
Chin, W. W., & Newsted, P. R. (1999). Structural Henseler, J., & Chin, W. W. (2010). A comparison of
equation modeling analysis with small samples approaches for the analysis of interaction effects
using partial least squares. In R. H. Hoyle (Ed.), between latent variables using partial least squares
Statistical strategies for small sample research (pp. path modeling. Structural Equation Modeling, 17
307–341). Thousand Oaks, CA: Sage Publica- (1), 82–109.
tions. Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A
Churchill, Jr., G. A. (1979). A paradigm for develop- new criterion for assessing discriminant validity
ing better measures of marketing constructs. Jour- in variance-based structural equation modeling.
nal of Marketing Research, 16 (1), 64–73. Journal of the Academy of Marketing Science, 43
Dalrymple, D. J. (1987). Sales forecasting practices: (1), 115–135. http://doi.org/10.1007/s11747-
Results from a United States survey. International 014-0403-8
Journal of Forecasting, 3 (3–4), 379–391. http:// Henseler, J., Ringle, C., & Sarstedt, M. (forthcoming).
doi.org/10.1016/0169-2070(87)90031-8 Testing measurement invariance of composites using
Danese, P., & Kalchschmidt, M. (2011a). The impact partial least squares. Retrieved from http://
of forecasting on companies’ performance: Analy- info.smartpls.com/data/uploads/micom.pdf
sis in a multivariate setting. International Journal of Hunt, S. D. (2000). A general theory of competition:
Production Economics, 133 (1), 458–469. Resources, competences, productivity, economic growth.
Danese, P., & Kalchschmidt, M. (2011b). The role of Thousand Oaks, CA: Sage Publications.
the forecasting process in improving forecast Hunt, S. D., & Davis, D. F. (2008). Grounding supply
accuracy and operational performance. Interna- chain management in resource-advantage theory.
tional Journal of Production Economics, 131 (1), Journal of Supply Chain Management, 44 (1), 10–
204–214. 21.
Davis, D. F., & Mentzer, J. T. (2007). Organizational Hunt, S. D., & Davis, D. F. (2012). Grounding supply
factors in sales forecasting management. Interna- chain management in resource-advantage theory:
tional Journal of Forecasting, 23 (3), 475–495. In defense of the resource-based view of the firm.
Dillman, D. A., Smyth, J. D., & Christian, L. M. Journal of Supply Chain Management, 48 (2), 14–
(2009). Internet, mail, and mixed-mode surveys: The 20.

Xxxx 2016 13
Journal of Supply Chain Management

Hunt, S. D., & Morgan, R. M. (1995). The compara- Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric
tive advantage theory of competition. Journal of theory. New York, NY: McGraw-Hill.
Marketing, 59 (2), 1–15. Oliva, R., & Watson, N. (2009). Managing functional
Johnson, R. E., Rosen, C. C., Chang, C.-H. D., Djurd- biases in organizational forecasts: A case study of
jevic, E., & Taing, M. U. (2012). Recommenda- consensus forecasting in supply chain planning.
tions for improving the construct clarity of Production and Operations Management, 18, 138–
higher-order multidimensional constructs. Human 151.
Resource Management Review, 22 (2), 62–72. Oliva, R., & Watson, N. (2011). Cross-functional
Kahn, K. B. (2002). An exploratory investigation of alignment in supply chain planning: A case study
new product forecasting practices. Journal of Pro- of sales and operations planning. Journal of Opera-
duct Innovation Management, 19, 133–143. tions Management, 29, 434–448.
Kahn, K. B., & Mentzer, J. T. (1994). The impact of Peng, D. X., & Lai, F. (2012). Using partial least squares
team-based forecasting. Journal of Business Forecast- in operations management research: A practical
ing Methods and Systems, 13, 18. guideline and summary of past research. Journal of
Kalchschmidt, M. (2012). Best practices in demand Operations Management, 30 (6), 467–480.
forecasting: Tests of universalistic, contingency Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS
and configurational theories. International Journal procedures for estimating indirect effects in sim-
of Production Economics, 140 (2), 782–793. ple mediation models. Behavior Research Methods,
Kerkk€anen, A. (2010). Improving demand forecasting Instruments, & Computers, 36 (4), 717–731.
practices in the industrial context. Lappeenranta Priem, R. L., & Swink, M. (2012). A demand-side per-
University of Technology. Retrieved from http:// spective on supply chain management. Journal of
www.doria.fi/handle/10024/59314 Supply Chain Management, 48 (2), 7–13.
Klassen, R. D., & Flores, B. E. (2001). Forecasting Ritzman, L. P., & King, B. E. (1993). The relative sig-
practices of Canadian firms: Survey results and nificance of forecast errors in multistage manufac-
comparisons. International Journal of Production turing. Journal of Operations Management, 11 (1),
Economics, 70, 163–174. 51–65.
Kusaba, K., Moser, R., & Rodrigues, A. M. (2011). Sanders, N. R., & Manrodt, K. B. (1994). Forecasting
Low-cost country sourcing competence: A concep- practices in US corporations: Survey results. Inter-
tual framework and empirical analysis. Journal of faces, 24 (2), 92–100.
Supply Chain Management, 47 (4), 73–93. Shapiro, B. P. (1977). Can marketing and manufactur-
Lee, T. S., & Adam, Jr, E. E. (1986). Forecasting error ing co-exist? Harvard Business Review, 55 (5), 104.
evaluation in material requirements planning Smith, C. D. (2001). Improving supply chain sales
(MRP) production-inventory systems. Management forecasting. In J. T. Mentzer (Ed.), Supply chain
Science, 32 (9), 1186–1205. management (pp. 155–182). Thousand Oaks, CA:
Leuschner, R., Rogers, D. S., & Charvet, F. F. (2013). Sage Publications.
A meta-analysis of supply chain integration and Smith, C. D., & Mentzer, J. T. (2010). User influence
firm performance. Journal of Supply Chain Manage- on the relationship between forecast accuracy,
ment, 49 (2), 34–57. application and logistics performance. Journal of
Makridakis, S., & Hibon, M. (2000). The M3-Compe- Business Logistics, 31 (1), 159–177.
tition: Results, conclusions and implications. Sousa, R., & Voss, C. A. (2008). Contingency research
International Journal of Forecasting, 16 (4), 451– in operations management practices. Journal of
476. Operations Management, 26 (6), 697–713.
Makridakis, S., Wheelwright, S. C., & Hyndman, R. J. Sterman, J. D. (1989). Modeling managerial behavior:
(1998). Forecasting: Methods and applications (3rd Misperceptions of feedback in a dynamic decision
edn). New York, NY: John Wiley & Sons. making experiment. Management Science, 35 (3),
McCarthy, T. M., & Golicic, S. L. (2002). Implement- 321–339.
ing collaborative forecasting to improve supply Stone, M. (1974). Cross-validatory choice and assess-
chain performance. International Journal of Physical ment of statistical predictions. Journal of the Royal
Distribution & Logistics Management, 32 (6), 431– Statistical Society. Series B (Methodological), 36 (2),
454. 111–147.
Mentzer, J. T., & Moon, M. A. (2005). Sales forecasting Wetzels, M., Odekerken-Schroder, G., & Van Oppen,
management: A demand management approach. C. (2009). Using PLS path modeling for assessing
Thousand Oaks, CA: Sage Publications. hierarchical construct models: Guidelines and
Moon, M. A., Mentzer, J. T., & Smith, C. D. (2003). empirical illustration. Management Information Sys-
Conducting a sales forecasting audit. International tems Quarterly, 33 (1), 11.
Journal of Forecasting, 19 (1), 5–25. Winklhofer, H., Diamantopoulos, A., & Witt, S. F.
Narasimhan, R., Jayaram, J. J., & Carter, J. R. (2001). (1996). Forecasting practice: A review of the
An empirical examination of the underlying empirical literature and an agenda for future
dimensions of purchasing competence. Production research. International Journal of Forecasting, 12
and Operations Management, 10 (1), 1–15. (2), 193–221.

14 Volume 0, Number 0
Forecasting Management Competence

Wisner, J. D., & Stanley, L. L. (1994). Forecasting He has published widely in leading journals and has
practices of JIT and non-JIT purchasers. European presented numerous papers at international confer-
Journal of Purchasing & Supply Management, 1 (4), ences. His papers have appeared in journals such as
219–225. Management Science, Decision Sciences, Journal of Opera-
Zotteri, G., & Kalchschmidt, M. (2007). Forecasting tions Management, International Journal of Production
practices: Empirical evidence and a framework for
Research, International Journal of Production Economics,
research. International Journal of Production Eco-
IIE Transactions, International Journal of Flexible Manu-
nomics, 108 (1), 84–99.
facturing Systems, European Journal of Operational
Research, Computers and Industrial Engineering, Interna-
tional Journal of Operations and Production Management,
Journal of Manufacturing Systems, Production and Inven-
tory Management, Journal of Intelligent Manufacturing,
Torsten Doering (M.B.A., State University of New Operations Management Review, and SME Technical
York—Buffalo) is a Ph.D. candidate at The State Papers. His contributions include the coedited book
University of New York—Buffalo and an assistant pro- Group Technology and Cellular Manufacturing: A State-of-
fessor at Daemen College. His research interests the-Art Synthesis of Research and Practice, Kluwer Aca-
include empirical supply chain management and demic Publishers, Boston/Dordrecht/London. He is a
manufacturing research with focus on demand plan- recipient of: Alfred Bodine/Society of Manufacturing
ning and integration in a global context. He is cur- Engineers Award for Studies in Machine Tool Eco-
rently working on the impact of culture and interfirm nomics; Joseph T.J. Stewart Faculty Scholarship, and
collaboration at county, industry, and firm levels the Sustained Achievement Exceptional Scholar Award
using hierarchical, longitudinal, and structural equa- at SUNY, Buffalo.
tion modeling approaches. Mr. Doering is the recipi-
ent of the 2015 Rising Star Award at the University of
Buffalo School of Management. SUPPORTING INFORMATION
Additional Supporting Information may be found in
Nallan C. Suresh (Ph.D., University of Cincinnati) the online version of this article:
is UB Distinguished Professor & Chair, Department of
Operations Management & Strategy at the School of Table S1. Operationalization of Constructs and
Management, State University of New York, Buffalo. Statistics.

Xxxx 2016 15

You might also like