DEVELOPMENT OF A DEMAND FORECASTING MODEL

FOR SPECIALTY PLANT NUTRITION SOLUTIONS

APPLICABLE TO INDIA

DISSERTATION SYNOPSIS

submitted by

RAHUL MIRCHANDANI

in partial fulfillment of the requirements for the

DOCTORAL PROGRAMME IN MANAGEMENT (Ph.D.)

under the guidance of

DR S R GANESH

Senior Professor

NARSEE MONJEE INSTITUTE OF MANAGEMENT STUDIES

Deemed University, Mumbai.

1

SYNOPSIS CONTENTS

Page #

1. Introduction 2

1.1 What Are Specialty Fertilizers? 2

1.2 Dissertation Scope 3

1.3 Need for the Study 3

2. Research Objectives 3

3. Variables 4

4. Research Methodology 5

4.1 Choosing an appropriate forecasting technique 5

4.2 Data Collection Methods 6

5. Sample & Respondents 8

6. Validity & Reliability 9

6.1 Validity 9

6.2 Inter-item Reliability 10

6.3 Parallel Form Reliability 10

7. Research Framework 11

Table of Objectives, Research Problems, Hypotheses, Approach,

Information Needs, Methodology & Tools

11

8. Research Results 14

8.1 Developing the Overall Demand Forecast Model 14

8.2 The Demand Forecasting Equation 16

8.3 Test of the Model 18

9. Implications 19

10. Limitations of the Study 20

11. Contributions of the Study 21

12. Suggestions for Future Research 21

Bibliography 23

2

Title of the Study: “Development of a Demand Forecasting Model for Specialty Plant Nutrition Solutions applicable to India”

1. Introduction Agriculture is one of the most complex and diverse elements of the economic fabric of India. The use of modern farming practices on a wider scale and integrated nutrient management practices are essential if India’s farmers wish to produce crops in line with the observed global standards of quantity and quality. If a farmer does not provide balanced ‘food’ to his crops, he cannot expect optimum levels of farm productivity. 1.1 What Are Specialty Fertilizers?

There are a total of sixteen elements recognized as essential plant nutrients, each having specific functions. Neither can a plant complete a healthy life cycle in the absence of even one of the essential nutrients, nor can they be replaced.

Note: Silicon (Si), the seventh micronutrient, has been excluded as there is no specific deficiency recorded of this nutrient in India. The application of nutrients like Nitrogen (through Urea), Phosphorus (through Di-Ammonium Phosphate) and Potassium (through Sulphate of Potash) is commonplace. These are nutrients required in large quantities and they are well recognized as major fertilizers. In fact, there are several instances observed of their rampant overuse, which is a serious cause for concern. Excess nutrients applied tend to reduce the efficiency of uptake of the other nutrients present in the soil, compounding the problems of deficiencies. However Nitrogen, Phosphorus & Potassium are not the only nutrients required by plants. Calcium, Magnesium and Sulphur are considered as secondary nutrients. Several other nutrients called micronutrients; though required in smaller quantities, are equally important for good growth and development of plants. The essential micronutrients are Boron (B), Chlorine (Cl), Copper (Cu), Iron (Fe), Manganese (Mn), Molybdenum (Mo) and Zinc (Zn). Plants take up all

3

these nutrients simultaneously and their requirements vary with type of plants, growth stages, yield potential, etc. All available sources of micro-nutrients will be classified as Specialty Plant Nutrition Solutions for the purpose of this study. 1.2 Dissertation Scope This Dissertation will undertake a systematic study to: a. Identify the factors that influence demand for specialty plant nutrition solutions (or

specialty fertilizers) in the Indian context. b. It would then attempt to use this knowledge to construct a predictive model for

forecasting demand for these specialty fertilizers. 1.3 Need for the study The need to be more productive has manifested itself in farmers adopting new methods of agriculture and the latest agricultural technology required to improve the economics of farming. Integrated Nutrient Management forms an integral part of this framework. As a result, demand situations for inputs required by the farm sector are no longer certain. Certainty, longer product life cycles and low competitive intensity are things of the past. The overall environment has become dynamic. (Singh, 2005) The environment is changing rapidly with the dynamism arising not simply from the interaction of the individual structural components of the industry, but also from the industry “field” itself. (Emery and Trist, 1965) New farming methods, increased awareness, understanding and acceptance for modern farming methods, including the application of specialty fertilizers, within the farming community have set the ‘industry’s “ground” itself in motion’ (Emery and Trist, 1965) creating an environment that can be aptly classified as a “turbulent field”. Demand has become uncertain, product life cycles have shortened, and competition has intensified. In such a situation, understanding demand, planning demand and linking supply with demand is crucial. (Singh, 2005) The paucity of demand forecasting models in the agricultural inputs sector is a major gap in the literature available to the Agribusiness Industry. There are models available that forecast output of individual categories of farm products as well as models that predict the impact of factors like rainfall, government policy, etc. on demand and supply of agricultural commodities. However, forecasting models for inputs are rare. Extensive search revealed only one model for forecasting global demand of commodity (major) fertilizers, developed by the Food and Agriculture Organization (FAO). There was no similar forecasting model available in the public domain prepared specifically for India. Considering the growing acceptance of new techniques and technologies of farming and the consequent impact on the demand for specialty fertilizers in India, development of a demand forecasting model for this Industry will be extremely useful. The study will also present a step by step methodology that managers can use to forecast demand in similar dynamic industry environments. The model will be especially useful to industries marketing specialty products or new product concepts where past demand data is unavailable, and the web of demand triggers is known but constantly changing. The study will identify and assess the individual impact of each factor affecting demand separately. Strategies may be formulated to ‘influence the influencers’ based on this information. Demand estimates would also assist in devising operational frameworks, planning inventories, increasing manufacturing and supply chain efficiency, etc. The study will therefore, add to the body of knowledge in the field of forecasting and aid planning within emerging markets.

4

2. Research Objectives This Study has the following primary Research objectives: To identify the factors that influence demand for specialty fertilizers in the Indian context To suggest a profile of the typical potential user (farmer) for specialty fertilizers To assess the extent of the relationship between the causal factors and demand for specialty

fertilizers To use above to generate a predictive model to forecast demand for specialty fertilizers

applicable to India. 3. Variables The following diagram illustrates the conceptual framework underlying the overall demand for specialty fertilizers. Variables have been classified into three heads:

(Conceptual Framework prepared with extensive inputs from: Aggarwal et al (2002), Alagh (2003), Bandyopadhyay & Perveen (2002), Bhalla (2001), Dhar & Kallummal (2004), Finck A. (2002), Foster & Rosenzweig (2003), Ghonsikar & Shinde (1997). Iyer (2003) Kamath (2002) Mahadevan (2003), Ray (2004), Sen (2001) and Subramanyam & Sudha (2002))

Independent Variables

Dependent Variable

ExtraneousVariables

Nature of farming practices

Affordability of the specialty nutrients

Recommendations and Word of Mouth

Awareness of specialty

nutrients

Field Trials of specialty

nutrients

Government Policy

Weather & Climate

Labour costs & shortages

Demand for

Specialty Plant

Nutrition Solutions

Financial capacity of farmer

Agri Output prices

Water availability

Infrastructure bottlenecks

Farmer attitudes

Trade channel influence

Type of Crop grown

5

4. Research Methodology 4.1 Choosing an appropriate forecasting technique A detailed review of forecasting techniques was undertaken in order to identify those that would be the most appropriate to this study. This was an essential step in arriving at the research methodology. A table summarizing the various techniques reviewed and the reasons why they were considered appropriate or inappropriate as part of the methodology for this study is presented below.

Table 1: Evaluation and Rationale for Choice of Forecasting Techniques Forecasting Technique

Decision regarding usage for a (near term) demand forecasting

model for specialty fertilizers

Rationale for decision

Econometric Methods

Do not use Primarily useful for long range forecasts; unimpressive results in short range forecasting; past demand data unavailable for extrapolation

Naïve Methods Do not use No assessment made about causality of specific factors

Causal Methods Method to be used Decomposition of ‘causal forces’ and assessment of impact of each causal factor on the demand for specialty nutrients will be especially useful in improving forecasting accuracy.

Intentions Surveys

Method to be used Intentions based forecasts more accurate than extrapolation of past sales; especially useful when past demand data not available; collecting ‘probability of purchase’ estimates from farmers is not complicated.

Delphi Technique

Do not use

Administering Delphi questionnaires anonymously and without the researcher being physically present in a rural setting would be a constraint.

Unaided Judgement

Do not use Experts were no better than chance while developing forecasts in complex situations.

Game Theory Do not use No research has directly tested the forecasting accuracy of game theory

Role Playing Do not use Useful only in understanding how people respond to exogenous pressures (government policy, weather, etc.)

Focus Groups & In depth interviews

Method to be used (only for understanding the decision context and trends)

Unstructured discussions rarely lead to accurate forecasts. However, they are of value in understanding market dynamics

Neural Nets Do not use High data requirements; results generated using computer intensive methods are often difficult to interpret & understand

√

√

√

6

Forecasting Technique

Decision regarding usage for a (near term) demand forecasting

model for specialty fertilizers

Rationale for decision

Data Mining Do not use

Little evidence of utility in forecasting

Segmentation Method to be used

Increases depth and improves accuracy of the overall forecast. The intentions survey will gather data based on crop wise segments. Thereafter, demand estimates will be made for every specialty nutrient separately and then summed to generate the overall forecast.

Rule Based Forecasting

Do not use Absence of empirically validated, fully disclosed prior rules that can be applied to the forecasts.

4.2 Data Collection Methods

A personally administered questionnaire was chosen as the data collection tool for phase one of the study. The respondents were assembled at multiple locations in meeting rooms where the questionnaire was available in multiple Indian languages. Respondents were free to choose the questionnaire in the language of their choice. The order of questions, overall format and scales were identical in all the questionnaire forms, irrespective of language. The questionnaires were distributed to all present. Every question was read out in English, Hindi and the local language. In case of the local language, help from a local speaker familiar with all the local language, English and Hindi was taken at each meeting location. Uma Sekharan (2003) states “wherever possible, questionnaires are best administered personally to groups of people because of these advantages:

(i) Establishes rapport with the respondents before administering the questionnaire (ii) Respondents were able to clarify any doubts that they had on any questions on the

spot. (iii) It provided the opportunity to introduce the research topic and motivate respondents

to offer frank answers. (iv) Administration was possible to a large group of respondents at the same time. (v) All completed responses were collected before the end of the hour long meeting. This

ensured a 100% response rate. (vi) The method is less expensive and less time consuming as conducting personal

interviews” To take advantage of these benefits, this method of questionnaire administration was chosen. Open ended questions were avoided completely in the survey questionnaire. This was to ensure uniformity across all respondents. Moreover, since respondents were reading and responding in multiple languages, open ended responses would complicate data analysis which would require translation of answers into English prior to coding. The interpretation of the translators would add bias to the responses and it was thought best to avoid this. Closed ended questions used in the survey form centered around ten broad focus areas:

(i) Ownership and size of farm land

√

7

(ii) Top three crops grown in the area (iii) Identifying any change in cropping patterns over the last two years (iv) Assessing level of usage of specialty plant nutrition across different crop types (v) Impact of individual factors affecting demand for specialty plant nutrition solutions (vi) Usage of specialty plant nutrition solutions in the previous season (vii) Probability of using specialty plant nutrition solutions in the next season, assuming

certain changes occur in the farming practices or due to external factors (measuring the ‘intention to try’)

(viii) Identifying who decides on agricultural input purchases (ix) Identifying which single influencer has the highest impact on the buying decision for

specialty plant nutrition solutions (x) Classification questions that asked for information regarding location (state, district,

village), contact numbers and size of the family unit.

The survey also requested the respondents to provide contact details of three persons in their area who could be contacted for In depth Interviews on the subject. This list was used as the sample frame for Phase Two of this study.

Considering the demographic profile of the respondents (farmers and agricultural input retailers from Rural India) and the diversity in languages, cultures and educational backgrounds, it was essential that all measurement scales be kept extremely simple. Questions in the survey only required “Yes/No” answers, Ranking top 3 crops as “1-2-3” and indicating responses using tick marks on a 4 point interval scale. The interval scales used were deliberately kept ‘unbalanced’, allowing no neutral choice point. This ensured that every respondent had to necessarily state if a particular factor had an impact on specialty plant nutrition demand or not. This was essential as the factor either had an impact on demand or did not have an impact on demand. There was no reason for a “may or may not have an impact” choice point. Also, considering that the respondents were not very familiar with answering detailed questionnaires or long rating or ranking scales, the number of intervals was limited to four. Increasing the number of choice points on the measurement scale would only confuse the respondents. The final two questions also asked for only tick marks to indicate the influencers in the purchase decision for agricultural inputs. Phase Two involved In depth Interviews and Focus Groups which provided ample opportunity to gain deep insights into the buying process, using open ended questions and probing. There was no structured questionnaire handed out to the respondents during this phase of the study. Respondents were met at their homes or on their farms to ensure that they are at ease during the process. It also ensured that they were able to speak at length on the subject. Each interview lasted close to an hour and was video taped (with permission) for ease of qualitative analysis. All respondents were familiar with Hindi and this helped the process of questioning and probing immensely. A check list of questions and focus areas to be covered during the In-depth Interviews was drawn up. This check-list was common across all respondents interviewed during this phase of the study. However the interview was not structured with rapid fire questions. Each respondent was asked to talk at length on the decision making process for agricultural inputs and this discussion was gradually focused onto specialty plant nutrition solutions. Respondents spoke at length about their farming practices as well. This provided deep insights into the minds of the farmers and the

8

factors that influence their buying behaviour. Each interview was videotaped and summarized into a concept map. 5. Sample & Respondents One of the primary objectives of this study was to identify the factors influencing demand for specialty plant nutrition solutions. Hence, the end consumers – i.e. farmers, need to be contacted to ascertain why they choose to purchase such inputs for use on their fields. The only method to directly ascertain farmers' intentions regarding fertilizer usage is through a sample survey. N.S. Parthasarathy (1994), stated in a publication for the United Nations Food and Agriculture Organisation, that “in the fertilizer business, it is possible to assess farmers' intentions through retailers.” Many retailers are located in villages and are in daily touch with farmers. “Many are themselves practicing farmers and have a good feel for agricultural prospects for the ensuing season or year.” (Parthasarathy, 1994) Being close to the scene of action, they have a good assessment of produce price trends, available purchasing power, likely crop shifts, etc. It should be possible to ascertain from each retailer directly or through a large sample, the likely demand for specialty fertilizers. It was decided to use agricultural input retailers as the respondents for this study. The retailer database of a specialty fertilizer marketing company having perhaps the widest distribution reach across India was selected as a sample frame. This was considered appropriate as it provided contact details of 44,000 retailers from 17 States of the country, providing an extremely wide geographical spread to the sample. From this sample frame, a cluster (area) sample of retailers was drawn. The Cluster (Area) sampling ensured the state wise geographical spread of the respondents. The number of respondents included in the sample from each state attempted to follow the same distribution as that of fertilizers consumed in that respective state and the area under cultivation within that State. The basic assumption here is that areas consuming more quantities of major fertilizers and having more area under cultivation have more agriculture and should therefore show higher consumption of specialty plant nutrition solutions. Once the number of respondents was decided, the selection of respondents (retailers) invited to attend the meeting was done at random. They were invited to a central location at a specified date and time. Once respondents gathered, the survey questionnaire was administered to all present. All retailers present at these meetings were handed out questionnaires in their choice of language. English questionnaires prepared were translated by professional copy writers into Telugu, Marathi, Hindi, Tamil, Bengali, Kannada, Oriya and Gujarati. The order of questions and format was kept identical. A total of 877 respondents answered the questionnaire. 81.9% of these respondents were farmers themselves, in addition to being agri input retailers (this is in line with the rationale put forth by Parthasarathy, 1994). The remaining 17% of respondents were not farmers, but purely agricultural input retailers. Due to their occupation, all 877 respondents were in regular touch with farmers in their respective areas. These 877 respondents represent 17 States and 201 districts of the country. Since the sample was essentially a non-probability sample, it was essential to check the sample for representation. This is essential to ensure the possibility of generalizing the results of the study.

9

To check on how representative this sample was in relation to the state-wise distribution of the total farming population, the sample distribution was correlated with secondary data on two separate criteria, namely, state-wise area under cultivation and state-wise fertilizer consumption. At a 99% Level of Confidence (α = 0.01), there was a statistically significant correlation observed between the total area under cultivation in the respective states and the number of respondents from that state included in the sample. This points to the fact that more the area under cultivation in a particular state, the more the number of respondents from that state were included in the sample. Further, at a 99% Level of Confidence (α = 0.01), there was a statistically significant correlation observed between the total fertilizer consumption in the respective states and the number of respondents from that state included in the sample. This points to the fact that more the fertilizer consumption in a particular state, the more the number of respondents from that state were included in the sample. After completion of Phase One of the sample survey as detailed above, a follow up qualitative study was conducted using in depth interviews and focus groups of farmers in the state of Maharashtra. This state had farmers with very wide crop diversity and would thus permit gathering of insights from growers of diverse crop types. Within the persons named by the survey respondents within Maharashtra state, selection of farmers to be interviewed was done on a ‘convenience’ basis. However, it was ensured that the farmers contacted during this phase grew multiple crops of a diverse nature on their fields. 6. Validity & Reliability 6.1 Validity In the case of the questionnaire used in this study, content validity, i.e., ensuring the inclusion of adequate and representative set of items to measure demand triggers for specialty plant nutrition solutions, was ascertained. The face validity of the questionnaire was ascertained by administering the questionnaire on a test group of ten ‘subjects’ (a group of agri input retailers, who were also farmers) within the Pune district of Maharashtra state. They all agreed that the items included did identify demand triggers and, on the face of it, appear to be appropriate measures of the concept. This criteria was assessed though it is viewed as a ‘basic and very minimum index of content validity’. (Sekharan, Uma 2003, pp 206) The Inter-Correlation Matrix gives an indication of how closely related the variables under investigation are. This matrix was drawn up for all variables, irrespective of whether or not the hypotheses are directly related to these analyses. (Sekharan Uma, 2003, pp 307) Not a single correlation coefficient in the entire matrix is above 0.455. This allows us to conclude that all variables are distinct and we can reasonably be sure of the validity of the measures. Further, if two variables that are theoretically stated to be related do not seem to be significantly correlated to each other in our sample, the validity and reliability of the measurement of the concepts under investigation would be suspect. A detailed check on the Inter-correlation matrix does not reveal any such discrepancies. This allows us to be reasonably certain of the overall validity of the measures used in the study.

10

6.2 Inter-item Reliability In the questionnaire, two sets of measures were gathered, both on a four point interval scale, to assess consistency across eight factors affecting demand for specialty plant nutrition solutions. To verify the reliability of the questionnaire responses, a test of consistency of the respondent’s answers to these items (independent measures of the same construct) was conducted. The most popular test of inter-item consistency reliability is the Cronbach’s coefficient alpha. (Sekharan Uma, 2003). This is generally used for multi-point scaled items and is therefore appropriate. The reliability coefficient, Cronbach’s Alpha reported is 0.6747. Generally, Alpha coefficients between 0.6 and 0.7 are considered ‘acceptable’ (Sekharan Uma, 2003, pp. 311) and hence we can conclude that there is ‘acceptable’ consistency in the responses of the respondents to questions measuring the same construct. We may therefore, take the responses as fit for detailed analysis and generalizations. 6.3 Parallel Form Reliability Responses were gathered in the survey to assess the probability of use of a specialty plant nutrition solution during the next season assuming a particular change occurs in the area. The responses were collected on a 4 point interval scale with options being “Will Not Purchase”, “May not purchase”, “May purchase” and “Will definitely purchase”. Since respondents generally do not tend to think in terms of probabilities, this scale served as an appropriate data collection tool. For analysis, these responses were converted into probabilities as follows:

Response on 4 point scale Probability of purchase assigned Will not purchase 0% or 0.00 May not purchase 33% or 0.33

May purchase 66% or 0.67 Will definitely purchase 100% or 1.00

Here, we now had two parallel sets of measures tapping the same construct. To check for the reliability, analysis using Cronbach’s Alpha coefficient was conducted. The reliability coefficient, Cronbach’s Alpha reported is 0.82. Generally, Alpha coefficients above 0.8 are considered ‘good’ (Sekharan Uma, 2003, pp. 311) and hence we can conclude that there is ‘good’ consistency in the responses on the 4 point interval scale and on the probability scale. We may therefore, take these responses as fit for detailed analysis and generalizations.

11

7. RESEARCH FRAMEWORK Table 2: Research Objectives, Research Problems, Hypotheses, Approach, Information Needs, Methodology & Tools

Research Objective Research Problem Hypothesis Approach Information Needs Methodology/Tools

To estimate the current usage levels of specialty nutrients by farmers of different crop types

What is the level of usage of specialty fertilizers by farmers of each crop type?

N.A. Gather usage levels within a particular area (as a percentage of farmers growing different crop types using specialty nutrients) using a 4-point scale (No usage – low usage – medium usage – high usage)

Total area under cultivation for each crop type Respondent estimates of the proportion of farmers having high usage, medium usage, low usage and no usage of specialty nutrients.

Structured Instrument (Questionnaire survey) Sample Size : 877

To identify the factors that influence demand for specialty plant nutrition solutions

What are the factors that influence purchase of specialty fertilizers?

H10 : There is no significant interrelationship between the variables influencing the demand for specialty fertilizers H11 : There is a significant interrelationship between the variables influencing the demand for specialty fertilizers

Factor Analysis

Farmer (end user) ratings on a 4-point scale indicating the level of impact (No impact – Low impact – Medium impact – High impact) that each of the influencing variables has on the demand for specialty fertilizers.

Structured Instrument (Questionnaire survey) Sample Size: 877

12

Research Objective Research Problem Hypothesis Approach Information Needs Methodology/Tools

What are the levels of influence that each of the variables has on the demand for specialty fertilizers?

N.A. Survey of Buyer Intentions

End user probability ratings (using a 4-point scale) assessing the probability of purchase of specialty nutrients, given the existence of each factor included in the list.

Structured Instrument (Questionnaire survey) Sample Size : 877

To suggest a basic profile of the typical potential user (farmer) for specialty fertilizers

What are the characteristics of a consumer who is most likely to purchase a specialty fertilizer?

H20 : There is no significant discriminating power in the variables H21 : There may be a significant discriminating power in the variables

Discriminant Analysis

Classification data (limited to farm acreage, type of crops grown and change in cropping patterns) Grouping Variable : current usage information

Structured Instrument (Questionnaire survey) Sample Size : 877

To ascertain the key decision makers determining the purchase of specialty nutrients

Who decides on the agricultural inputs that are being purchased for your farm? Who or what influences the farm inputs purchase decision the most?

N.A.

Frequency distributions

Listing of all influencers Indication of most crucial influencer in the decision making process

Structured Instrument (Questionnaire survey) Sample Size : 877

13

Research Objective Research Problem Hypothesis Approach Information Needs Methodology/Tools

To gain in depth insights into the decision making process regarding specialty nutrients

What are the dynamics that are involved in the decision to purchase specialty nutrients?

N.A. Expert opinion (for validation of quantitative survey and deeper qualitative insights)

Responses in line with in-depth interview checklist

Number of in-depth interviews : 12 (Respondents growing different types of crops and having differing area under cultivation selected for in-depth interviews & focus groups)

To use above to generate a near term predictive model to forecast demand for specialty plant nutrition.

Based on the study of various influencing variables, prepare a forecasting equation to estimate demand for specialty nutrients (dependent variable), given existence and impact of each causal variable.

Construct Demand Forecasting equation

Total area under cultivation, current usage levels, stated intentions (probability of purchase), agronomic requirements (from government notifications)

Prepare predictive equation Prepare pessimistic, optimistic and average estimates of specialty fertilizers demand for each nutrient separately, in quantity and market value terms.

To estimate the accuracy of the demand estimate

What is the overall accuracy of the estimate of demand for specialty fertilizers?

Estimate of total demand for specialty nutrients derived from other sources

Compare demand estimates and evaluate model accuracy.

14

8. Research Results The key research results gathered from the quantitative and qualitative phases of this study are summarized below: 1. The core factors that influence demand for specialty nutrients are market prices of

agricultural output (past and current season), usage of hybrid seeds, nature of farming practices, recommendations (by other farmers, private companies, etc.) and others (weather, external funding, size of area under cultivation, age of farmer, financial strength of farmer, etc.)

2. Based on the ‘intention to buy’ data, a farmer is most likely to purchase specialty fertilizers

if he switches over to using hybrid seeds. The next six influencers (in descending order of importance) are neighbouring farmer’s reports, recommendations by company staff, retailer recommendation, sales promotion schemes, change over to horticultural crops and installation of micro irrigation systems.

3. Insights from farmers revealed that advertising and recommendations by government

officials) have no significant influence on demand. Women also have very little influence over the buying decision for agricultural inputs, including specialty nutrients, though majority of them are working on the fields.

4. Half of the farmers make purchase decisions regarding specialty nutrients themselves.

Elder male family members and agricultural input retailers are the two next most important decision makers.

5. A basic model to identify potential users who are likely to use specialty fertilizers has been

drawn up using crop type, area of farm land and change in cropping patterns as the basis. The model has accurately identified users 78% of the times. The model has used only three variables. Adding more variables could improve the predictive ability of the model further.

6. Given awareness about the concept and having conducted or witnessed a successful trial

using specialty nutrients (two necessary conditions to be fulfilled before any demand arises), the mean probability of purchase, taking all factors collectively, is estimated at 73.13%.

7. The actual purchase behaviour indicates that mean purchase intentions were understated by

11.8%. This difference must be incorporated into the forecasting model. 8.1 Developing the Overall Demand Forecast Model The demand forecasting model being proposed shall take into account the following:

(i) Total Market potential for the product group, i.e. specialty nutrients. (ii) Level of awareness of the concept among the consumers (iii) Extent of ‘demonstrated performance’ of the product using trials amongst the aware

consumers (or target market) (iv) Existence, impact and importance of each of the influencing factors that play a role

on the purchase decision for the product (v) Affordability of the product, having a bearing on actual buying behaviour (vi) Availability of the product, having a bearing on actual buying behaviour

15

(vii) Willingness to pay for the product, having a bearing the actual buying behaviour (viii) Error in estimate resulting due to overstated or understated intentions (ix) General (agronomic) requirement of the specialty nutrients (per unit of land area) as

established by the Government. (x) Market Price of the specialty nutrient sources, in order to convert the estimate of

quantity demanded into a monetary estimate of market size. Awareness about specialty nutrients is a pre-requisite that must be satisfied before any farmer becomes a potential user for specialty nutrients. Awareness about the concept is a key determining factor for demand. Unless utility of the concept is known and understood, there is no chance that a farmer will purchase a specialty nutrient. The forecasting model will thus incorporate awareness (A) as a dichotomous variable, where A = 0 if farmer is unaware about the concept of specialty nutrients, and A = 1 if farmer is aware about the concept of specialty nutrients. When assessing demand forecast for such products within any group of farmers or for any geographical territory, level of awareness must be ascertained. A representative sample of farmers could be asked whether they are aware about this concept or not. Their ‘yes – no’ responses once analyzed will provide a good estimate of awareness levels in the area. It must be borne in mind that awareness levels in every area or group of farmers will vary. Once a farmer is made aware, he will always conduct a trial on his own field within a small area before using the specialty nutrients over the entire area under cultivation. Successful conduct of this trial is also a critical determining factor before demand from a particular farmer becomes a reality. If the trial on his field is unsuccessful, the specialty nutrients will never be purchased. Demand will materialize only if the trial is successful, i.e., the field results are favourable. Word of mouth spreads after successful trials are conducted within the immediate geographical area. The forecasting model will need to consider whether such field trials have been successful, using a dichotomous variable ‘T’, where , where T = 0 if farm trials on specialty nutrients have been un-successful, and T = 1 if farm trials on specialty nutrients have been successful. Once again, enquiries within the farming community in the area will give insights into this variable. It must be borne in mind that demand in any area for specialty nutrients will be zero if trials have either not been conducted or witnessed by farmers or if the trials are unsuccessful. Twelve other influencing variables have also been studied and their impact on purchase probabilities for specialty fertilizers has been ascertained during the farmer intentions survey. Based on this, the “probability that an individual farmer will demand specialty nutrients

during a particular season” may computed as: DPr = A(0,1) x T(0,1) x Σ (Xi Pi)

n

where, A = 0 if farmer is unaware about the concept of specialty nutrients, and A = 1 if farmer is aware about the concept of specialty nutrients T = 0 if farm trials on specialty nutrients have been un-successful or not yet conducted, and T = 1 if farm trials on specialty nutrients have been successful Pi = mean probability of purchase for factor ‘i’ (assessed using intentions survey responses)

i = 1

k

16

Xi = dichotomous variable indicating the existence of influencing factor ‘i’ (Xi = 0 if factor is not valid for the farmer, Xi = 1 if factor is valid for the farmer) ‘i’ ranges from 1 to k; k = number of influencing variables (k = 12 in this study) n = number of valid influencing factors, i.e., number of factors where Xi = 1 The above DPr estimate is valid where all factors are (or assumed to be) of equal importance, or if the priority of importance is not known to the forecaster. To get a more accurate forecast, it is useful to assign weights to each of the influencing factors. An assessment of the importance of each factor can be ascertained using factor analysis. Alternatively, a qualitative assessment of importance by speaking with groups of potential users is possible. Weightages (Wi) may be assigned to each influencing variable (Xi). The above expression can factor in these weightages (as assigned by the forecaster) as follows:

DPr = A(0,1) x T(0,1) x Σ Wi Xi Pi

Σ Wi

Σ Wi is the sum of weightages assigned. Note that Wi must necessarily be 0 wherever Xi = 0 Since DPr is a probability estimate, 0 < DPr < 1, with DPr moving closer to 1 as probability of purchasing specialty nutrients increases. To ascertain the probability that a group of farmers will demand specialty nutrients during a particular season, the above equation will continue to be appropriate, except that all variables will need to be ascertained for the entire group of farmers being targeted. Alternatively, if the group is small in size, individual probabilities may be ascertained and then averaged. Further, as discussed in section 5.5, the above probability of demand for specialty nutrients would need to be adjusted (PAct – PInt) [if available] would need to be built into the final probability estimate, if available. Thus, the Adjusted DPr = DPr + (PAct – PInt ) [0 < Adjusted DPr < 1] Usually, in case of initial estimates or first time studies, the difference between ‘actual purchase’ and ‘stated intent to buy’ would not be available. Over time, when forecasts are repeatedly developed for a certain area or group of farmers, such an ‘adjustment factor’ would become known to the forecasters (when they assess the accuracy of the forecasts) and can be applied. 8.2 The Demand Forecasting Equation After arriving at the ‘overall probability of purchase’ estimate (preferably adjusted for the difference between actual purchase behaviour and stated intentions), the overall demand forecast for specialty nutrients may be arrived at using the function below: QDj = (Area X AW%) X TR% X Adjusted DPr X Agj

i = 1

i = 1 k

k

17

Where, QDj = Demand (in quantity terms) of the j’th specialty nutrient Area = Total Area under cultivation in the area for which the forecast is being prepared AW% = Percentage of farmers in the area who are aware about the concept of specialty nutrients TR% = Percentage of aware farmers in the area who have conducted or witnessed a successful trial using specialty nutrients Adjusted DPr = Overall probability of purchase of specialty nutrients during a particular season (as estimated using a survey of farmers intentions), adjusted for (PAct – PInt), i.e. the difference between actual observed purchase behaviour (PAct) and stated intentions (PInt) [ascertained by comparing the mean intentions gathered during the intentions survey and comparing with the actual buying behaviour recorded during the season, taking a sample of respondents surveyed] Agj = General agronomic requirement of the j’th specialty nutrient (in terms of quantity per unit of land area per year – for e.g., kilos/hectare/year) Note that within the Adjusted DPr = DPr + (PAct – PInt), the PAct factor deals with issues of availability and affordability.

(i) If the specialty nutrient is priced such that the farmer who intends to purchase finds it unaffordable, the PAct estimate will decline.

(ii) If the specialty nutrient is unavailable to the farmer who intends to purchase it (either due to stock-outs, inventory shortages or lack of reach of the distribution network), the PAct estimate will decline.

(iii) If the respondent farmers, overstate their intentions to purchase during the survey (perhaps due to social desirability bias or auspices bias, etc), the PAct estimate will even out the overstated intentions estimates. This holds true even in case the respondents understate their intentions for any reason during the survey.

Considering the advantages of using the Adjusted DPr, only in case the adjustment factor is not available, DPr should be used in its place within the forecasting equation. The expanded demand forecasting equation for all specialty nutrients would be as under:

M = Σ ( QDj x Yj )

where, M = total demand for specialty nutrients in the area under study in money terms QDj = quantity demanded of the j’th specialty nutrient (in quantity terms; for eg. metric tons or kilos, etc.) m = the total number of specialty nutrients whose demand forecast is being prepared Yj = average current market price of the j’th specialty nutrient (in money terms; for eg. Rupees per metric ton or Rupees per kilo, etc.)

QDj = (Area x AW%) x TR% x A(0,1) x T(0,1) x Σ Wi Xi Pi + (PAct – PInt) X Agj

Σ Wi

Thus, QDj = f (Area, AW, TR, DPr, Agj) [all terms within the function as described above]

i = 1

i = 1 k

k

j = 1

m

18

The following table indicates the concepts being addressed by the various variables included in the forecasting equation.

Table: Variables and Concepts Variable in Equation

Concepts being addressed by the variable

Remarks

AREA Total Market potential for the product group, i.e. specialty nutrients.

This is the total addressable potential : total area under cultivation

AW% Level of awareness of the concept among the consumers

Can be ascertained using a survey of respondents in the area

TR% Extent of ‘demonstrated performance’ of the product using trials amongst the aware consumers

Can be ascertained using a survey of respondents in the area

Xi, A, T Existence of each of the influencing factors that play a role on the purchase decision

A=1 if awareness exists, A=0 otherwise; T=1 if successful trial conducted or witnessed, T=0 otherwise, Xi = 1 if particular influencer is relevant or present, Xi = 0 otherwise

Pi Impact of each of the influencing factors that play a role on the purchase decision

Impact on buying decision of each factor Xi is measured in terms of ‘probability of purchase’ in case the factor exists or manifests itself

Wi Importance of each of the influencing factors that play a role on the purchase decision

Weightage may be assigned by the forecaster subjectively/ qualitatively

PAct Affordability If the farmers find that the product is not affordable, demand will not materialize and PAct will fall

PAct Availability If there is a supply bottleneck or if the specialty nutrient is unavailable at the point of sale for whatever reasons, demand will not materialize; PAct will fall

PAct Willingness to pay If the farmers are unwilling to pay for the specialty nutrient for whatever reasons, demand will not materialize and PAct will fall

PAct – PInt Error in estimate resulting due to overstated or understated intentions

PAct – PInt will account for any overstated or understated intentions expressed by the respondents

Agj

General (agronomic) requirement of the specialty nutrients (per unit of land area)

The requirement of each nutrient has been notified by the Government from time to time. Gazetted notifications can be used as the basis here.

Yj Market Price of the specialty nutrient sources

Average market price (MRPs) are available from the company price lists

8.3 Test of the Model One way to assess accuracy of a forecasting model has been to examine the agreement among judgmental forecasts. For example, Ashton (1985), found that the agreement among the individual judgmental forecasts was a useful proxy for accuracy.

19

In order to test the accuracy of the results of this forecast, a comparison was made with the Industry’s judgmental forecast as follows: The Indian Micro-fertilizers Manufacturers’ Association (IMMA) estimates that the micronutrients consumption in India stands at 0.87% of the major fertilizer consumption. (Note that the world average is 4%) Since there is no published quantitative estimate of market size or consumption of micro-fertilizers, an estimate of the micronutrient demand calculated at 0.87% of major fertilizer consumption works out to: Micronutrient Consumption = 0.87% x (All India Major Fertilizer Consumption) = 0.87% x (184 lakh tonnes)* = 1.60 lakh tonnes *Source : CMIE Indian Harvest Database The ‘average demand’ estimate based on the model works out to 1,78,919 MT, i.e. 1.78 lakh tonnes The difference between the two estimates (above) works out to approx 11.25% It must be noted that the IMMA estimate of percentage of micro-fertilizer consumption vis-à-vis major fertilizers was made in the year 1995. Ever since, there has been no similar estimate prepared by the Association. Literature search has not revealed a similar estimate by any other Institution. It is very reasonable to assume that awareness levels and hence usage of specialty fertilizers has grown higher than this 0.87% level. The quantity estimate of micronutrients as per our forecasting model works out to 0.967% of major fertilizer consumption. This increase of micronutrient usage (as a percentage of major fertilizer consumption) of approx. 0.097% over a period of 11 years will largely explain the difference in demand estimates. The accuracy of the model may hence be considered ‘adequate’. 9. Implications This model can be applied to industries beyond specialty fertilizers. It can be used for any specialty product where building awareness about the concept and its utility is critical. For example, in the pharmaceuticals industry, when a new drug is introduced, the model may be applied to estimate demand for the new drug in the short term. The ‘Area’ may be substituted with ‘Population’ (read as ‘target population’, in case the drug is relevant to only a particular age group or gender or geographic area, etc.) and Agronomic requirement (Agj) may be substituted with ‘prescribed dosage’. In case the drug is a curative drug, a dichotomous variable ‘S’ would need to be included (see QDc below) where S=1 if the disease or symptoms exist, S=0 otherwise. In case of a preventive drug, this specific term need not be included (see QDp below). The factors (Xi) would include recommendation by doctor, press reports on efficacy of the drug, international approvals, etc. A survey of buyer intentions can be conducted to ascertain the probability of usage (Pi) of the new drug, subject to each of the listed factors. A weightage (Wi) may also be assigned to each factor.

20

AW% and TR% continue to remain ‘estimates of awareness about the drug’ and ‘percentage of aware/target consumers who have successfully tried/witnessed trials of the drug’, respectively. Thus, for preventive drugs sold over the counter (with no specific prescription from a medical practitioner required), the demand estimate would be:

QDp = (Popln x AW%) x TR% x A(0,1) x T(0,1) x Σ Wi Xi Pi + (PAct – PInt) x Dose

Σ Wi

For curative drugs which are sold over the counter (with no specific prescription from a medical practitioner required) the demand estimate would include the extra variable S (as described above):

QDc = (Pop x AW%) x TR% x S(0,1) x A(0,1) x T(0,1) x Σ Wi Xi Pi + (PAct – PInt) x Dose

Σ Wi

Similarly, the model could be applied across other industries as well. The general requirements for applicability of this forecasting model are that:

i. The product or service is not a commodity or an impulse purchase item. ii. It must have a specific, identifiable usage and benefit which can be demonstrated by

means of a trial or demonstration. iii. The product should be considered ‘important’ by the consumers, either due to the extent

of monetary outlay required to purchase the product or due to its utility. iv. The markets should typically be in the introductory and growth stages. As markets

become mature, concepts tend to get ‘commoditized’, usage of the product becomes part of the consumers’ habits, routinized buying behaviour sets in and differences in the relative importance of factors blur. As an assessment of the impact of the influencing factors on the purchase decision is required, the model cannot be applied when markets reach maturity. This is because, when markets reach this stage, consumers purchase the product by force of habit and do not pause to think of ‘influencing factors.

10. Limitations of the Study This study has surveyed buyer intentions for specialty fertilizers across India. The ‘probability of purchase’ estimates and importance (weightage) of individual factors have been evaluated on an All India basis. There may be certain changes in these estimates and weightages within individual states. Considering this limitation, the model has been applied and demand estimates prepared only on an All India basis. In case, state specific demand estimates are required, assessment of state specific purchase intentions would become necessary using a representative sample from each state. The structure of the forecasting equation would however remain unchanged.

i = 1

i = 1 k

k

i = 1

i = 1 k

k

21

The assessment of difference between actual purchase behaviour and stated intentions is critical to the accuracy of the model. This ‘adjustment factor’ addresses issues of affordability, availability, ability and willingness to buy as well as under and overstated intentions. In case this factor is not available to users of the model, the accuracy of the demand estimate would be lower. The qualitative phase of this research identified a few factors that were not specifically included in the survey questionnaire. ‘Probability of purchase’ estimates exclude the impact of these variables. The factor analysis results indicate that close to two thirds of the influencing variables have been explained by the survey (see section 5.2.3). The impact of the remaining variables has not been specifically identified by this study. These could perhaps be identified in a follow up study and included as additional variables within the forecasting equation. It must be however borne in mind that while adding and assessing the impact of more variables would increase accuracy of the model, it would add to respondent fatigue and model complexity. 11. Contributions of the Study The specialty nutrients industry has been in existence in India since the 1970s. However, it is still rather small in terms of absolute market size. The application of such a model will be extremely useful to companies operating within this industry. Considering the fact that this industry and companies within it are growing fast, planning operations using a good estimate of near term demand scenarios will lead to establishing long term operational efficiencies. A similar study or demand estimation model is currently not available to this industry. From a strategic perspective, understanding the impact and importance of the individual variables that affect demand will be extremely useful. The results of the study would be useful to other agribusiness companies selling specialized products, like new molecules of crop protection products, new generation seeds, etc. Companies can identify the core factors and develop strategies to try and ‘influence the influencers’. For those variables that are not within the organization’s control but of high importance, tracking the influencers would become possible. The demand forecasting model proposed by the study can also find application across a wide spectrum of nascent and growth markets. It illustrates a step by step methodology that marketers can use to plan for markets where published data regarding past demand is unavailable. It is an attempt to move beyond extrapolation and other econometric models for forecasting demand. 12. Suggestions for Future Research This model is useful to estimate demand in the short term only. It is dependent on a survey of buyer intentions which gets more subjective as the time horizon increases. A respondent cannot be expected to give an indication of the probability of purchase of a product beyond the near term (say six months to one year). Future research could seek to extend this model to try and develop demand estimates for the medium and long term. The model has been developed and tested in this study on the specialty nutrients industry in India. Research opportunities exist in applying the model on the specialty nutrients industry overseas, and also comparing its applicability across different geographies. Considering the fact that the

22

specialty nutrients industry is at varying levels of maturity in different countries, cross country comparisons would be interesting. Factors that affect the demand for specialty nutrients and their levels of importance may differ across countries. They will also change over time. Some factors may lose their importance. New factors may emerge. Continuous research on the identifying the nature, evaluating the impact and estimating the importance of these factors that affect demand would need to be conducted to keep the forecasts accurate. Moreover, the model can also be applied to any other industry where the product or service being offered is new and concept marketing is involved. Application of the model across industries with a portfolio of specialty offerings will be useful. Another area of continuing importance would be to continuously track the intentions survey data and actual purchase patterns. Over time, it may become possible to estimate (within confidence limits) this level of difference between stated intentions and actual buying behaviour. Recognizing that the profile of buyers is different across different industries, it is only fair to suggest that these estimates would be largely industry or sector specific. However, attempts to arrive at such an estimate and also identify the factors that cause this difference would be very useful.

23

REFERENCES & BIBLIOGRAPHY

Agricultural Economics Allen, P. Geoffrey (1994). “Economic Forecasting in Agriculture”. International Journal of Forecasting, 10 (1994) pp. 81-135. Armstrong, Scott J. (1994). “The Fertile Field of Meta-Analysis: Cumulative Progress in Agricultural Forecasting.” International Journal of Forecasting, 10 (1994), pp. 140-147. Foster A.D. & Rosenzweig, M.R. (2003). Agricultural Productivity Growth, Rural Economic Diversity, and Economic Reforms: India, 1970-2000. D. Gale Johnson Memorial Conference, October 25, 2003. Kurian, Vinson (2004). “Weather still best guide to rural demand.” Hindu Business Line Online. http://www.blonnet.com/2004/10/18/stories/2004101800901300.htm (accessed Jan 9, 2006) Mahadevan, Renuka (2003). Productivity Growth In Indian Agriculture: The Role Of Globalization And Economic Reform. Asia-Pacific Development Journal Vol. 10, No. 2, December 2003. Ministry of External Affairs (2003): The Indian Economy. http://www.divest.nic.in (accessed on January 6, 2005) Ray, Sankar (2004) : WTO Regime may bring uncertainties in Indian Agriculture. NetGuru India Business News, Sept 23, 2004. Singh, N.K. (2003) : How India Registered Growth in the age of Global uncertainties. The Indian Express, 24 May 2003. Singh, Rakesh (2005). “Predicting demand in an uncertain world.” Hindu Business Line Online http://www.thehindubusinessline.com/catalyst/2005/10/06/stories/2005100600060200.htm (accessed January 9, 2006) Agricultural Technology Aggarwal, P.K. et al (2002) : Impact of Climate Change Scenarios on Indian Agriculture. Indian Agricultural Research Institute, New Delhi-110012 & Central Research Institute for Dryland Agriculture, Hyderabad. Alagh, Yoginder K. (2003) : Water Development & Management. http://www.ficci.com/ficci/ mediaroom/speeches-presentations/2003/oct/oct21-water-ykalagh.ppt. (accessed on January 6, 2005) Bandyopadhyay, Jayanta & Perveen, Shama (2002): The Interlinking of India’s Rivers – Some questions on the Scientific, economic & Environmental dimensions of the proposal. Centre for Development and Environment Policy, Indian Institute of Management Calcutta. Bhalla, G.S. (2001) : The Nature of Agricultural Development in India. http://isidev.nic.in/ pdf/nature.pdf. Jawaharlal Nehru University. (accessed on January 6, 2005)

24

Dhar, Biswajeet & Kallummal, Murali (2004) : Trade Liberalisation & Agriculture – Challenges before India. International Development Economics Associates. Asia-Europe Dialogue & Partner - www.ased.org Directorate of Economics and Statistics (2006). “Statistics at a Glance : State-wise consumption of fertilizers” http://agricoop.nic.in/statistics/consum4a.htm (accessed January 20, 2006) FAOSTAT data, (2004). FAO World Agricultural Information Centre – last updated 20 December 2004. http://faostat.fao.org/faostat/form?collection=Production.Crops. Primary&Domain=Production &servlet=1&hasbulk=&version=ext&language=EN. (accessed on January 6, 2005) Finck A. (2002) : Fertilizers & their efficient use. Institut f. Pflanzenernaehrung u.Bodenkunde, University Kiel, Germany. Ghonsikar, C.P. & Shinde, V.S. (1997). Nutrient Management practices in Crops and Cropping Systems. Scientific Publishers, Jodhpur. International Fertilizer Industry Association (2004): SUMMARY REPORT-World Agricultural Situation and Fertilizer Demand, Global Fertilizer Supply and Trade 2003/04 – 2008/09. 72nd IFA Monthly Conference, Marrakech, Morocco. International Fertilizer Industry Association (2004): Micronutrients Symposium highlights stronger role for fertilizers. Fertilizers & Agriculture – IFA Journal, May 2004, p.p.3. Iyer, Ramaswamy R. (2003) : Linking of Rivers – Vision or Mirage. World Water Institute – River Links Dialogue. Kamath, K.V. (2002) : The Challenge of Interlinking India’s Rivers. Features – Samachar on Sify.com Kanwar, J.S. & Katyal, J.C. (1997). Plant Nutrient Needs, Supply, Efficiency and Policy Issues : 2000-2025. National Academy of Agricultural Sciences, New Delhi. Katyal, K.C. (1995) : Agronomic Management of Drought – The Indian Context. Global Grains Legumes Drought Research Network. News & Views - Volume 3 Issue 2. Khachatryan, A. & Von Oppen, M (2003). Market Access And Plant Productivity In Indian Agriculture. Deutscher Tropentag, Gottingen, Oct 8 – 10, 2003. Ministry of Commerce & Industry, Government of India (2001) : Conference of Chief Ministers on WTO Agreement on Agriculture & Food Management. May 30, 2001. http://commin.nic.in/doc/wto-may01.pdf. (accessed on January 6, 2005) Murthy, Laxmi (2004): Mapping Food Insecurity. InfoChange News and Features, February 2004. Rajesh, Kochhar (2000) : Water – The Ground Reality – Factors that caused the crisis. The Tribune, 22 May 2000.

25

Sen, Abhijeet (2001) : A Whole Crop of Uncertainties. Frontline – Vol 18, Issue 2. Sen, Abhijeet (2001) : Agriculture : A Long Road Ahead. The Hindu, July 19, 2001. Singh, R.B. (2003) : Land Use Change, Diversification of Agriculture & Agroforestry in North West India. Department of Geography, University of Delhi. http://www.icar.org.in/ncap/ publications/workshopprocedd/wsp5/chapter7.pdf. (accessed on January 6, 2005) Subramanyam K. V. & Sudha M. (2002) : Diversification of small farms through horticultural crops. Indian Institute of Horticultural Research. http://www.icar.org.in/ncap/publications/workshopprocedd/ wsp1/chapter11.pdf. (accessed on January 6, 2005) Statistical Outline of India (2003-04). Tata Services Limited, Department of Economics & Statistics, January 2004. Consumer Behaviour and Marketing Agarwal Nitin & Agarwal Manish (2003). “Theory of Trying – Implications for Marketing New-Concept Products.” IIMB Management Review, Dec 2003, pp. 15–21. Bagozzi, Richard P, and Paul R Warshaw, (1990) “Trying to Consume”, Journal of Consumer Research, 17, Sept, pp 127-133. Bentler, P. M. & G. Spechart (1979), “Models of Attitude-Behavior Relationships,” Psychological Review, 86, pp 452-464. Bomberger, William A. (1996) "Disagreement as a Measure of Uncertainty." Journal of Money, Credit and Banking 28 (Aug 1996): pp. 381-392. Carroll, Christopher D. (2002) “The Epidemiology of Macroeconomic Expectations”. Center on Social and Economic Dynamics, CSED Working Paper No. 25. http://www.brookings.edu/es/ dynamics/papers/epidemiology/epidemiology.htm (accessed January 3, 2006) Fishbein, M. & Ajzen, I. (1975), Belief, Attitude, Intention, and Behavior. Reading, MA: Addison-Wesley. Hair, J.F., Bush, R.P., Ortinau, D.J. (2006). Marketing Research within a Changing Information Environment, Third edition. Tata Mc-Graw Hill Publishing Company Ltd., New Delhi. Hawkins, D.I., Best, R.J., Coney, K.A. (2002). Consumer Behaviour : Building Marketing Strategy. Eighth Edition. Tata McGraw-Hill Publishing Company, New Delhi. Larréché, Jean-Claude and R. Moinpour (1983), “Managerial judgment in marketing: The concept of expertise,” Journal of Marketing Research, 20,110-121. Malhotra, Naresh K. (2001). Marketing Research : An applied orientation. Third edition. Pearson Education.

26

Morrison, D. G. (1979), “Purchase intentions and purchase behavior,” Journal of Marketing, 43, 65-74. Shapiro, Alan C. and Armstrong J. Scott (1974). “Analyzing Quantitative Models.” Journal of Marketing, Vol. 38, No. 2, (1974), pp. 61-66 Silk, S. J. & G. L. Urban (1978), “Pre-test market evaluation of new packaged goods: A model and measurement methodology,” Journal of Marketing Research, 171-191. Solomon, M.R. (2004). Consumer Behaviour : Buying, Having and Being. Sixth edition. Pearson Education. General Management & Strategy Armstrong, J.Scott (1967). “Derivation of Theory by Factor Analysis.” The American Statistician, 21 (December), 1967, pp 17-21. Ashton, Alison H. (1985), Does consensus imply accuracy in accounting studies of decision making? Accounting Review 60, 173-185. Callopy Fred and Armstrong J. Scott (1992). “Expert Opinions About Extrapolation And The Mystery Of The Overlooked Discontinuities.” International Journal of Forecasting 8, pp 575-582 Edwards, W. (1982). “Conservatism in human information processing”, in Kahneman, D., Slovic, P. & Tversky, A. (eds.), Judgment under uncertainty: heuristics and biases. Cambridge, UK: Cambridge University Press. Emery, F E and Trist, E L (1965). “The Causal Texture of Organizational Environments,” Human Relations, 18(1), pp 21-32. Hogarth, Robin M., "A note on aggregating opinions," Organizational Behavior and Human Performance, 21 (1978): pp 121-129. Khandwalla, P N (1977). The Design of Organization, New York: HBJ. McLarney Carolan (2003). “Strategic Planning Processes in Chaotic Environments: How to Calm a Turbulent Sea.” Vikalpa, Vol 28, No.1, Jan-Mar 2003, pp. 27–45. Ray, Sougata (2004). “Environment-Strategy-Performance Linkages: A Study of Indian Firms during Economic Liberalization” Vikalpa • Vol 29 • No 2 • April - June 2004, pp 9-23 Rich, Robert and Tracy, Joseph (2003). “Modeling Uncertainty: Predictive Accuracy as a Proxy for Predictive Confidence.” Federal Reserve Bank of New York Staff Reports, no. 161. Sekharan, Uma (2003). Research Methods for Business – A Skill Building Approach. Fourth edition. John Wiley & Sons, Inc. Walters, David (2003). “The Causal Texture of Organisational Environments Revisited”

27

Thompson, J D (1967). Organizations in Action: Social Science Bases of Administrative Theory, New York: McGraw-Hill. Planning and Forecasting Arkes, M. R. (2001), “Overconfidence in judgmental forecasting,” in J. S. Armstrong (Ed.) Principles of Forecasting. Norwell, MA: Kluwer Academic Publishers, pp. 495-515. Armstrong, J.Scott (1978). “Forecasting with Econometric Methods: Folklore versus Fact” Journal of Business, 51 (4), 1978, pp 549-564 Armstrong, J. Scott (1980). “The Seer-Sucker Theory: The Value of Experts in Forecasting.” Technology Review, June/July, 1980, pp 16-24. Armstrong J. Scott (1983). “Strategic Planning and Forecasting Fundamentals” in Kenneth Albert (ed.), The Strategic Management Handbook. New York: McGraw Hill, 1983, pp. 2-1 to 2-32. Armstrong, J. Scott (1985), Long-Range Forecasting: From Crystal Ball to Computer. John Wiley, New York (2nd ed.). Armstrong, J. Scott, Brodie Roderick and McIntyre, Shelby H. (1987). “Forecasting Methods for Marketing: Review of Empirical Research.” International Journal of Forecasting, 3 (1987), 335-376, North Holland. Armstrong, J. Scott (1989). Combining Forecasts: The End of the Beginning or the Beginning of the End? International Journal of Forecasting (1989), 5, pp. 585-588 Armstrong, J. Scott, and Collopy, Fred, (1992), “Error measures for generalizing about forecasting methods: Empirical comparisons,” International Journal of Forecasting, 8, 69-80. Armstrong, J. S. (2001), Principles of Forecasting. Boston: Kluwer Academic Publishers. Armstrong, Scott J. (2001). “Standards and Practices for Forecasting”. Principles of Forecasting: A Handbook for Researchers and Practitioners. Norwell, MA: Kluwer Academic Publishers, 2001 Armstrong J.Scott (2002). “Assessing game theory, role playing, and unaided judgment”, International Journal of Forecasting 18 (2002), pp 345–352. Carbone, R. and J.S. Armstrong (1982) “Evaluation of extrapolative forecasting methods: Results of a survey of academicians and practitioners,” Journal of Forecasting, 1, pp 215-217. Clemen, Robert T. (1989), “Combining forecasts: A review and annotated bibliography,” International Journal of Forecasting, 5, 559-583. Collopy, F., J.S. Armstrong (1993), “Causal Forces: Structuring Knowledge for Time-series Extrapolation,” Journal of Forecasting, 12 (1993), pp 103-115. Collopy, F., J.S. Armstrong and M. Adya (1994), “Principles for, examining predictive validity: The case of information systems spending forecasts,” Information Systems Research. 5, pp. 170-179.

28

Coopersmith, L.W. (1983) “Forecasting time series which are inherently discontinuous,” Journal of Forecasting, 2, pp 225-235. Dalrymple, D.J. (1987) “Sales forecasting practices,” International Journal of Forecasting, 3, pp 379-391. Fildes, Robert (1985), Quantitative forecasting - the state of the art: Econometric models, Journal of the Operational Research Society 36, 549-580. Fischoff, Baruch and D. MacGregor (1982), Subjective confidence in forecasts, Journal of Forecasting 1,155-172. Gettys, C. F., Kelly, C., & Peterson, C. R. (1982). “The best-guess hypothesis in multistage inference”, in Kahneman, D., Slovic, P. & Tversky, A. (eds.), Judgment under uncertainty: heuristics and biases. Cambridge, UK: Cambridge University Press. Green Kesten C.and Armstrong, J.S. (2005). “Demand Forecasting: Evidence-based Methods.” Forthcoming Chapter in : Strategic Marketing Management: A Business Process Approach, edited by Luiz Moutinho and Geoff Southern. Harrison, P.J. and C.F. Stevens (1971) “A Bayasian approach to short-term forecasting,” Operational Research Quarterly, 22, pp 341-362. Kalwani, M. U. &. Silk, A. J. (1982), “On the reliability and predictive validity of purchase intention measures,” Marketing Science, 1, 243-286. Kumar V., Morwitz Vicki G. And Armstrong J. Scott (2000). “Sales Forecasts for Existing Consumer Products and Services: Do Purchase Intentions Contribute to Accuracy?” International Journal of Forecasting, 16, pp 383-397 Mentzer, J.T. and J.E. Cox (1984) “Familiarity, application and performance of sales forecasting techniques,” Journal of Forecasting, 3, pp 27-36. Rowe, Gene and Wright, George (1999). “The Delphi technique as a forecasting tool: issues and analysis” International Journal of Forecasting 15 (1999), pp. 353–375 Schnaars, S.P. (1986) “Situational factors affecting forecast accuracy,” Journal of Marketing Research, 21, pp 290-297. Tsay, R. (1988) “Outliers, level shifts, and variance changes in time series,” Journal of Forecasting, 7, pp 1-20. Witt, S. and C. Witt, 1992, Modeling and Forecasting Demand in Tourism. Academic Press, London. Yokum Thomas J. and Armstrong J.S. (1995). “Beyond Accuracy: Comparison of Criteria Used to Select Forecasting Methods” International Journal of Forecasting, 11 (1995), pp 591-597. Yokum, J. Thomas, Collopy, F., J.S. Armstrong (2004). “Decomposition by Causal Forces: A Procedure for Forecasting Complex Time Series,” International Journal of Forecasting, 21, pp 25-36.