International Review of Business and Finance

ISSN 0976-5891 Volume 2 Number 1 (2010), pp. 97–102

© Research India Publications
http://www.ripublication.com/irbf.htm

Use of the ARIMA Model for Forecasting Pigeon Pea

Production in India

Rachana Wankhade1, Suvarna Mahalle2, Sonal Gajbhiye3,

and V.M. Bodade4
1,2,3,4
Senior Research Fellow (Agril. Econ.) Dr. PDKV, Akola, India

Abstract

Pigeon pea is an important pulse crop in India and being the largest producer
of Pigeon pea in the world, India contributes to around 85 percent of the
world’s total production. It is imperative to assess scientifically the accurate
future production potentials of this crop on the basis of past trends. Thus, the
study has been made to forecast the production of Pigeon pea in India up to
the year 2015.by using ARIMA (autoregressive integrated moving average)
model. At present it is the most sophisticated model for forecast. The results
showed that production of Pigeon pea in India would be 2.73452 M. Tons in
2015. In the year 2009 price of Tur dal touched the sky, price of any
commodity ultimately related to production. So it is necessary to study the
future movement of Pigeon pea production.

Keywords: ARIMA; Forecasting; Production; Pigeon pea; India.

Introduction
Pigeon pea is an important pulse crop in India and being the largest producer of
Pigeon pea in the world, India contributes to around 85 percent of the world’s total
production. It is a protein rich staple food and consumed in the form of split pulse as
Dal. Pigeon pea is second largest pulse crop in India, accounting about 20 percent of
total pulse production, The crop ranks sixth in the world in dry land legume
production.
Auto Regressive Integrated Moving Average (ARIMA) is the most general
class of models for forecasting a time series. Different series appearing in the
forecasting equations are called “Auto-Regressive” process. Appearance of lags of
the forecast errors in the model is called “moving average”. (ARIMA) model was
introduced by Box and Jenkins in 1960 for forecasting variables. In the year 2009
price of Tur dal touched the sky, price of any commodity ultimately related to
98 Rachana Wankhade et al

production. So it is necessary to study the future movement of Pigeon pea

production by using past trends

Methodology
Respective time series data for this study were collected from Government
Publications, (Agricultural Statistics at a glance,2008).Box and Jenkins (1976) linear
time series model was applied. Auto Regressive Integrated Moving Average
(ARIMA) is the most general class of models for forecasting a time series. Different
series appearing in the forecasting equations are called “Auto-Regressive” process.
Appearance of lags of the forecast errors in the model is called “moving average”
process. The ARIMA model is denoted by ARIMA (p,d,q), where “p” stands for the
order of the auto regressive process, ‘d’ is the order of the data stationary and ‘q’ is
the order of the moving average process.
Δdyt =δ+θ1 Δdyt-1+θ2 Δdyt-2+----- + θpyt-p+e t-1 α et-1-α2 et-2 αq et-2 (1)
Where, Δd denotes differencing of order d,i.e.,
Δyt = yt-yt-1, Δ2yt=Δyt-Δt-1 and so forth, Y t-1 ----- yt-p are past observations
(lags), δ,θ1 -------- θp are parameters (constant and coefficient) to be estimated similar
to regression coefficients of the Auto Regressive process (AR) of order “p” denoted
by AR (p) and is written as
Y = δ+θ1 y t-1+θ2 y t-2 + ---------- + θpy t-p +et (2)
Where,et is forecast error, assumed to be independently distributed across time
with mean θ and variance θ2e, et-1, et- 2 ------ et-q are past forecast errors, α1, --------
-- αq are moving average (MA) coefficient that needs to be estimated. While MA
model of order q (i.e.) MA (q) can be written as
Yt = et-α1 αt-1 - α2et-2 ------------- αqet-q (3)
The major problem in ARIMA modeling technique is to choose the most
appropriate values for the p, d, and q. This problem can be partially resolved by
looking at the Auto correlation function (ACF) and partial Auto Correlation Functions
(PACF) for the series (Pindyk & Rubinfeld, 1991). The degree of the homogeneity,
(d) i.e. the number of time series to be differenced to yield a stationary series was
determined on the basis where the ACF approached zero.
After determining “d” a stationary series Δd yt its auto correlation function and
partial autocorrelation were examined to determined values of p and q, next step was
to “estimate” the model. Using the results of ARIMA (p,q,d), forecast the Tur
production in India from 2008 up to 2015

Result and Discussion

To fit an ARIMA model requires a sufficiently large data set. In this study, we used
the Pigeon pea production data for the period 1950-51 to 2007-2008. The maximum
production of Pigeon pea in India was obtained in 2007-08 year (3.09 million tones)
and minimum in 1950-51 year (1.72 million tones). The ARIMA model was applied
according to three steps namely Identification, Estimation and Verification.
Last fifty seven years data of Pigeon pea production was used for modeling
Use of the ARIMA Model for Forecasting Pigeon Pea Production in India 99

purpose. The model specification involved the plots of the auto correlation function
(ACF), partial auto correlation function (PACF) and the plot of the differenced series.
Auto correlation function indicated the order of the autoregressive components ‘q’ of
the model, while the partial correlation function gave an indication for the parameter
p. First step was to check the stationarity of the data. The time series plot for
production showed an increasing trend. Auto correlation function of the series showed
non-stationary as auto correlation function did not fall as quickly as the log (k)
increased. To make the series stationary for production, difference series was used
and first difference series of Pigeon pea production showed stationarity. After this the
values of autoregressive (AR) parameter “p” and moving average (MA) parameter ‘q’
was determined from correlograms of partial autocorrelation function and the auto
correlation function, respectively. Partial auto correlation function of the first
differenced series of production was used to determine the parameter “p”. It was
observed that partial auto correlation function fell after lag 2. Thus the value of “p”
was decided 2 for production which gave good results consequently, the respective
values of p,d,q were determined for ARIMA i.e. ARIMA (2,1,0)

Model estimation
With the help of SPSS computer package ARIMA (1, 1, 1) models was found to be
estimated for Pigeon pea Production in India.

Diagnostic checking
For diagnostic checking of the estimated models, different diagnostic checks were
applied for whether these were properly fitted or not. Which is given in table 1.

Table 1: Values of AIC and BIC.

ARIMA (p d q) AIC BIC

111 19.79 25.92
211 21.03 29.20
212 21.32 31.54
101 23.49 29.67

It is shown from the table that the most suitable model is ARIMA (1, 1, 1) this
model has the lowest AIC and BIC values

Residual Analysis
One of the indicators of the properly fitted model is that of scattered residuals in a
rectangular shape around the zero at horizontal level. The time series plot of residuals
of production data showed scatter trend, therefore, models were fitted properly by
residual
100 Rachana Wankhade et al

Table 2: Estimates of the fitted ARIMA model.

Estimates Std Error T Approx sig

Non- Seasonal AR1 -0.15554435 0.18045731 -0.8619454 0.39252919
lag MA1 0.70368794 0.13216211 5.3244301 0.00000202
Constant 0.01569487 0.00989606 1.5859713 0.11858501
Number of Residual 57
Number of parameters 2
Residual Df 54
Adjusted Residual sum of 4.2483160
Squares
Residual sum of Squares
Residual variance 0.07741866
Model std. Error 027824208
Log – likelihood -6.8987496
Akaike’s Information 19.797499
Criterion (AIC)
Schwartz’s Bayesian 25.926653
Criterion (BIC)

Table 3: Autocorrelations and partial autocorrelations of residuals.

Lag Autocorrelation Std. Error Box-Ljung Statistic df Sig.

1 -0.028 0.129 0.046 1.000 0.830
2 -0.077 0.128 0.413 2.000 0.813
3 0.053 0.127 0.589 3.000 0.899
4 0.041 0.126 0.695 4.000 0.952
5 0.146 0.124 2.077 5.000 0.838
6 -0.018 0.123 2.099 6.000 0.910
7 -0.052 0.122 2.279 7.000 0.943
8 -0.078 0.121 2.696 8.000 0.952
9 0.091 0.119 3.282 9.000 0.952
10 -0.100 0.118 4.005 10.000 0.947
11 -0.060 0.117 4.267 11.000 0.961
12 -0.026 0.116 4.318 12.000 0.977
13 -0.117 0.114 5.369 13.000 0.966
14 -0.006 0.113 5.371 14.000 0.980
15 0.050 0.112 5.571 15.000 0.986
16 -0.052 0.110 5.790 16.000 0.990
Use of the ARIMA Model for Forecasting Pigeon Pea Production in India 101

Lag Partial Autocorrelation Std. Error

1 -0.028 0.132
2 -0.078 0.132
3 0.049 0.132
4 0.038 0.132
5 0.158 0.132
6 -0.006 0.132
7 -0.034 0.132
8 -0.106 0.132
9 0.071 0.132
10 -0.132 0.132
11 -0.034 0.132
12 -0.041 0.132
13 -0.096 0.132
14 -0.034 0.132
15 0.077 0.132
16 -0.028 0.132

Figure 1: ACF of residual of fitted Figure 2: PACF of residual of fitted

ARIMA model. ARIMA model.

Table 4: Forecasted values of Pigeon pea production in India up to year 2014-15.

Year Estimated Production Lower CL Upper CL

2008-09 2.56797 2.00479 3.13115
2009-10 2.66730 2.09709 3.23752
2010-11 2.66999 2.07604 3.26394
2011-12 2.68771 2.07331 3.30210
2012-13 2.70309 2.06784 3.33833
2013-14 2.71883 2.06286 3.37480
2014-15 2.73452 2.05787 3.41116
102 Rachana Wankhade et al

Forecasts of Pigeon pea production

ARIMA (1,1,1) was taken for seven year ahead and forecasts for Pigeon pea
production which are given in table 4 along with 95% confidence interval values.
Forecasts of Pigeon pea production will increases to some extent in future i.e. 2008-
09 is 2.49479 million tones up to the year 2014-2015 it will be accepted 2.73452
million tones. With lower and upper limits of 2.05787 million tones and 3.41116
million tones respectively.

Conclusions
ARIMA model offers a good technique for predicting the magnitude of any variable.
Its strength lies in the fact that the method is suitable for any time series with any
pattern of change and it does not require the forecaster to choose a prior value of any
parameter. Its limitations include requirement of a long time series. Often it is called a
13 ‘Black Box’ model. Like any other method, this technique also does not guarantee
perfect forecasts. Nevertheless, it can be successfully used for forecasting long time
series data.
In our study ARIMA (1, 1, 1) model was best suited for estimation of Pigeon pea
production data. From the forecast values obtained the developed model, it can be
said that forecasted production will increases to some extent in future i.e. In 2008-09
production of Pigeon pea was 2.49479 million tones up to the year 2014-2015 it will
be 2.74 million tones.

References
[1] A JRG Commodities Research Publication,2009,”Soybean and Soy oil Price
Forecast 2009-10. Commodities Research Desk, JRG Wealth Management
Limited, JRG House, Cochin-682071 Kerala, India.
[2] B.N. Mandal, 2005. Forecasting sugarcane Production in India with ARIMA
model. Interstat. www.interstat.statjournals.net.
[3] Box, G.E.P., and G. M. Jenkins. 1970. Time series analysis: forecasting and
control. Holden Day, San Francisco, CA.
[4] Government of India, 2008. Agricultural statistics at a glance, 2008. Dept of
Agriculture and cooperation, Ministry of Agriculture. N. Delhi
[5] Winters, L. A. and Sapsford, D. 1990. Primary Commodity Prices: Economic
Models and Policy, Cambridege University Press, Cambridge.