Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
Sale Forecasting of Merck Pharma Company using ARMA Model
Nawaz Ahmad
Visiting Professor at PAF KIET, Karachi
Fouzia Nasir
PhD Scholor at Indus University, Karachi
Usman Aleem
Assistant Professor at PAK KIET, Karachi
Abstract
This study aims to develop a stochastic framework of model to forecast future sales for pharmaceutical industry.
In this regard, the study focuses on Merck Pharmaceutical monthly sales data. This study examines the Sale
forecasting models. The study includes monthly data published in the annual reports of the company from Jan.
2008 to Dec. 2012.The time series diagram shows unequal means over the time period that suggests the data is
stationary. Having transformed the data, ARMA (1, 1) model is applied which shows that there will be increase
in sales by $6.784m given that in the last month sales were $1bn. On the contrary, last month’s residual has an
adverse effect on current month sales up to the extent of $432.942m. In this study AR (1) and MA (1) both the
processes are significant at 1%
Keywords: Sales forecast, ARMA (1, 1), Pharma Industry
1. Introduction
Forecasts are needed throughout an organization -- and they should certainly not be produced by an isolated
group of forecasters. Neither is forecasting ever "finished". Forecasts are needed continually, and as time moves
on, the impact of the forecasts on actual performance is measured; original forecasts are updated; and decisions
are modified, and so on.
Forecasting is the process of making statements about events whose actual outcomes (typically) have
not yet been observed. A commonplace example might be estimation of some variable of interest at some
specified future date. Usage can differ between areas of application: for example in hydrology, the terms
"forecast" and "forecasting" are sometimes reserved for estimates of values at certain specific future times, while
the term "prediction" is used for more general estimates, such as the number of times floods will occur over a
long period.
Risk and uncertainty are central to forecasting and prediction; it is generally considered good practice
to indicate the degree of uncertainty attaching to forecasts.Although quantitative analysis can be very precise, it
is not always appropriate. Some experts in the field of forecasting have advised against the use of mean square
error to compare forecasting methods.
Sales forecasting is an essential tool in the master budget. The sales budget is one of the most
important tools because the accuracy of other budgets depends on the accuracy of the sales budget. The sales
budget is dependent upon sales forecasts, therefore analyzing past patterns of sales, general economic conditions
and other factors aid in creating the sales budget. Sales forecasting is done by predicting sales under a set of
conditions based on past sales and future outlooks.
In creating a sales forecast, past performance and analysis of expected market conditions are the two
major factors that are taken into consideration. The sales budget is dependent upon sales forecast therefore
creating an accurate sales forecast will be beneficial to Guillermo. Sales forecasts are also supported by
documentation on how much the client plans on spending and the profit that will be received for the sales. It is
essential that Guillermo use every effort when creating the sales forecast to assure that his budgeting is done
correctly. (Jock, 2008).
The sales forecast is essential to assure inventory levels are accurate, staffing plans meet the
requirement and suppliers, customers and investors are pleased with the outcome of the product and the amount
being distributed. There are many risks that can be associated with sales forecasting due to the financial risk
that can be associated because of producing too much inventory and having to many or too few resources to meet
the sales that are going out. (Jock, 2008). It is essential that Guillermo take past information from his Flex
Budget and have an understanding of the demand for his product based on future estimation if he plans to be
successful with his business venture. Forecasting is not based on intuition but on calculations and are an
essential tool in the master budget in order to prepare for future business ventures.
In planning, one must make forecasts. In making forecasts, you consider a number of possibilities
(called “states” of the world). One possibility is that business conditions will be “normal”, another that the
30
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
economy will go into a mild recession and business suffers a little, and so on. To each possibility one associates
a likelihood (or “probability”). If we have taken into account all possible outcomes, then these probabilities must
sum to unity.
Each possible “outcome” implies certain things about your business – e.g., if the economy experiences
a mild recession your sales may drop. When all these outcomes are considered some average (“expected value”)
of your future sales is obtained.
Forecasts assist the finance group as they develop revenue plans, determine appropriate expense levels,
and forecast the profitability of the company. The operations group uses forecasts to develop a production
schedule, to make component buying decisions, and to plan for any required capacity changes needed to meet
demand.
Developing a sales forecast for existing products can easily be arrived at by conducting a statistical
analysis of historical sales data and then combining this information with anticipated changes in market
dynamics, sales organization structure and pricing. Forecasting sales revenue and product utilization for novel
medical technologies becomes much more difficult due in part to the lack of historical sales data and the
unknowns associated with a new product in the marketplace.
Using input from market based assumptions and company related parameters, a spreadsheet-based
model can be built which allows the user to more accurately forecast sales revenue and product demand. With
these models, users can determine the effect that changes to baseline assumptions can have on the forecast.
As a standalone company, Merck & Co. faces a tough future of declining sales. Data monitor forecasts
that the proposed merger with Schering-Plough will succeed in returning Merck to a positive sales growth
outlook.
If management delivers on its promise of an additional $3.5bn of annual cost savings beyond 2011,
Data monitor calculates that the combined company will see operating profit growth rate accelerate to a strong
6.9% compound annual growth rate (CAGR) 2008-13.
As it is evident from Figure 5, in terms of generated revenues for 2008, Merck’s main three areas of
focus are cardiovascular, respiratory and infectious diseases. Coupling that with the knowledge that out of the
$12bn portfolio that will be exposed to generic competition by 2015, 29% comes from cardiovascular, 36% from
respiratory and seven percent from infectious diseases, the need for a drastic deal that will reshape Merck’s
future becomes apparent.
Within the cardiovascular arena, Merck’s hypertension portfolio (Cozaar and Hyzaar) also holds great
importance with sales approaching $3.6bn in 2008. However, patent expiries will all but eliminate this revenue
stream from 2010 onwards. These imminent patent expiries, in combination with the challenges faced by the
cholesterol franchise have left Merck in a highly exposed position with regards to its cardiovascular products.In
addition, the resulting company is expected to change its focus in the cardiovascular arena from primary care
indications to indications mainly covered by specialists.
Merck has performed an impressive entry in the diabetes therapy area leapfrogging Novartis to bring
the first, and still only, DPP-4 (dipeptidyl peptidase-4) inhibitor in the market. With sales of $1.4bn in 2008, two
years after launch, Januvia (sitagliptin) has had the second most successful launch in the cardiovascular and
metabolic diseases arena, behind only Pfizer’s mega-blockbuster Lipitor. Merck has already launched a
combination with metformin (Janumet) and is developing a combination with pioglitazone (MK-0431c). The
FDA changed the requirements for approval of antidiabetic medications late in 2008, raising the bar in terms of
proving cardiovascular safety and at the same time ensuring Januvia enjoys an even longer period in the market
without any true competitors, Dr. Karachalias says.
“Schering-Plough has virtually nothing to offer in the diabetes arena both in terms of marketed
products and in terms of developmental pipeline. Inevitably, Merck’s focus in the therapy will be somewhat
diluted post-merger.
2. Literature review
Although quantitative techniques are arguably very useful and often should be part of a company’s forecasting
process, they have certain weaknesses that can be counterbalanced by the use of qualitative forecasting. (Fulcher,
1998; Moon and Mentzer, 1999; Helms et al., 2000). Quantitative time-series techniques are designed to identify
and forecast trends and seasonal patterns in data and to adjust quickly to changes in these trends or patterns.
Their limitation is that Assortment forecasting method 141 they do not consider contextual information, such as
price changes (Mentzer and Schroeter, 1994). Regression analysis makes it possible to take such factors into
consideration, but the complexity of the method and its significant data requirements limit its use (Lapide, 1999).
Neither of the methods does well in dealing with changes that have never happened before, or that have
happened before but for which no data exist in the system. This is where expert judgment can add significant
value to the forecasting process. (Mentzer and Bienstock, 1998).
Situations in which expert judgment is needed include, in addition to the price changes mentioned
31
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
above, assortment changes, promotions, competitor activities, and product introductions. The best information
concerning these situations oftentimes resides with the company’s marketing and sales personnel (Fulcher, 1998;
Fosnaught, 1999; Moon and Mentzer, 1999; Helms et al., 2000; Jain, 2000; Reese, 2000).
Although both researchers and practitioners seem to agree that sales force involvement in forecasting
is important, benefiting from it in practice can be difficult. Several motivational, organizational, and tool-related
obstacles have been identified. In their in-depth study of the sales forecasting management practices at 33
companies, Moon and Mentzer (1999) found there to be some resistance from salespeople concerning their
forecasting responsibilities in almost all of thecompanies studied. Many salespersons felt that it was not their job
to forecast and that time spent on forecasting was time taken away from their real job of managing customer
relationships and selling products and services. Similar observations have been reported by Helms et al. (2000)
and Reese (2000).
According to Reese (2000), these motivational problems are often aggravated by the lack of
forecasting incentives; salespersons are seldom rewarded for producing accurate forecasts. Moreover, Moon and
Mentzer (1999) claim that even when companies get salespeople to forecast, they tend to do a relatively poor job.
As they put it: even when the salespeople are provided with a history of their customers’ demand patterns, they
frequently will either see patterns that do not exist, or will fail to see patterns that do exist (Moon and Mentzer,
1999). Based on their research, Moon and Mentzer (1999) have compiled a set of guidelines for overcoming the
barriers that hinder companies from fully benefiting from sales force involvement in forecasting. They suggest
that companies should: Make forecasting part of the salespeople’s job by including forecasting as a part of their
job descriptions, creating incentives for high performance in forecasting, and providing feedback and training.
Minimize game playing by making forecasting accuracy an important outcome for salespeople and clearly
separating sales quotas from forecasts.
Keep it simple by asking salespeople only to adjust statistically generated forecasts rather than
producing forecasts from scratch and by providing them with adequate tools that enable them to complete their
forecasting work as efficiently as possible. . Keep it focused by having the salespeople deal only with the
products and customers that are truly important and where their input can significantly affect the company’s
supply chain.
The first two recommendations concern organizational and motivational factors, such as rewards, job
descriptions, and training. The other two are about creating forecasting processes and tools that support sales
force involvement by making forecasting simpler, more efficient, and more focused on the products and
customers that really matter. These latter ones are the focus of this paper.
3.
Methodology
3.1 Data and Variable
The data extracted from the annual report of Merck pharmaceutical company. The selected data is based on
monthly sales from Jan 2008 to Dec 2012.
3.2 Data analysis
The selection and application of correct prediction methodology has always been an important issue of planning
and control for most companies and organizations. Often, the financial well-being of the entire operation rely on
the accuracy of the forecast since such information will likely be used to make interrelated budgetary and
operative decisions in areas of personnel management, purchasing, marketing and advertising, capital financing,
etc. If, on the other hand, has historically experienced signature pattern upward and downward sales, then the
complexity of the task is compounded prediction. Data is analyzed via E – views 6, an econometric package.
3.3 Model
We use Correlogram followed by Unit root test for Stationarity. Then Auto Regressive Moving Average –
ARMA (1, 1)model to find whether trailing sales and first lag of error term has significant impact while sales is
forecasted.
3.4 Hypothesis
H1: Trailing sales has no impact on current sales.
H2: First lag of error has no impact on current sales.
32
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
4.
www.iiste.org
Results and Discussion
Date: 06/04/15 Time: 12:59
Sample: 2008M01 2012M04
Included observations: 52
Autocorrelation
. |******|
. |******|
. |***** |
. |***** |
. |***** |
. |**** |
. |**** |
. |**** |
. |**** |
. |*** |
. |*** |
. |** |
. |** |
. |*. |
. |*. |
. |*. |
.|. |
.|. |
.|. |
.|. |
.*| . |
.*| . |
.*| . |
.*| . |
Partial Correlation
. |******|
. |*. |
. |*. |
.|. |
. |*. |
.|. |
.|. |
.|. |
. |*. |
.*| . |
.|. |
.|. |
.*| . |
.*| . |
. |*. |
.*| . |
.|. |
.|. |
.|. |
.*| . |
.|. |
.|. |
.|. |
.|. |
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
AC
PAC
0.874
0.789
0.735
0.674
0.646
0.605
0.564
0.516
0.494
0.437
0.381
0.345
0.269
0.193
0.169
0.108
0.073
0.054
0.003
-0.057
-0.095
-0.113
-0.154
-0.170
0.874
0.106
0.114
-0.020
0.129
-0.032
0.006
-0.061
0.097
-0.157
-0.033
-0.007
-0.155
-0.135
0.140
-0.159
0.071
0.004
-0.055
-0.144
0.049
0.051
-0.058
-0.018
Q-Stat
42.035
76.953
107.92
134.47
159.44
181.81
201.67
218.68
234.63
247.39
257.33
265.67
270.87
273.63
275.80
276.71
277.14
277.38
277.38
277.67
278.48
279.67
281.98
284.90
Prob
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
As we can observe a decreasing pattern in auto correlation function (ACF) which further gets into
negative and makes a wave shape, it depicts sales data is stationary. ACF is associated with moving averages.
Furthermore, partial autocorrelation function (PACF) is significant at first lag which suggests that only first lag
of sales ie, last month’s sales has a significant effect on current sales as it is associated with auto regression.
The time series diagram shows monthly sales for the period from 2008- 2011 which shows unequal
means over the time period that suggest the data is stationary.
33
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
Date: 06/04/15 Time: 13:39
Sample: 2008M01 2012M04
Included observations: 51
Autocorrelation
***| . |
.|. |
. |*. |
**| . |
. |*. |
.|. |
. |*. |
.*| . |
.|. |
.|. |
.*| . |
. |** |
.*| . |
.|. |
. |*. |
.*| . |
.|. |
. |*. |
.|. |
**| . |
. |*. |
.*| . |
.|. |
. |*. |
Partial Correlation
***| . |
.*| . |
. |*. |
.*| . |
.|. |
.|. |
. |*. |
.*| . |
.|. |
.|. |
. |*. |
. |** |
.|. |
.|. |
. |*. |
.|. |
.*| . |
.|. |
. |*. |
**| . |
.|. |
.*| . |
.|. |
.|. |
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
AC
PAC
-0.377
-0.031
0.203
-0.226
0.097
-0.029
0.123
-0.172
0.070
0.066
-0.073
0.233
-0.110
-0.051
0.169
-0.141
-0.022
0.092
0.005
-0.223
0.162
-0.070
-0.052
0.208
-0.377
-0.202
0.137
-0.115
-0.005
-0.060
0.188
-0.131
0.004
0.017
0.087
0.214
0.073
-0.064
0.139
0.001
-0.090
-0.018
0.089
-0.217
-0.034
-0.171
0.007
0.065
Q-Stat
7.6867
7.7410
10.062
13.011
13.568
13.620
14.549
16.413
16.727
17.010
17.369
21.146
22.012
22.206
24.349
25.890
25.929
26.615
26.617
30.962
33.339
33.798
34.054
38.376
Prob
0.006
0.021
0.018
0.011
0.019
0.034
0.042
0.037
0.053
0.074
0.097
0.048
0.055
0.074
0.059
0.056
0.076
0.087
0.114
0.056
0.043
0.052
0.064
0.032
Since the sales data was stationary at level, the first difference was taken and again the data was run
and now it has been observed that sales data become non stationary as there is no loop observed in ACF.
Furthermore, PACF is also significant for lag 1.
Since ACF and PACF both are significant at first lag, and ACF follows MA process and PACF follows
AR process therefore Auto Regressive Moving Average i.e. ARMA (1, 1) process is the most suitable model to
forecast monthly sales which follows first lag of sales and first lag of error term.
4.1 ARMA Model
Dependent Variable: SALES
Method: Least Squares
Date: 06/04/15 Time: 13:58
Sample (adjusted): 2008M03 2012M04
Included observations: 50 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C
SALES(-1)
E(-1)
511.0677
1.006784
-0.432942
3199.698
0.054696
0.143863
0.159724
18.40686
-3.009405
0.8738
0.0000
0.0042
R-squared
Adjusted R-squared
S.E. of regression
Sum squared resid
Log likelihood
F-statistic
Prob(F-statistic)
0.884200
0.879272
4210.816
8.33E+08
-486.6706
179.4354
0.000000
Mean dependent var
S.D. dependent var
Akaike info criterion
Schwarz criterion
Hannan-Quinn criter.
Durbin-Watson stat
34
58387.60
12118.87
19.58682
19.70155
19.63051
2.203031
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
In ARMA output, both first lag of sales and error term is significant at 1%. There will be increase in
sales by $6.784m given that in the last month sales were $1bn. On the contrary, last month’s residual has an
adverse effect on current month sales up to the extent of $432.942m.
Moreover, there is no sample error as the difference between r-square and adjusted r-square is as low
as 0.5%. Overall model is also significant as F-statistics is quite high (179.43). There is no evidence of strong
auto correlation as Durbin-Watson statistics is closer to 2.
5. Conclusion
This study concludes that the sales data has unit root which has overcome via taking first difference.
Furthermore, the pattern of ACF indorses the presence of unit roots whereas PACF suggests that only first lag of
error term is significance. There will be increase in sales by $6.784m given that in the last month sales were
$1bn. On the contrary, last month’s residual has an adverse effect on current month sales up to the extent of
$432.942m. Both the variables are significant at 1%.
6. References
Armstrong, J.S. (2001), “Selecting forecasting methods”, in Armstrong, J.S. (Ed.), Principles ofForecasting,
Kluwer Academic Publishers, Norwell, MA.
Basu, S. and Schroeder, R.G. (1977), “Incorporating judgments in sales forecasts: application ofthe Delphi
method at American Hoist and Derrick”, Interfaces, Vol. 7 No. 3, pp. 18-27.
Bunn, D. and Wright, G. (1991), “Interaction of judgemental and statistical forecasting methods:issues&
analysis”, Management Science, Vol. 37 No. 5, pp. 501-18.
Cadeaux, J.M. (1997), “A closer look at the interface between the product lines of manufacturersand the
assortments of retailers”, International Journal of Retail & DistributionManagement, Vol. 25 No. 6, pp.
197-203.
Dalrymple, D.J. (1988), “Sales forecasting practices”, International Journal of Forecasting, Vol. 3No. 3, pp. 37991.
Edmundson, R.H., Lawrence, M.J. and O’Connor, M.J. (1988), “The use of non-time series data insales
forecasting: a case study”, Journal of Forecasting, Vol. 7 No. 3, pp. 201-11.
Fildes, R. (1991), “Efficient use of information in the formation of subjective industry forecasts”,Journal of
Forecasting, Vol. 10 No. 5, pp. 597-617.
Fosnaught, K. (1999), “The strategic power of consensus forecasting: setting your organizationup to win”,
Journal of Business Forecasting Methods & Systems, Vol. 18 No. 3, pp. 3-7.Freeland, K. (2003),
Personal communication, 14 August.
Fulcher, J. (1998), “A common vision”, Manufacturing Systems, Vol. 16 No. 2, pp. 88-94.
Helms, M.M., Attain, L.P. and Chapman, S. (2000), “Supply chain forecasting – collaborativeforecasting
supports supply chain management”, Business Process Management Journal,Vol. 6 No. 5, pp. 392-407.
Herbig, P., Milewicz, J. and Golden, J.E. (1993), “Forecasting: who, what, when and how”, Journalof Business
Forecasting Methods & Systems, Vol. 12 No. 2, pp. 16-21.
Holmstro¨m, J. (1998), “Handling product range complexity: a case study on re-engineeringdemand forecasting”,
Business Process Management Journal, Vol. 4 No. 3, pp. 241-58.
Jain, C.L. (2000), “Editorial: which forecasting model should we use?”, Journal of BusinessForecasting Methods
& Systems, Vol. 19 No. 3, pp. 2-4.IJPDLM34,2156
Kahn, K.B. (1998), “Benchmarking sales forecasting performance measure”, Journal of BusinessForecasting
Methods & Systems, Vol. 17 No. 4, pp. 19-23.
Lapide, L. (1999), “New developments in business forecasting”, Journal of Business ForecastingMethods &
Systems, Vol. 18 No. 2, pp. 13-14.
Lim, J.S. and O’Connor, M. (1996), “Judgmental forecasting with time series and causal information”,
International Journal of Forecasting, Vol. 12 No. 1, pp. 139-53.
McClelland, A.S., Raman, A. and Fisher, M. (2000), “Supply chain management at World Co Ltd.”,Harvard
Business School Case N9-601-072.
Makridakis, S. (1988), “Metaforecasting: ways of improving forecasting accuracy and usefulness”, International
Journal of Forecasting, Vol. 4 No. 3, pp. 467-91.
Mentzer, J.T. and Bienstock, C.C. (1998), “The seven principles of sales-forecasting systems”,Supply Chain
Management Review, Fall, pp. 76-83.
Mentzer, J.T. and Kahn, K.B. (1995), “A framework of logistics research”, Journal of BusinessLogistics, Vol. 16
No. 1, pp. 231-50.
Mentzer, J.T. and Schroeter, J. (1994), “Integrating logistics forecasting techniques, systems, andadministration:
the multiple forecasting system”, Journal of Business Logistics, Vol. 15No. 2, pp. 205-26.
Moon, M.A. and Mentzer, J.T. (1999), “Improving salesforce forecasting”, Journal of BusinessForecasting, Vol.
35
Research Journal of Finance and Accounting
ISSN 2222-1697 (Paper) ISSN 2222-2847 (Online)
Vol.6, No.21, 2015
www.iiste.org
18 No. 2, pp. 7-12.
Muir, J.W. (1979), “The pyramid principle”, Proceedings of 22nd Annual Conference, AmericanProduction and
Inventory Control Society, pp. 105-7.
Peterson, R.T. and Jun, M. (1999), “Forecasting sales in wholesale industry”, Journal of BusinessForecasting
Methods & Systems, Vol. 18 No. 2, pp. 15-17.
Plossl, G.W. (1973), “Getting the most from forecasts”, Production and Inventory Management,Vol. 14 No. 1,
pp. 1-16.
Reese, S. (2000), “The human aspects of collaborative forecasting”, Journal of BusinessForecasting Methods &
Systems, Vol. 19 No. 4, pp. 3-9.
Safavi, A. (2000), “Choosing the right forecasting software and system”, Journal of BusinessForecasting
Methods & Systems, Vol. 19 No. 3, pp. 6-10.
Sanders, N. and Manrodt, K. (1994), “Forecasting practices in US corporations: survey results”,Interfaces, Vol.
24 No. 2, pp. 92-100.
Sparkes, J.R. and McHugh, A.K. (1984), “Awareness and use of forecasting techniques in Britishindustry”,
Journal of Forecasting, Vol. 3 No. 2, pp. 37-42.
Webby, R. and O’Connor, M. (1996), “Judgemental and statistical time series forecasting: a reviewof the
literature”, International Journal of Forecasting, Vol. 12 No. 1, pp. 91-118.Assortment forecasting
method157.
36