Climate Volatility, Wheat Productivity and Food Security:

A Quantile Regression Analysis

BABAR HUSSAIN1, USMAN ALI2, SANIA SHAHEEN3, LAL K.ALMAS4

ISLAMABAD

Shorkot City, Punjab,

Abstract: - Climate change's effects on food crop production are a serious concern due to its linkages with food

insecurity. This study attempts to investigate the question of whether and to what extend climate volatility has

affected the yield of a major staple crop, the wheat, in the District Faisalabad, the largest agricultural city in

Pakistan. Daily base data of temperature and rainfall over the past 33 years is collected, and the average and

volatility measures of climate conditions are calculated for the whole crop period as well as for the vegetative

cope with uncertain climate shocks.

Key-Words: - Climate volatility, temperature, rainfall, food security, quantile

Received: April 9, 2023. Revised: September 7, 2023. Accepted: November 2, 2023. Published: December 6, 2023.

1 Introduction

Environmental degradation and climate change

issues are considered the greatest threat to human

resources, and agriculture production, [4], [5].

Climate change impacts are global in scope and

unprecedented in scale, [6], [7].

The repercussions of such climatic changes are

visible from the intolerable weather conditions,

more frequent floods, episodes of extreme

temperature and rainfall, food shortage, etc, [8]. It

is anticipated that the concentration of GHGs will

to the level of the pre-industrial era, and this will

of such variations. For instance, temperature rise,

changes in rainfall and precipitation patterns, and

other extreme weather events may disrupt

agriculture productivity, food availability, and

quality,

[10], [11].

comprehending the impacts of climatic variability

on food crop production is indispensable to adopting

preventive measures. As observed in the 5th report

of IPCC (the Intergovernmental Panel on Climate

Change, [7], climate change in the last three decades

has negatively affected crops, reducing production

by up to 5%, and climate volatility is the main

responsible factor driving food price instability in

recent years. However, despite the importance of the

topic, very limited research is devoted to exploring

they ignore the unprotected changes or the volatile

nature of environmental variables, such as sudden

rainfall during the harvesting period or a lack of

temperature. The recent climatic changes are more

volatile and require a fresh investigation of the

relationship between climate volatility and food

crop productivity.

Although these studies provided important

insights and guidelines for future research, the main

shortcoming is that they ignore the unprotected

productivity.

This study is intended to inquire about a case

from Pakistan to spotlight the vulnerability of a

major staple crop in particular, and the agriculture

sector in general. The contribution of the agriculture

sector to GDP is about 24% of GDP. This sector

employs half of the total labor force in the country

and is the third-largest source of foreign exchange

earnings. The sensitivity of Pakistan to climate

change is obvious from, for example, extreme

conditions are appearing in Pakistan for the last two

decades, [18], [19], [20].

The shortage of food stuffs along with

that food insecurity is becoming more severe in

rural areas that are already facing higher food prices

and shortages in the country, [23].

This study contributes to the broader debate on

the climate change-food security nexus by exploring

whether and to what extent the capricious pattern of

environmental indicators exerts an effect on wheat

productivity, a major food crop not only in Pakistan

but also in other regions of the world. This paper

has several distinctive features. First, we calculate

indicators daily to calculate average and volatility

measures of rainfall and temperature from the

largest agriculture-producing district of Pakistan,

Faisalabad. Methodologically, we utilized the

quantile regression (QR) approach. This method has

an advantage over the other existing techniques as it

helps to explain the relationship among the variables

at the different points of data distribution, rather

than focusing only on the single average parameter

estimation.

whole discussion.

patterns, caused by nature or anthropogenic

activities, which persist for a longer time period,

[24]. Rise in Population size, deforestation, and

climate change, [9], [25].

Though there might still be a debate on the

degree and sources of climate change, its actuality

has mainly been accepted on a scientific basis, [24],

[26].

Climate change instigated a shift in seasons, a rise in

temperature and sea level, and thus resulting in

punishing weather events like floods, storms, and

Agriculture is one of the basic economic

activities that not only provides food to human

beings but also a source of industrial raw materials.

Unfortunately, agriculture is one of the vulnerable

sectors to climate change owing to its greater

dependence on weather and climatic conditions,

[30]. The vulnerability of the agriculture sector is

associated with several interrelated factors, such as

variations in rainfall patterns, sunshine hours,

temperature, humidity, droughts, and storms etc.

is exposed to population pressure and several

climatic factors including changes in rainfall and

temperature patterns, harvesting time, and water

stress. All these factors have the potential to alter

agriculture productivity and yield, and thus have

substantial implications for food security, [33], [34],

[35].

Recent quantitative research in this field utilizes

different methods and climatic indicators to

empirically examine the food crop response to

to any change in climatic conditions. [37], study the

effects of rainfall and temperature on agriculture

productivity using Ethiopian household survey data.

The findings show that the temperature’s effects are

significant and nonlinear; while the impact of

precipitation on productivity is less prominent as

compared to temperature. Besides, the effects of

temperature are not the same across food crops,

indicating the crop-specific effects of climatic

indicators. [38], conclude that the average

substantial loss of China’s grain production. [17],

predicted about 32% fall in wheat productivity in

Mexico due to changes in rainfall patterns. More

food-insecure by affecting food crops’ productivity

directly or indirectly through changes in

temperature, rainfall, precipitation, and other related

conditions, [41], [23], [42], [43], use metrological

data of rainfall, temperature, sunshine hours, and

relative humidity to show that all these factors have

a favorable impact on productivity during the

reproductive stage.

[9], use 50 years’ time series data and find no

negative effect of climate change on wheat yield.

different across different crops and districts in the

province.

[45], employ average monthly data to show that

the temperature has significantly positive impacts on

both the vegetative and reproductive stages of the

rice crop; while the rainfall is harmful only for the

reproductive phase. [46], utilize annual time series

data of 19 districts of Pakistan and report a negative

effect of temperature hikes on wheat yield. [47],

collected primary data from 442 farmers and

wheat farmers to show a negative impact of

temperature rise on mean wheat yield. [50],

conducted a survey of 150 farmers from a province

of Pakistan and found that the wheat yield response

to changing climatic conditions is different across

the districts in the province. More recently, [45],

shows that temperature anomaly has a significantly

negative effect on the economic efficiency of

rainfed wheat farmers; while rainfall appears to

exert a significantly positive effect.

ambiguous as some report negative, [38], [48],

while others conclude a positive impact of climate

change on food-crop yield, [9], [32], [47].

annually-averaged measures, [53], of the climatic

variables. Different from previous studies, this study

collected daily base data of climatic conditions and

calculated their average and volatility measures to

examine if it is the climate volatility that matters the

most to food-crop yield. In addition, a non-

parametric QR approach is utilized to quantify the

impact of climate volatility on food security by

focusing on wheat yield.

In summary, the extant literature provides very

Besides, these studies are based on the cross-

of the farmers, [47], [48], [51], [54]. Although some

studies have utilized time series data, their analysis

is susceptible to aggregation bias as they use

monthly-averaged, [52], or even annually-averaged

measures, [14], of the climatic variables. Different

from previous studies, this study collected daily

base data of climatic conditions and calculated their

average and volatility measures to examine if it is

3 Data and Methodology

parameters along with some other non-climatic

factors to quantify the impact of climate volatility

on the wheat crop - the major staple food in

Pakistan. The initial daily base data during 1981-

2013 for district Faisalabad is collected from the

Pakistan Meteorological Department, Faisalabad

station. This daily-based data is utilized to calculate

the average and volatility measures of rainfall and

temperature. Following, [27], we assess rainfall

volatility (RV) and temperature volatility (TV) from

the coefficient variation i.e., the standard deviation

of the rainfall series divided by the mean of the

Temperature (y) Volatility,

It should be noted that we sort out the data

duration by including only those months that

encompass the duration of the wheat crop in the

Faisalabad region. In other words, the total duration

eliminated. In addition to calculating average and

analyzing the dynamics of crop growth. AUC also

accounts for various omitted economic variables.

For instance, if the farmer expects a higher price for

includes time trends as an additional regressor to

avoid spurious regression, [55], [56]. A description

of the variables is given in Table 1.

Table 2. Descriptive statistics

is a vector of climatic factors. AUCt and At represent

annual trends and AUC, respectively. The model

specifications to estimate the impact of average

climate change and volatility are as follows:

(3)

where Yieldt is the wheat yield per hectare, Trend

captures the annual time trend, and AUC is the area

under wheat cultivation. Rain and Tem represent the

average of rainfall and temperature, respectively,

during the whole crop duration; while VR and VT

respectively are the volatility measures of rainfall

temperature. As aforementioned, we have also made

a separation in the wheat crop growth period i.e.,

Veg_TM +

(5)

(6)

This study employs the QR approach to analyze the

impact of climatic variations on wheat yield

distribution. This methodology helps analyze the

differential impact of climatic indicators across the

different points of wheat yield distribution and also

facilitates computing the marginal effects of other

regressors on the conditional distribution of the

respondent variable. In reality, several phenomena

require analysis of the differential distributional

quantiles of a distribution to study the impact on the

tails of the distribution. This is in contrast to the

traditional ‘average-based’ regressions which yield

distribution.

In this particular context, the QR is more

appropriate to study the effects of fluctuations in

climatic variables on the different conditional

distributions of wheat yield because different stages

of the crop growth might require different levels of

temperature and rainfall, [15]. Besides, the QR is

found to address the issue of heteroscedasticity by

estimating different coefficients for the quantiles,

[2]. The distortion arising from the outliers in data is

provides a very malleable tool to inspect the

association between the variables, without placing

any constraints on the functional form. Following,

[57], the QR model is written as,

where x is the vector of independent variables, β

error term. θ󰇛󰇜 is the θth conditional

quantile for y given x. Distinct from the traditional

least square methods, the QR minimizes the

absolute sum of the error for a particular quantile of

defined as;

yield different values against any given value of θ.

The linear programming technique is utilized to

solve the problem by using the whole sample. The

QR production function in the present context can

where the θth conditional quantile of wheat yield is

given by θ󰇛󰇜, X denotes the set of

explanatory variables. In this study, we estimate a

vector of coefficients, , for the four quantiles, i.e.,

model.

Augmented Dicky Fuller (ADF) and Philips Perron

(PP) tests to confirm the stationary of the observed

series. From the results given in Table 3, it is

revealed that yield, rainfall, vegetative stage rainfall,

and reproductive stage temperature are stationary at

the level at 1% level of significance; while average

temperature, rainfall for the whole crop period, and

the vegetative stage temperature are stationary at the

5% level of significance. The results of the

remaining variable also show that all variables are

Before proceeding to the main QR analysis, it is

useful to draw some non-parametric evidence by

using graphs. Figure 2 displays a linear relationship

Fig. 2: Linear regression lines between climate

change and wheat yield

Rainfall is a crucial input and an important

driving force behind the wheat crop growth in the

rain-fed as well as in the irrigated settings. In the

when the crop is ready to be harvested, is harmful to

crop productivity because, at that time the sunshine

hours and maximum temperature are needed for the

crop ripeness; while rainfall at this ripening stage

deteriorates grain quality, such as damaged grains or

viviparous germination. Therefore, we separated the

data into two parts, the vegetative stage and

reproductive stage, to get a clearer picture of these

effects.

their association with wheat yield. It reveals that the

volatility of both rainfall and temperature is

favorably related to yield, suggesting that the

volatility of climatic conditions is desirable rather

than their permanent existence or absence.

Turning to the regression estimations, Table 4

represents the results of coefficient estimation for

the 25th, 50th, 75th, and 95th percentiles of the

yield distribution. The estimates reveal that wheat

yield responds differently to climate change across

decade as there has been a substantial increase in

wheat yield per hectare for the last ten to fifteen

years (see also Figure 1). The results reveal a

negative relationship between wheat yield and

average maximum temperature at the lowest

quantile, although insignificant; while the

coefficient of temperature becomes positive at the

95th quantile. Specifically, the impact of average

temperature is significantly greater and positive for

the highest quantile, indicating a rise of about 2.68

period of the study and the highest quantile captures

the more recent wheat yield trends, the findings

suggest that average temperature is not a threat to

disproportionally associated with agriculture

coefficient is increasing with higher quantile; while

positive impact of the Trend is not surprising, given

farmers are risk-averse, take more care of their

maximize output, and thus avoid a diminishing

return to scale. On the other hand, large farmers use

wage laborers and also face greater uncertainty and

concept than the average climatic change, as the

former represents uncertain variations in weather;

while the latter is related to the gradual changes in

weather captured through averaging the parameters.

This is also depicted from the results in Table 5

which are different from those of Table 4. The

difference is seen not only from the magnitude but

also from the signs of coefficients. The results show

significantly negative coefficients of both rainfall

and temperature at the highest quantile, while

for the crop. Besides, as we move from the lower to

the upper quantile, the size, and sign of the

maximum temperature coefficient are also

changing; while the coefficient of rainfall is

negatively associated with every quantile and its

magnitude is the highest for the highest quantile, as

also concluded by, [36].

In other words, the volatility in climatic factors

i.e., maximum temperature and rainfall is highly and

negatively associated with wheat yield in the last

The results discussed so far (Table 4 and Table 5)

were based on the overall wheat crop growth period,

and there is a caveat that the estimations might be

sensitive due to ignoring the different stages of

wheat crop growth. That is, the temperature and

rainfall requirements of the wheat crop are different

Starting from the average measures, the results

reported in Table 6 show that the rainfall is

significant and positive only at the highest quantile

for both the vegetative and reproductive stages,

confirming our previous findings that average

rainfall is positively associated with wheat yield.

However, the impact of temperature across the crop

growth stages is not the same. It is significantly

positive for the vegetative stage but negative for the

reproductive stage. Specifically, the vegetative stage

are not the same for the vegetative and reproductive

stages. Similarly, we found a greater effect of rain at

the vegetative stage than that of the reproductive

stage. A millimeter rise in rainfall at the vegetative

stage increases yield by about 5.8; while the same

change at the reproductive stage brings only a 0.6

mounds increase. This indicates that the rainfall is

more beneficial at the early stages of the crop

growth period i.e., the flowering stage. Nonetheless,

these estimations are based on the average measures

Table 7. In the case of the lowest quantile, we find a