Research Article

Major Crops Forecasting Area, Production and Yield Evidence from Agriculture Sector of Pakistan

Dilshad Ahmad*, Muhammad Irfan Chani and Asad Afzal Humayon

Department of Management Sciences, COMSATS Institute of information Technology Vehari Campus, Vehari, Pakistan.

Received | March 16, 2017; Accepted | July 26, 2017; Published | August 10, 2017

*Correspondence | Dilshad Ahmad, Department of Management Sciences, COMSATS Institute of information Technology Vehari Campus, Vehari, Pakistan; Email: [email protected]

Citation | Ahmad, D., M.I. Chani and A.A. Humayon. 2017. Major crops forecasting area, production and yield evidence from agriculture sector of Pakistan. Sarhad Journal of Agriculture, 33(3): 385-396.

DOI | http://dx.doi.org/10.17582/journal.sja/2017/33.3.385.396

Keywords | Wheat, Rice, Maize, Cotton, Sugarcane, Forecasting

Introduction

According to United Nation (UN), millennium declaration of Millennium Development Goals (MDGs) “Eradicate Extreme Hunger and Poverty” and now food security has become a global concern in the current scenario of the world. It has anticipated world population will increase around 9.3 billion in 2050 and the immense increase in the population of Pakistan is expected in coming decades (UN, 2011). Pakistan population is currently 19.1 million 2015 and anticipated to increase 242.06 million in 2030 and 300 million in 2050 (PBS 2014-15). Increasing population requires more food, clothing, shelters and other necessities of life. Dynamics of population stresses economic resources and dependency of young on the economy of the country. Agriculture is a major source of providing food and employment to the population of the country. Major crops facilitate in the provision of food basket to the population. There is direct relation between population growth and food requirements, so it is necessary to focus on enhancing the agricultural productivity.

Agriculture is not only a major source direct employment, it also provides raw material to export oriented industries of Pakistan. Agriculture sector contributes 20.9% of GDP and employing a source of 43.5% of rural population. Official statistics reveals that growth rate of agriculture sector rose from 2.7 percent in 2013-14 to 2.9 percent in 2014-15, but the alarming fact is that the production of major food crops showed negative growth during the same period (PBS 2014-15).

Major crops of agriculture sector wheat, rice, cotton, maize, and sugarcane have reported 25.6% of value added in overall agriculture and 5.3% of GDP in Pakistan. Wheat crop is considered major crop in agriculture sector by cultivation area and it contributes 10% of value added in agriculture and 2.1% of GDP. A decrease in wheat cultivation area and production is observed. The wheat cultivation area reduced from 9199 thousand hectares in 2013-14 to 9180 thousand hectares in 2014-15. Wheat production decreased from 25979 thousand tonnes in 2013-14 to 25478 thousand tonnes in 2014-15. Rice crop is accounted for 0.7% of GDP and 3.2% agriculture value added. Increase in rice cropped area and production has recorded. Area rose from 2789 to 2892 thousand hectares and rice production from 6798 to 7005 thousand tonnes from during the fiscal year 2014-15. Maize is a prominent food crop, which accounts for 0.4% of GDP and 2.1% agriculture value added. Maize cropped area and production have decreased. Area shrank to 1130 from 1168 thousand hectares declined to 4695 from 4944 thousand tonnes (PBS, 2014-15).

Cotton crop performs a prominent role in foreign exchange earning exporting and provision of raw material to local textile industry. Cotton crop shares 1.5% of GDP and 7.1% of value added in agriculture. Cotton cropped area has increased to 2961 thousand hectares to 2806 thousand hectares increase and its production rose to 13.983 million bales from 12.769 million bales from 2013-14 to 2014-15 (PBS, 2014-15). Sugarcane is considered a cash crop and it contributes 0.6% of GDP and 3.1% of value added in agriculture. Decrease in sugarcane crop area and production has been reported from 2013-14 to 2014-15. The cropped area reduced to 1141 thousand hectares from 1173 thousand hectares decrease and its production decreased to 62.7 million tonnes from 67.5 million tonnes during the year (PBS, 2014-15).

Intensification in population is a critical issue to agrarian economies justifying nutrition requisite to masses of the country and raw material to the industrial sector. Low productivity in agriculture sector is a crucial one to developing economies. Developing agricultural countries have to import cereals to meet population nutrition requirements. Numerous factors are playing a role in such marginal rotations. Environmental changes are playing a pivot role in agricultural productivity. Environmental changes and policy matters are playing a key role in agro-based decisions in the current scenario. Forecasting plays a vital role in notifying future availability of resources and requirements of the country and vision to adjusting gaps and policy formulation to meet future requisites. Forecasting estimates of major crops will provide supportive measures to state, formulating policies relevant to trade, prices, and protective measures surplus and scarcity of food and raw material in future. In the agriculture sector, major crops are the main source of food staple to masses and provision of foreign exchange earnings and raw material to the industrial sector of the country. The study has focused on forecasting major crops (area, production, and yield) and per capita food availability for 21 years 2014-15 to 2034-35. Major crops are playing a vital role in the agriculture sector of Pakistan. This study has employed ARIMA model forecasting interlinked with previous sixty-seven years data of major crops from 1947-48 to 2014-15.

Literature review

Agriculture has foremost worth the economic growth of developing countries. Major crops impact economic growth of developing countries it is notable to be aware of production and yield of crops for economic stability. Forecast food availability and population growth rate are consider as prerequisites for preventing alarming food insecurity in future and formulating policy measures for a viable solution. Most of studies have forecasted and detected curtailed in food crops and cash crops production in Pakistan. Hamid et al. (1987) and Muhammad (1989) reported forecast play vital role in growth in area and yield of major crops except some limits. Numerous studies have forecasted major crops and vegetables with a variety of econometric models, Exponential Smoothing Approach, Regression Analysis and ARIMA model. ARIMA model approach has remained the focus of massive studies for the forecasting object. Muhammad (1992) has applied ARIMA model approach to forecast rice production in Pakistan and suggest proper measures to increasing exports. Saeed et al. (2000) have forecasted wheat crop area and production employing ARIMA model. Karim et al. (2005) have forecasted wheat crop production through Regression Analysis of Bangladesh. Sabir and Tahir (2012) have forecasted wheat crop production with supply-demand projection using Exponential Smoothing model for Pakistan. Badmus et al. (2011) study has forecasted maize cultivated area and production in Nigeria with the application of ARIMA model. Mehmood and Ahmad (2013) employed ARIMA model and investigated a forecasted area of mangoes in Pakistan while Iqbal et al. (2014) forecasted trading behaviour of food crop of four leading SAARC countries. Arivarasi et al (2015) study has forecasted vegetable trends and production of two vegetable growing zones in the Chennai state of India using ARIMA model. Badar et al. (2015) have used ARIMA model to forecast major food crops including wheat, rice, and maize, area, production and yield in Pakistan. Ali et al. (2015) have forecasted production and yield of cotton and sugarcane using ARIMA model.

In literature, a bulk of studies has focused, forecast of various crop wheat, rice, sugarcane, maize, cotton, and vegetables. Plenty of work has provided a fruitful contribution to literature while gap still prevails to the significant impact of forecasting crops. Numerous studies have fixated production and yield of crops while omitted cropped area forecasting except study of Badar et al. (2015). Cropped area plays a vital role in agriculture sector productivity and preliminary element in resource allocation. Studies in the literature have focused only limited major crops single crop, cash crops or food crops and not simultaneously forecasted all major crops. Studies have forecasted area, production, and yield while no justification of forecasted food availability per capita except study of Abbas et al. (2007) and Zulfiqar and Hussain (2014) forecasted wheat production gaps to approach state future food security in Pakistan. Research gap prevails to forecast major food crops and per capita food availability. In this study, five major crops are simultaneously forecasted area, production, and yield for 21 years 2014-2035. Per capita food availability of food crops forecasted to provide deep insight to bridge such research gap. Exponential Smoothing Approach or Regression Analysis has rarely used to forecast in studies. Plentiful studies in literature preferred ARIMA model as Azhar et al. (1972), Boken (2000), Saeed (2000), Muhammad (2001), Iqbal (2005), Karim (2005), Badmus (2011), Sabir and Tahir (2012), Ali et al. (2013), Mehmood and Ahmad (2013), Abid et al. (2014), Amin (2014), Zulfiqar and Hussain (2014), Ali (2015), Badar (2015) have applied ARIMA model to forecast based on valid estimation. ARIMA approach considers preferable for forecasting due to such feathers deal any type of data and estimation comparatively to other approaches.

Matrials and Methed

For the purpose of forecasting of production, area and yield of major crops, i.e. wheat, rice, maize, cotton and sugarcane, the study has used the long range of annual time series data. Our data set covers almost all history of Pakistan and it ranges from 1947-48 to 2013-14. The last few years are skipped due to unavailability of reliable statistic at the time of data analysis. The data for all three dimensions, including production, area and yield, of five major crops has been collected from various statistical supplements of Agriculture Statistics of Pakistan published by Pakistan Bureau of Statistics (PBS).

Keeping in view the aims of the study and nature of statistical information, various methods of forecasting have been considered at the time of data analysis. The Autoregressive Integrated Moving Average (ARIMA) model has chosen for analysis due to its suitability to our dataset and non-stationary nature of time series to be forcasted. ARIMA is univariate model which has several advantages over its multivariate alternates. This model suitable even for non-stationary time series, has great statistical power for reliable forcasting from small datasets and requires data for only time series being forcasted but not for its determinants (Brockwell and Davis, 2003; Hamilton, 1994; Tsay, 2005; Wei, 2005).

The major issue in the use of ARIMA is selection of appropriate ARIMA(p,d,q) models where ‘p’ refers to the number of autoregressive terms, ‘d’ represents order of integration and ‘q’ denotes the moving average terms to be selected for more precise forcast. For the purpose of model selection Box and Jenkins (1970) approach is used. This approach helps in ARIMA model identification, estimation, diagnostic checking and forecasting of both stationary and non-stationary time series. This approach differentiates ARIMA and ARMA model for stationary and non-stationary time series respectively.

In Box-Jenkins approach, stationarity or order of integration (d) of the data is checked through formal unit root tests like Augmented Dickey-Fuller (ADF) test, Phillips-Perron (PP) test etc. Graphs of Partial Autocorrelation Function (PACF) and

Table 1: Augmented dickey fuller test (adf) unit root test results.

Autoregressive (AR) general form according to order (p) as mentioned in Equation (1).

Model AR (p) explained with “t” time

Xt : Value of response variable at time “t”; Xt-1, Xt-2…Xt-p: Symbolize different time lags of respective variable; λ0, λ1, λ2…..λp: Signify coefficients of the model; Ut: Mentions error term of time “t”.

Moving Average MA (q) generalized form mentioned in Equation (2).

While βt as constant mean of series

ψ1+ …ψq: Mentioned coefficient of estimated error term; zt: error factor.

After combining Autoregressive and Moving Average models formed as ARMA model (Equation 3)

Use of differenced time series (∆Xt) according to order of integration, like I (1) first-differenced stationary time series forms the ARIMA model.

Results and Discussion

This section operationalized the proposed ARIMA method used for forecasting in the study. As a first step of Box-Jenkins approach for ARIMA model selection, ADF unit root test is applied for checking the stationarity or non-stationarity of the time series being forecasted. Sixty-seven empirical annual observations for each series are used in this analysis and forecasting. Stationarity or non-stationarity check through unit root analysis will help us in choosing between ARMA and ARIMA for further processing and forecasting. The estimates of unit root test applied on the series of production, area and yield of wheat, cotton, sugarcane, rice and maize are reported in the Table 1.

Table 2: Phillips perron unit root test.

Table 3: Estimation of wheat area, production and yield.

Table 4: Estimation of rice area, production and yield.

Estimates of unit root test in Table 1 indicate that all fifteen variables consisting of wheat crop area, production, yield, rice crop area, production, yield, maize crop area, production, yield, cotton crop area, production, yield and sugarcane crop area, production and yield are non-stationary at level based on ADF test statistics and level of significance. This indication suggests that ARMA model is not suitable for forecasting of these variables and we will have to move for ARIMA models for reliable forecasting. The results of the last three columns reveal that all the variables became stationary when they are used in first differenced form. It suggests that ARIMA (p, 1, q) model should be used for forecasting as all the variables have order of integration I (1).

Wheat crop area, production and yield estimation are mentioned with appropriate models in Table 2. According to Akaike Information Criteria(AIC), ARIMA(2,1,4) model for wheat area, ARIMA(1,1,2) model for wheat production and ARIMA(1,1,2) model for wheat yield are considered appropriate models for forecasting of these three variables.

Estimation of rice crop area, production, and yield are presented in Table 3 with selected models. In this table, ARIMA (3, 1, 4) model for rice area, ARIMA (0,1,1) model for rice production and ARIMA(0,1,1) model for rice yield are selected as appropriate models for forecasting of these variables.

Table 5: Estimation of maize area, production and yield.

Table 6: Estimation of cotton area, production and yield.

Findings of appropriate ARIMA models of maize crop area, production and yield are reported in Table 4. ARIMA(0,1,4) model for maize area, ARIMA(0,1,1) model for maize production and ARIMA(2,1,4) model for maize yield are selected as appropriate models for forecasting of these series.

Table 5 reported appropriate ARIMA models and their estimates for cotton crop area, production and yield. ARIMA(4,1,2) model for cotton area, ARIMA(4,1,1) model for cotton production and ARIMA(0,1,2) model for cotton yield are chosen as suitable forecasting models for cotton related variables.

ARIMA models for sugarcane area, production, yield and their estimates are given in Table 6. ARIMA (2,1,1) model for sugarcane area, ARIMA(3,1,1) model for sugarcane production and ARIMA(0,1,3) model for the sugarcane yield are considered appropriate models for forecasting sugarcane area, production and yield respectively.

Diagnostic checking of fitted models is considered next step in pursuit of models fitted and estimated through Box-Jenkins (1976) approach. Autocorrelation function (ACF) and Partial Autocorrelation function (PACF) of plotted residual pointed out fitted model. For major crops area production and yield ACF and PACF of plotted residual are found within limits and proves fitness of models (see Appendix-A). Forecasting of major crops area, production, and yield is made through above estimated ARIMA Models. Major crops forecasted values of wheat, rice, maize, cotton, sugarcane, area, production and yield for the year 2014-15 to 2034-35 presented in Table 7.

Wheat crop forecast values indicate that wheat cropped area will reach to 11927 thousand hectares, wheat production will rise to 47797 thousand tonnes and wheat yield will be increased to 4007 kg/ hectare till the year 2035. These forecast estimates are represented in Figure 1. These forecast findings are consistent with the studies of Iqbal et al. (2005), Sher and Ahmad (2008), Amin et al. (2014) and Badar et al. (2015).

Table 7: Estimation of sugarcane area, production and yield.

Rice crop forecasted values are shown in Figure 2. The forecasted rice area will reach to 3383 thousand hectares, rice production will rise to 13103 thousand tonnes and rice yield will increase to 3881kg/ hectare by the year 2035. These findings are accounted parallel to the study of Badar et al. (2015).

Figure 3 shows the maize crop forecasted values. Maize area will be 1343 thousand hectares, maize production will increase to 11251 thousand tonnes

and maize yield will rise to 8377 kg/ hectare till 2035. These forecasted findings are similar to the studies of Badar et al. (2015).

Cotton crop forecasted area, production, and yield are reported in Figure 4. Forecasted values of cotton crop area will be 3410 thousand hectares, cotton production will rise to 16755 thousand bales and cotton yield will increase to 836 kg/ hectare by 2035. All these estimates are consistent with Ali et al. (2015).

As the Figure 5 above have reported the Sugarcane forecasted area, production, and yield. Sugarcane forecasted area will rise to 1421 thousand hectares, sugarcane production will increase to 77257 thousand tonnes and sugarcane yield will be 54368 kg/ hectare by the year 2035. These reported estimates are similar to the studies of Masood and Javid (2004), Yaseen et al. (2005) and Ali et al. (2015) (Table 9).

Forecasted values of the population for the year 2014 to 2035 are shown in Figure 6, which indicate that forecasted population would be 258.4 million in 2035.

Per capita forecasted estimates of food crops like wheat, rice and maize are given in Figure 7. Availability of per capita wheat will be 185 kg, rice 50.8 kg and maize 43.5 kg for the year 2035 such findings are aligned with study of Zulfiqar and Hussain (2014).

Table 9: Forecasted estimation of per capita food of major food crops from 2014-2035.

There is supplementary material associated with this article. Access the material online at: http://dx.doi.org/10.17582/journal.sja/2017/33.3.385.396

References

Abbas, M., A.D. Sheikh, M. Shahbaz and A. Afzaal. 2007. Food security through wheat productivity in Pakistan. Sarhad J. Agric. 23: 1240-1247.

Abid, S., I. Raza, A. Khalil, M.N. Khan, S. Anwar and M. A. Masood. 2014. Trend analysis and forecasting of maize area and production in Khyber Pakhtunkhwa, Pakistan. European Academic Research. 2(4): 4653-4664.

Ali, S., N. Badar and H. Fatima. 2015. Forecasting production and yield of sugarcane and cotton crops of Pakistan for 2013-30. Sarhad J. Agric. 1 (31).

Ali, H., H. Ali, Z. Faridi, and H. Ali. 2013. Production and forecasting trends of cotton in Pakistan: an analysis view. J. Basic Sci. Res. 3(12): 97-101.

Amin, M., M. Amanullah and A. Akbar. 2014. Time series modeling for forecasting wheat production of Pakistan. J. Anim. Plant Sci. 24(5):1444-1451.

Arivarasi, R and M. Ganesan. 2015. Time series analysis of vegetable production and forecasting using ARIMA model. Asia. J. Sci. Technol. 06 (10): 1844-1848.

Azhar, B.A., G.M. Chaudhry and M.A. Shafiq. 1972. A model for forecasting wheat production in the Punjab. Pak. Dev. Rev. 11(2): 407-415.

Badmus, M.A and O.S. Ariyo. 2011. Forecasting cultivated areas and production of maize in nigeria using ARIMA model. Asia. J. Agric. Sci. 3(3): 171-176.

Badar, N., H. Fatima., A. Jabbar and M. Asif. 2015. Major food crops production and yield forecast in Pakistan. Pak. J. Agric. Res. 28: 295-303.

Boken, V.K. 2000. Forecasting spring wheat yield using time series analysis: A Case Study of Canadian Prairies. Agric. J. 92(6): 1047-1053. https://doi.org/10.2134/agronj2000.9261047x

Box, G. and G. Jenkins. 1970. Time series analysis: Forecasting and Control, Holden-Day, San Francisco.

Box, G. and G. Jenkins. 1976. Time series analysis: Forecasting and control, 2nd edition. Holden-Day, San Francisco.

Brockwell, P.J. and R.A. Davis. 2003. Introduction to time series and forecasting. Springer. New York.

Dickey, D.A. and W.A. Fuller. 1979. Distribution of the estimator for autoregressive time series with a unit-root. J. Am0 Stat. Assoc. 74: 427-431. https://doi.org/10.1080/01621459.1979.10482531

Hamid, N., P. Thomas, V. Alberto and G. Suzanne. 1987. The wheat economy of Pakistan setting and prospectus. USAID/ International food policy research institute, Washington DC. pp: 140.

Hamilton, J.D. 1994. Time Series Analysis. Princeton University Press, Princeton, 7-71.

Iqbal, M.Z. Hussain and T. Mahmood. 2014. Forecasting trade behaviour of major food crops in four leading SAARC countries. Pak. Economy. Soc. Rev. 52(2): 209-226.

Iqbal, N., K. Bakhsh. K. Maqbool and A.S. Ahmad. 2005. Use of the ARIMA model for forecasting wheat area and production in Pakistan. Int. J. Agric. Biol. 2: 352-354.

Karim. R, A. Awal and M. Akhtar. 2005. Forecasting of wheat production in Bangladesh. J. Agric. Soc. Sci. 1:120-122.

Masood, M.A. and Javid, M.A. 2004. Forecast models for sugarcane in Pakistan. Pak. J. Agric Sci. 41: 1-2.

Mehmood, S. and Z. Ahmad. 2013. Time series model to forecast areas of mangoes from Pakistan: An application of univariate ARIMA model. Acad. Contemp. Res. J. 2(1): 10-15.

Meyler, A., G. Kenni and T. Quinn. 1998. Forecasting Irish inflation using arima models. Economic analysis, Research and Publications Department, Central Bank of Ireland, Dublin. pp: 46.

Muhammad, F., M. Siddique, M. Bashir and S. Ahmad. 1992. Forecasting rice production in Pakistan using ARIMA models. J. Anim. Plant Sci. 2: 27-31.

Muhammad, K. 1989. Description of the historical background of wheat improvement in Baluchistan, Pakistan (1909-1980) Rachis. 8: 10-12.

Muhammad, N. 2001. Factor affecting area and yield of wheat crop a case study. Pak. J. Agric. Econom. 4(2): 66-76.

Pakistan Bureau of Statistics. 2014-15. Economic Survey of Pakistan, statistical supplement finance division, Economic Advisors Wing, Islamabad.

Sabir, H.M. and S.H. Tahir. 2012. Supply and demand projections of wheat in Punjab for the Year 2011-12. Interdisciplin. J. Contemp. Res. Bus. 3: 800-808.

Saeed, N., A. Saeed., M. Zakir and T.M. Bajwa. 2000. Forecasting of wheat production in Pakistan using ARIMA models. Int. J. Agric. Biol. 2: 352-353.

Sher, F. and E. Ahmad. 2008. Forecasting wheat production in Pakistan. Lahore J. Econom. 13(1): 57-85.

Tsay, R.S. 2005. Analysis of Financial Time Series. Wiley New York. https://doi.org/10.1002/0471746193

United Nations. 2011. World population prospects the 2010 revision: Volume I: Comprehensive Tables. New York: Department of Economic and Social Affairs, Population Division, United Nations.

Wei, W.W.S. 2005. Time Series Analysis: Univariate and Multivariate Methods. 2nd Edition. Addison-Wesley, New York.

Yaseen, M, M. Zakria, I. Shahzad, M.I. Khan and M.A. Javed. 2005. Modeling and forecasting the sugarcane yield of Pakistan. Int. J. Agric. Biol. 7(2): 180-183.

Zulfiqar, F., and Hussain. A. 2014. Forecasting Wheat production gaps to assess the state of future food security in Pakistan. J. Food Nut. Disord. 3:3.