You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305808463

ARIMA Modelling for Forecasting of Rice Production: A Case Study of Pakistan

Article · December 2015

CITATIONS READS
2 756

4 authors:

Khalid Khan Gulawar Khan


Lasbela University of Agriculture, Water and Marine Sciences Lasbela University of Agriculture, Water and Marine Sciences
37 PUBLICATIONS   141 CITATIONS    16 PUBLICATIONS   52 CITATIONS   

SEE PROFILE SEE PROFILE

Sarfaraz Ahmed Shaikh Abdul Salam Lodhi


Indus Center for Sustainable Development Balochistan University of Information Technology, Engineering and Management Sci…
26 PUBLICATIONS   33 CITATIONS    8 PUBLICATIONS   17 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Federalism and nationalism View project

Financial regulations and banks View project

All content following this page was uploaded by Khalid Khan on 16 June 2017.

The user has requested enhancement of the downloaded file.


Lasbela, U. J.Sci. Techl., vol.IV , pp. 117-120, 2015 ISSN 2306-8256
RESEARCH ARTICLE

ARIMA Modelling for Forecasting of Rice Production: A Case


Study of Pakistan

Khalid Khan *1, Gulawar Khan 1, Sarfaraz Ahmed Shaikh2, Abdul Salam Lodhi 3
and Ghulam Jilani4
1
Faculty of SSM&IT, LUAWMS, Uthal, Balochistan
2
School of Economics, Huazhong University of Science and Technology (HUST), Wuhan, China
3
Faculty of Management Sciences, BUITEMS, Quetta
4
Faculty of Agriculture, LUAWMS, Uthal, Balochistan

Abstract:- This paper is based on empirical research which forecasts the rice production in Pakistan
wherein we have applied the Box Jenkins ARIMA methodology for forecasting. For this study, we used time
series data from 1993 to 2015. The diagnostic tests showed that ARIMA set (2, 1, 1) is more suitable
combination for forecasting. This study clearly indicates that forecasting is very useful for policy makers to
anticipate future needs of grain and export. Moreover, this would also prove helpful in shaping the national
policy of economic growth and self-sufficiency in food grain.

Key words: Rice Production; Autoregressive Integrated Moving Average (ARIMA), Forecasting

INTRODUCTION Rice is the second biggest staple grain


crops produced around Pakistan. It is
Pakistan is an agricultural therefore, Pakistan is one of the biggest
country. Agriculture is the backbone of exporters of rice. In the last fiscal year
Pakistan’s economy. For instance, in (from July 2014 to March 2015) the rice
2014-15 it (agriculture) accounts 20.9% export produced US$ 1.53 billion of
of the Gross Domestic Product (GDP) foreign exchange.3 It shows that rice
and 43.5% of the rural households are earned precious foreign exchange for
directly and indirectly engaged with Pakistan each year.
agriculture sector.1 In 2014-2015 rice was cultivated
Varieties of cash crops are grown in on 2891 thousands hectares of land,
Pakistan in its versatile weather which shows the increase of 3.6% in
throughout the year. Though, rice is one cultivable land under the rice crop,
of the important cash crops of Pakistan consequently this year (2014-2015)
and its contribution to agricultures’ GDP Pakistan produce 7005 thousand tons of
in the form of value added is 3.2%. It rice, the highest ever production of
contributes 0.7% to overall GDP.2 Pakistan with 3% of growth (ESP 2014-
_________________________________ 15). This vast increase in the rice
production is due to various factors such
Corresponding author: khalidk82@yahoo.com as the availability of water and its timely
use. The availability of water also
1
Economic Survey of Pakistan (ESP), Government of
Pakistan, (GoP) 2014-15
2 3
ESP, Government of Pakistan, 2014-1 ESP, Government of Pakistan, 2014-15

117
brought new land under cultivation 2030. As a result, this exercise will allow
including the use of hybrid rice varieties the policy makers to predict the future
and optimal use of scares resources. shortage of food grain in the country and
Rice is one of the main staple could take necessary measures in future
foods of the large portion of the world’s to avoid food shortage. Furthermore, the
population. It (rice) plays an important forecasts of rice production will also
role to meet the desperate needs of food support to save the scares and precious
demand both in urban and rural areas. A resources of the country.
lot of efforts have been made by the
government of Pakistan to increase the MATERIALS AND METHODS
productivity of rice to meet food demand
and boost export. The aim of this study is to project
Traditionally, forecasting has been the future expected values of rice
remained the most important part of production by applying ARIMA model.
structural econometric modeling because This study used the annual data of rice
generally it focus on the one-dimensional production from 1993 to 2015. There are
time series model known as a lot of econometrics techniques available
Autoregressive Integrated Moving for forecasting like Shryock et al (1973),
Average (ARIMA) models (see Box and Agrawal (2000) and Jan et al (2007).
Jenkins 1970). This model is widely They used different nonlinear, linear, first
applicable for time series forecasting, and high degree regression and
simulation and distributions (See Brown, exponential logistic regression for the
1959; Holt et al, 1960; and Saeed et al, purpose of forecasting but among them
2000). Furthermore, it is a generalize ARIMA model are widely used and
form of values weighted moving average acceptable.
exponentially. Therefore, we preferred to use
There are several approaches for ARIMA model for forecasting of rice
specific cases of ARIMA models production. It is because ARIMA model
identified by Box Jenkins (1970) and is one of the common techniques which
others. Mees and Geweke (1982) discuss make possible forecasting without the
the methods for univariate models while help of independent variables. ARIMA
others like Jenkins and Watts (1968), technique combines the two
Yule (1926), Quenouille (1949), and specifications into one equation i.e.
Bartlett (1964) emphasize the application Autoregressive Process (AR) and Moving
of ARIMA models. Average Process (MA).The ARIMA
In this paper we applied the model based on linear equation
ARIMA model for the forecasting of rice introduced by Box and Jenkins (1970).
production in Pakistan.4 It is a systematic The general structure of ARIMA
model and will allow us to predict the rice model is; ARIMA (p, d, q), where ‘p’ and
production for current year and onward to ‘q’ are the order of the autoregressive and
moving average process respectively
while ‘d’ is the order of the stationarity.
4
ARIMA model is one of the common techniques Hence, then the mathematical for the
which make possible forecasting without the help ARIMA (p, d, q) model is:
of independent variables. ARIMA technique
combines the two specifications into one equation
i.e. Autoregressive Process (AR) and Moving
Average Process (MA).

118
Zt  C  (F1Zt 1  .....  Fp Zt  p )  (1at 1  .....  qat q )  at different stochastic series are fitted on the
2nd order.
Z t 1.....Z t  p
Where: ‘C’ is the constant,

are lags and ‘


 i ’ are the parameters of the

equation t a
is the random error.
However, the AR model of order P is:

Z t  C  F1 Z t 1  F2 Z t 2   p at  p  at Table: I. Results of stationarity based


on ADF test
Similarly the MA structure of order q is: AV WT (-1)
TS 3.26 0.66 -3.28*
Zt  at  1at 1  2at  2 .....   p at  p P-Values -0.41 -0.80 -3.6**
Conclusion NST NST ST
RESULTS AND DISCUSSION Note: **p<0.05, * p<0.10, Actual variable (AV), With
Trend (WT), 1st difference (-1), Not stationary (NST),
Stationary (ST), Test Statistics (TS)
We used the annual time series
data of rice production from 1993 to 2015 In addition, we also applied
for forecasting purpose. All the data is different graphical validation techniques
extracted from Agriculture Statistics of i.e. histogram, normal probability plot
Pakistan (2014-15), Government of and residual plots. The ACF indicates the
Pakistan, Statistical Division, Pakistan values of q=1 because the correllogram of
Bureau of Statistics, Islamabad. ACF after lag one drops rapidly,
The forecasting of ARIMA model however, the correllogram of PACF drops
involved different steps such as promptly after lag two, hence the value of
identification of the model or p=2. Thus, the recommended
specification of the model. Firstly, we parsimonious ARIMA model for the rice
checked the date for stationary and data is ARIMA (2,1,1) whereas: p=1, d=1
examine the Autocorrelation Function and q=1.
(ACF) and Partial Autocorrelation Table II contains the estimated
Function (PACF) of the model. Secondly, parameters of ARIMA model (2,1,1) and
we used the estimation process of the its goodness of fit tests. The brief results
model. Thirdly, we checked the data are as under
diagnostically and come up with
forecasting. The Augmented Dickey Table:II. Estimated results of ARIMA
Fuller (ADF) test was applied to test the Model
data for stationarity. The table I offer the Type Parameters S.D TV
results of the ADF test, which indicates AR(1) -0.0513 0.1623 -0.3160
that the rice productivity is stationary at AR(2) -0.5408 0.1321 4.0936
fist difference i.e. I (1), therefore, for MA(1) 0.6100 0.1803 3.3830
ARIMA model, d=1’. N 27
Adjusted Box and Pierce  statistics
After the stationarity analysis, for 2
the identification of the model; we also
construct different plots of ACF and Lag 12 24 36
PACF.Hence, after the model 2 17.12(6) 29.21(17) 37.2(32)
identification the ARIMA model of
119
Note: S.D standard deviation, T-value (TV), Number of ARIMA (2, 1, 1). Hence, we did
observation (N)
forecasting for next 15 years on the basis
In table II row number seven reported the of this model. The forecasted values of
test statistics of adjusted Box and Pierce ARIMA (2, 1, 1) model shows that in
(1970) for rice production. The value of future the rice productivity will lead to
the test statistics is 17.8 for lag 12 while increase 40% till 2030 as compare to
the corresponding p-value is 0.7197 2015. Thus, the results of the study help
which indicates that it is insignificant at 5 the policy makers, ministry of agriculture
% level of significance. as well as chamber of commerce for
By means of ARIMA (2,1,1) the future planning and optimal use of scare
forecasted values of 15 years for rice resources.
production, with 95% of confidence
interval are reported in table II. REFERENCES
Bartlett, M.S., 1964. “On The Theoretical Specification
Table: III Forecasts for rice production of Sampling Properties of Autocorrelated
with confidence interval Time Series”. J. Roy. Stat. Soc., B 8: 27–41.
Year LV UV FV Brown, R.G., 1959. Statistical Forecasting For Inventory
2015 7025 7388 7,346 Control. New York, McGraw-Hill,
2016 7211 7853 7,828 Economic Survey of Pakistan, Government of Pakistan
2017 7412 7531 7,475 2014-15.
2018 7015 7100 6,554 Holt, C.C., F. Modigliani, J.F. Muth and H.A. Simon,
2019 7093 7251 7,151 1960. Planning, Production, Inventores, and
2021 7341 7651 7,520 Work Force. Prentice Hall, Englewood Cliffs,
2022 7300 7756 7,697 NJ, USA.
2023 7950 8318 8,219 Jenkins, G. M. and D.G. Watts, 1968. Spectral Analysis
2024 8102 8200 8,110 and its Application, Day, San Francisco,
2025 8201 8385 8,235 California, USA.
2026 8011 9688 9,624
2027 9236 9612 9,555 Meese, R. and J. Geweke, 1982. A Comparison of
Autoregressive Univariate Forecasting
2028 8515 8930 8,832 Procedures for Macroeconomic Time Series.
2029 8900 8671 8,208 Unpublished Manuscript, University of
2030 9211 9501 9,470 California, Berkeley, CA, USA.
Note: Authors’ own calculations, Production ‘000’tones
Source: Lower Value (LV), Upper Value (UV) and Saeed. N., Saeed. A. and Zakria. M., 2000. Forecasting
Forecasted Value (FV) of Wheat Production in Pakistan using Arima
Models, International Journal of Agriculture
& Biology., 4 352–353.
CONCLUSION
Yule, G.U., 1926. “Why Do We Sometimes Get
Nonsence-corrleations Between Times Series.
The study recommended that the A study in Sampling and the Nature of
most suitable and parsimonious model for Series”. J. Roy. Stat. Soc., 89: 1–69.
rice forecasting in case of Pakistan is
Received on 3rd March 2015, Accepted 25th Jun 2015
Manuscript can be viewed online on www.luawms.edu.pk

120

View publication stats

You might also like