ABEDULLAH, Shujaat FAROOQ, Farah NAZ

Pakistan Institute of Development Economics (PIDE), Islamabad 44000, Pakistan

Abstract Sustainable income growth and poverty reduction remain critical challenges at the forefront of research in Pakistan,particularly in rural areas.To overcome these challenges, the role of rural transformation (RT) has emerged and gained importance in recent years.The present study is based on district-level data and covers the period from 1981 to 2019.The study attempts to quantify the role of rural transformation in boosting rural per capita income and alleviating rural poverty in the country.The study also aims to explore the impact of stages of rural transformation on rural per capita income and rural poverty alleviation.The empirical findings reveal that rural transformation (RT1 and RT2) is essential in enhancing rural per capita income and alleviating rural poverty.The role of the share of high-value crops (RT1) is more pronounced than the share of non-farm employment (RT2) in boosting rural per capita income and poverty alleviation.The trend of larger contribution of RT1 to enhance rural per capita income also continued at 2nd stage of rural transformation.In the case of poverty reduction, at 3rd stage of rural transformation, the role of RT2 is dominant.Our results indicate that districts at higher stages of rural transformation (both RT1 and RT2) tend to correlate positively with increased rural per capita income and reduced poverty rates, suggesting that progress in rural transformation is associated with improved economic conditions.However, it is important to note that this correlation does not necessarily imply a direct causal relationship between rural transformation and these economic outcomes; other factors may have influenced this relationship.In addition, the welfare impacts are more noticeable among the districts where a simultaneous shift from grain crops to cash crops and from farm employment to non-farm employment is observed.The study provides baseline information to learn experiences from fast-growing districts and to replicate the strategies in other districts, which boosts the RT process that may increase rural per capita income and enhance poverty reduction efforts.

Keywords: rural transformation, rural income, rural poverty, agriculture, panel analysis

1.lntroduction

Sustained economic growth and poverty alleviation are the central themes of public policies in most developing countries, where most of the population resides in rural areas (Huang 2018).In many countries, the rural transformation (RT) acts as a driving force for rural poverty alleviation as it operates through a rising share of high-yield crops (Timmer 2017; Fan 2019) and livelihood diversification by increasing the share of nonfarm activities (Reardonet al.2007; Haggbladeet al.2010; Otsuka and Fan 2021).Huang and Shi (2021)suggest that faster rural transformation contributes to rural poverty reduction in many developing countries.Menget al.(2020) provide empirical evidence that cash crop cultivation has a positive and significant impact on household farm income.However, its success largely depends on public policies (Doroshet al.2003),modern farming techniques (Spielmanet al.2016),gender inclusiveness, investment in rural infrastructure,sectoral backward and forward linkages, and livelihood opportunities available to the rural population (Farooq and Younais 2018).Also, rural development through E-commerce is considered an efficient channel to reduce rural poverty (Penget al.2021).

Rural transformation is crucial for Pakistan, where 63%(146 million) of the population resides in rural areas (World Bank 2021).The agriculture sector employs nearly half of the rural workforce despite 63% of the rural households being landless.On the other hand, non-farm activities contribute more than half of the total rural income,indicating they are an important source of income,especially for landless poor rural households (Farooq and Younais 2018).

Despite a significant reduction in rural poverty over time, there is a higher concentration of poverty in certain regions, i.e., interior Sindh, south Punjab, south Khyber Pakhtunkhwa, and Baluchistan.These areas mostly lack dynamic agricultural activities, a healthy non-farm sector,and resource diversification, i.e., urbanization, industry,and a healthy services sector to absorb the labor force newly entering the job market.The high poverty in these regions can be reconciled from three aspects.First,agriculture mostly faces a host of shocks, including crop failure, droughts, floods, and price fluctuations.Second,a large share of the rural labor force is employed in the traditional agriculture sector, but few efforts have been made to modernize it.Moreover, the ability of the nonfarm sector to absorb the additional labor force is also limited.Third, the quality of non-farm activities is largely compromised due to poor rural infrastructure and limited market linkages (Arif and Shujaat 2014).It is critical to review the rural and structural transformation processes in the country where the former is linked with resource diversification (i.e., a shift in agriculture from traditional to high-value crops) both at the household and regional levels.The latter dialogues on the rising share of agriculture in GDP through labor productivity (Huang and Shi 2021; Otsuka and Fan 2021).

Various studies in Pakistan found that rural poverty reduction is subject to the speed of rural transformation(Hussein and Suttie 2016; World Bank 2018).Rural transformation is mostly taking place through livestock,evidenced by the rising share of livestock in the agriculture sector from 57% in the 1970s to 77% in 2019(GoP 2020).Rural non-farm activities, on the other hand, have also been rising, and it is the main source of employment in rural areas.However, the speed and quality of jobs vary across regions, and the overall pace may be slow compared to China and various other Asian countries.Several reasons may hold.

First, the country devised various agricultural policies to enhance productivity and make it a modern sector,1These policies in Pakistan include National Agricultural Commission of 1987, the National Agricultural Policy of 1991,the Agricultural Perspective and Policy 2004, and the National Vision 2025 and 2030.but the implementation remains a major challenge.As a result, they did not fabricate the desired results.A high level of poverty exists in pure agricultural zones (Arif and Ahmad 2001).Second, the land is highly unequally distributed among rural masses, where only 37% of farmers own some land.Despite three land reforms(1959, 1972, 1977), a major share of farmland is cultivated under sharecropping arrangements - making it difficult to induce modern agricultural policy actions (Heston and Kumar 1983; Nasimet al.2014).Third, the credit markets are not playing an active role where most benefits are limited to big farmers by excluding the small ones (Malik and Nazli 1999; Qureshiet al.2004; Amjad and Hasnu 2007).Other critical challenges include expanding the capacity and knowledge to manage the natural resources(soil fertility and environment) and adopting efficient water use technologies by increasing public-private investments in this sector (Briscoeet al.2005; Laghariet al.2012).

Moreover, structural transformation (ST) relates to restructuring the economic activities towards the industry and services from basic agriculture-related activities.Such transformation contributes to increasing production,urbanization, domestic and international trade expansion,comprehensive division of labor, increased specialization,and declined share of agriculture in national GDP.For a better understanding of the RT process, the concept of ST is vital in the sense that RT is entrenched within the ST.Structural transformation over the last four decades (1978 to 2018) shows a declining share of agriculture in both the GDP and employment, indicating a transformation in farm and non-farm employment.The declining gap between agricultural share in GDP and employment highlights the rise in labor productivity in Pakistan.The analysis confirms the process of structural transformation, though the process is prolonged and requires speeding up the rural transformation process for sustainable growth.All this necessitates analyzing the country’s path and speed of rural transformation.The current study analyses the speed and direction of rural transformation and its impact on rural income and poverty over the last four decades.The analysis is carried out at the district level by using the panel dataset of three rounds.

The rest of the paper is organized as follows.Section 2 displays data and methods.Section 3 uncovers the results and discussions that include structural and rural transformation at the district level for the 1981-2019 period, speed of rural transformation, and correlation analysis based on econometric results.Section 4 concludes the study along with recommendations.

2.Data and methods

2.1.Theoretical framework

A conceptual framework maps the expected relationship between variables to draw coherent conclusions.Our conceptual framework attempts to map the process of rural transformation (Fig.1).The rural transformation(RT) is precisely measured by the share of high-value agriculture (RT1) and the share of non-farm employment(RT2).However, the process of rural transformation is governed by various drivers.The most significant and prominent drivers of RT include the policy stimulus,investment, and institutions (Morriset al.2017; Abreu and Mesias 2020; Casiniet al.2021).

Fig.1 Conceptual framework.Source: Author’s own summary.

The most prominent outcomes of this process are income growth and poverty reduction, which are the focus of this paper.The rural transformation acts through a rising percentage of high-value crops and the promotion of quality-based rural non-farm activities are the primary driving factors for rural poverty reduction (Timmer 2017;Fan 2019; Otsuka and Fan 2021).These outcomes may happen through various channels, as explained in Fig.1 and documented in the literature.Division of labor, specialization, and diversification leads to income growth (Jessop 2013; Callaghy 2019), and the rural transformation through these channels may also fruitfully lead to rural income growth.

According to Deininger and Squire (1998), the enabling factors (i.e., education, health, skills, and infrastructure)can promote equity through income redistribution.The likelihood of differential levels of rural transformation,such as advanced, normal, or lagging, due to disparities in the endogenous capabilities and characteristics of rurality and access to resources, can ultimately lead to regional development imbalances like income inequality(Liao and Chen 2017; Luet al.2020).However,numerous studies have shown that the rural investment sub-system increases agricultural gross domestic product (GDP) and decreases rural poverty through the implementation of supportive policies (Heet al.2016;Watts 2020).Other significant factors that effectively enhance rural transformation include expanding rural industry, strengthening and encouraging agriculture scaleintensive management, and growing rural employment and labor.However, it is also important not to overlook environmental protection and declining natural resources in rural transformation (Jiaoet al.2018).

2.2.Data collection procedure and methodology

The data used in this study were collected at three points in time (1981, 1998, and 2019) from different sources,including the Population Census and various sampledbased national surveys (e.g., the Pakistan Social and Living Measurement Survey (PSLM)).In addition, the provincial Agricultural Statistics were used to acquire the data on crop production and its value.

The main variables used in the study are rural income,Rural Deprivation Index (RDI, an indicator of poverty),literacy rate, employment rate, and urbanization.The literacy rate is defined for males and females aged 10 years and above by using the standard definition in the country where a literate is a person who can read and write with a good understanding.Similarly, the employment rate is defined separately for males and females by calculating the percentage share of employed to the respective gender working age population.Urbanization reflects the share of the urban population in the total population by following the official numbers.

RT is defined by two variables: RT1and RT2.The RT1may include cotton, oil, sugar, horticulture, livestock, and aquaculture in gross agricultural output values, excluding forestry (Huang and Shi 2021).Considering data availability, we made a few modifications to calculate RT1,where the high-value crops include cotton, sugar crops,horticulture, tobacco, oil seeds, and livestock.It is worth mentioning that sugar beet was added while calculating the value of sugarcane.Horticulture includes fruits and vegetables.The value of meat and milk was taken in livestock.The value of castor seed, groundnut, linseed,R&M seed, sesamum, soybean, and sunflower was taken to calculate the value of oil seeds.The fishery’s contribution is negligible and was excluded due to data unavailability.The low-value crops include grains and pulses.The grains include wheat, rice, maize, bajra,jawar, and barley; pulses include mong, mash, masoor,and gram.The total value of agriculture output was calculated by summing up the value of low- and highvalue crops against 62 items.The value of each crop was calculated by multiplying the production of a particular year with its respective prices.The data for calculating RT1were extracted from multiple sources.These sources include: Provincial Crop Reporting Services, Agriculture Marketing Information Services, Directorate of Agriculture Punjab, Ministry of Food Security and Research, Pakistan Census of Livestock (1986, 1996, and 2006) Special Reports, Agricultural Census Organization of Government of Pakistan and Provincial Agricultural Marketing Departments.

The RT2was calculated by subtracting the employment in agriculture, forestry, hunting, and fishing from the total employment in a particular district and dividing it by the total employment of that district.The data sources at the district level include Census Reports of 1981, 1998, and PSLM 2019.The two outcome variables are per capita rural income and rural poverty.The data on rural income at the household level was extracted from PSLM 2019 at the district level.PSLM provides data onward from 2004 at the district level.Therefore, to generate data for 1981 and 1998, we used the 2004 data series from PSLM as the base year.Using the GDP (Current US Dollar), the annual income growth rate is estimated between 1981 and 1998 and between 1998 and 2004.The per capita rural income variable for 1998 was generated by using backward compounding with the annual GDP growth rate(1998 and 2004).The 1981 series was generated using backward compounding with the annual GDP growth rate(1981 and 1998).Finally, the consumer price index (CPI)with the base year 2010 was used to find the per capita rural income in real terms for 1981, 1998, and 2019.

The official poverty estimates in Pakistan are only available at the national and regional (rural-urban)level, and it follows a consumption-based methodology.2The Planning Commission of Pakistan measured the official poverty line by using the Pakistan Integrated Household Survey(PIHS) 1998-1999 dataset, based on 2 350 calories per adult equivalent per day.Mahmoodet al.(2019) have calculated the subjective poverty at the household level by using sampled data.Unfortunately, such household-level subjective poverty measures are unavailable at the district level at the three points of time required for our panel analysis.Moreover,among other methods of poverty calculation, the notable methods are “the non-food consumption share approach”(Haq and Bhatti 2010) and the asset-based approach(Ullah and Chisti 2023).Again, these methods of poverty estimation are not applicable in our case because of data limitations.Hence, to overcome these data limitations, we have followed the Human Development Index (HDI) and Multidimensional Poverty Index (MPI)approach to develop RDI.We selected 10 indicators in four dimensions (education, health, living standard, and employment) at three points in time (1981, 1998, and 2019).The selection of indicators is largely derived from the availability of data as above mentioned points to construct panel:

(1) The education dimension has two indicators:the percentage of children not going to school and the percentage of adults lacking a primary education.

(2) The health dimension has one indicator: the percentage of the population not having access to the Rural Health Centre.

(3) The living standard dimension comprises five deprivation indicators: a) percentage of households that experience crowding (more than 3 persons live in a room); b) percentage of households that lack brick walls;c) percentage of households that lack safe drinking water;d) percentage of households that lack electricity; and e) percentage of households that lack efficient cooking sources.

(4) The employment dimension has two indicators: the percentage of unpaid family workers and the percentage of workers concentrated only in the agriculture sector.

Following the literature (UNDP 2014; Attariet al.2018),we estimated each indicator in percentage (ranging from 0 to 100).However, indicators are taken in reverse order, i.e., a higher percentage reflects high deprivation.Following the HDI approach (Desai 1991; Anand and Sen 1994), we used the geometric mean approach by giving equal weights to four dimensions.Initially, districts were ranked against each of the indicators by using maximum and minimum values (by following HDI methodology).

Poverty and inequality do not solely depend on rural transformation; certain other factors can impact its pace.Keeping this in view, we used some control variables in the analysis, including literacy rate (the percentage of the population of age group 10 and above who can read and write against the total population within the same age group), the employment rate (the percentage of the employed population (10 years and above) against the total population within the same age group), and urbanization (the percentage of the urban population against the total population).The data for these control variables were derived from various publications of the Pakistan Bureau of Statistics.

According to the latest Census 2017/18, there are 121 districts in four provinces.We excluded 42 districts from the analysis due to different reasons.Specifically,23 districts either have no or negligible agriculture; six districts do not experience changes in rural transformation mainly because they have constantly been growing fruits and vegetables since 1980 without changes; five districts are purely urban areas without agricultural land;data on important variables are missing in the remaining eight districts.Over time, new districts were announced by dividing the large districts into small ones due to administrative reasons.Hence, 79 districts in 2019 were merged back into 40 districts of 1980 (base year),implying that our analysis is based on 40 districts, which represent 79 districts of the 2017-2018 census.

The summary statistics of RT1reported in Table 1 indicate that the share of high-value agriculture is 74.2,which is higher than the expectation.However, it is mainly because of the high share of livestock contributing more than 50% in RT1.The average value of RT2is 42.8.Being an agrarian economy, it is showing a reasonable share of non-farm employment.The average real rural per capita income of Pakistan is PKR24 175.2(US$105.63), which is the average of the three periods considered in our analysis, and it is much less than the neighboring economy of China (US$2 379.26).Upon looking at the mean literacy rate for males and females,the former (45.3) is greater than the latter (20.5) due to the low average investment in female education in almost all districts of Pakistan.Similarly, the comparison of mean values of employment rate for males and females surprisingly depicts that the latter (6.6) is significantly less than the former (60.7).It is again due to less investment in female education that restricts the entry of women into job markets.

Table 1 Summary of variables used in the analysis

It is worth mentioning that the data on urbanization is available only at three points in time (1981, 1998,and 2019) during the last 40 years.Therefore, the analysis (both bi-variate and multi-variate) is covered for the subject time framework.Our mean statistic ofurbanization reveals that cities in Pakistan are expanding at the rate of 20%, which is alarming because it is quite challenging to accommodate such a large population every year in cities.It is putting pressure on food security because most productive agricultural lands are taken away to provide shelter to the migrants.Further, the availability of clean drinking water, jobs, transport, and shelters is quite challenging.

3.Results and discussion

3.1.Regional trends

Figs.2 and 3 demonstrate overtime changes of RT1and RT2, respectively, in selected districts at three points in time (1981, 1998, and 2019).There is a rising trend of RT1in almost all the districts of Punjab, particularly in Sargodha, Rawalpindi, and Faisalabad, where it went close to 90% in recent years.Similarly, districts Jacobabad, Shikarpur, and Larkana in Sindh have shown a tremendous improvement in RT1over time compared to other districts of Sindh.While growth in RT1has been noted in other districts of Sindh, the pace of progress has been relatively slower.

In Khyber-Pakhtunkhwa (KP), district Mardan is observed to be a leading district in RT1, and similarly,districts Bannu and Kohistan show some exceptional improvement in RT1.High vegetable and fruit production is observed in district Kohistan, whereas district Bannu improved its RT1, primarily due to a rising share of fruits and vegetables as well as milk production due to the establishment of a research institute in the area almost two decades ago - aiming to promote new varieties of fruits and vegetables.RT1in district Nasirabad almost doubled between 1998 and 2019 period.It is among the few districts of Balochistan having a crop production base where both the share of high-value crops and livestock went up.

Our district-level analysis reveals that the share of rural labor employment in the non-farm sector has been continuously rising over the last four decades, almost in all districts (Fig.3).The pace, however, varies over time.For example, in district Jhang, RT2was reasonably growing during the 1981 to 1998 period but then started to decline during the last two decades.Overall, there is a rising share of non-farm employment, suggesting that the country has witnessed a rural transformation.The comparison of RT1and RT2also suggests that the pace of RT2has been growing more swiftly than RT1in most districts, especially in Punjab Province.Larkana and Sanghar districts in Sindh have witnessed a high growth rate of RT2during the 1981-2019 period compared to the first two decades.In KP, the districts Kohat, Bannu,D.I.Khan, Dir, Mardan, Chitral, and Swat have had an exceptional growth rate of RT2during the last two decades.Similarly, in Balochistan, the district Nasirabad has had a reasonably high growth rate of RT2during the last two decades.

Fig.3 Share of rural non-farm employment across districts for 1981, 1998, and 2019.

Our analysis of per capita income over time reveals that all the districts have improved their per capita real income (Fig.4).Interestingly, at the initial point of comparison (i.e., in 1981), the value of per capita income was close to each other in all districts as they were almost same in terms of infrastructure and endowment and they lacked certain spill-over effects of urbanization, industry,and remittances.Several districts have succeeded in improving their income in an exceptional manner,especially during the last two decades, mainly due to a host of factors.For example, district Rawalpindi in Punjab holds multiple favorable externalities, including urbanization, industry, vibrant services sector, and remittances.Both the Sukkur and Larkana in Sindh are agricultural districts, and both these districts receive several infrastructure-related development projects from the provincial government.On average, the increase in income during the last four decades in all districts of Punjab, Sindh, KP, and Baluchistan is 4.38, 4.15, 4.92,and 5.21 times, respectively.Our results reveal that the income is growing almost at the same pace in most districts.

Fig.4 Per capita income of rural households across districts for 1981, 1998, and 2019.

Fig.5 indicates the reduction in rural poverty as measured through the RDI.The analysis shows that rural poverty has declined in every district over the last four decades.Unlike China (considered a benchmark for Asian countries), the pace of poverty reduction is slow in most of the districts of Pakistan.In 1981, Kohistan,Mardan, Nasirabad, and Bannu were the most deprived districts, with the RDI above 70%.However, a likely decline in poverty is observed in Rawalpindi, Bannu,Mardan, Sargodha, Chitral, and Dir, where the RDI fell to 25 or below in 2019.There are slight variations in poverty reduction in the rest of the districts of all four provinces.

Fig.5 Rural Deprivation Index across districts for 1981, 1998, and 2019.

3.2.Stages of rural transformation

We have defined three stages of rural transformation,both for RT1and RT2, by using certain thresholds.These thresholds are defined based on the difference in maximum and minimum values (dx=Max-Min) where stage 1 indicates a value less than min+dx/3, stage 2 is recognized for the values lie between min+dx/3 and min+2dx/3.Finally, the transformation will be regarded in stage 3 if it crosses the limit of min+2dx/3.The majority of the districts are clustering in stage 2 in terms of both RT1and RT2, whereas there are also some districts in stage 3 but only in terms of RT1(Fig.6).

Fig.6 Stages of rural transformation.

The typology analysis unfolds the relationship between rural income growth and the speed of rural transformation.The rural income growth is measured by using the average annual growth rate of rural per capita income for 1981, 1998, and 2019.Similarly, the speed of RT is measured as the average annual percentage point change in RT1and RT2for the same years.Though growth rate implications sacrifice the initial year value, at the same time, it provides beneficial information in terms of speed of transformation in connection to the rural income growth at the district level.We set specific thresholds for rural income growth by defining it in terms of “fast,” “moderate,”and “slow” speed.On the other hand, the RT1and RT2are also categorized by “fast” and “slow” based on their median values.For example, if the average percentage point change (for RT1and RT2) is greater than the median,speed is considered “fast,” otherwise “slow.” The results of the typological analysis are shown in Tables 2 and 3.

It is observed that Rawalpindi, Larkana, Sahiwal, and D.I.Khan have fast rural transformation in terms of RT1and high speed of income growth (Table 2).One reason for high-income growth in these districts is due to the low value of income in the initial years.The high speed of rural transformation in these districts is mainly observed due to the increasing share of the livestock sector and moving towards the cultivation of high-value crops such as cotton in Sahiwal, vegetables in Rawalpindi and D.I.Khan, and sugarcane in Larkana.

Table 2 Typology of district-level rural transformation (RT)based on the speed of RT1 and rural income growth for the years 1981, 1998, and 20191)

District Rawalpindi, Larkana, and D.I.Khan are again those districts that are fast in terms of RT2, which indicates that both farm and non-farm activities are more vital in these districts for the survival of the households.The households in these districts are well-off in terms of their income and consumption patterns.

Moreover, an attempt is also made to investigate the relation of slow RT1with fast, moderate, and slow rural income growth (Table 2).There is not a single district that falls in the slow category of rural transformation in terms of RT1and experiencing a high/fast per capita income.It reveals that RT1is a key factor that contributes towards the elevation of per capita income of rural households.We may observe that most districts facing slow/moderate rural income growth in terms of RT1are also falling in the same category of slow/moderate income growth in terms of RT2.Those districts that are slow in rural transformation (RT1and RT2) and income growth have neither proper infrastructure nor modernized agriculture facilities, which collectively kept these districts backward(GoP 2020).

Moving forward, the districts that are slow in RT2are witnessed with slow and moderate rural income growth(Table 3).It is observed that only the Sahiwal district is an exception where RT2is slow with fast per capita income growth.The sole reason is that district Sahiwal is fast in terms of RT1(Table 2), contributing to boosting per capita income.Particularly, the livestock sector is growing rapidly in Sahiwal because the Sahiwal breed is getting popular nationwide because of its high milk production.The rest of the districts indicate a slow increase in the share of non-farm employment (RT2).This demonstrates that growth in rural income is either moderate or slow.It is logical that slow rural transformation (RT2) or shifting from agriculture to non-agriculture sector results in moderate or slow rural income growth.It is observed that the majority of the districts that are slow in rural transformation (bothin RT1and RT2) are facing moderate/slow income growth(Tables 2 and 3), implying that high-income growth is strongly linked with the high speed of rural transformation.

Table 3 Typology of district-level rural transformation (RT)based on the speed of RT2 and rural income growth for the years 1981, 1998, and 2019

3.3.Econometric analysis

Multi-variate analyses were employed to quantify the impact of rural transformation on per capita rural income and poverty alleviation.Two types of analysis were carried out.First, the direct impact of rural transformation on real rural per capita income and poverty alleviation was analyzed as indicated in eq.(1), and second, the pace of rural transformation was quantified by including the multiplicative terms of stage dummies with RT1and RT2as given in eq.(2).Two dependent (outcome) variables(i.e., log of real rural per capita income and rural poverty)are considered in each equation.This implies that two different models (as indicated in eqs.(1) and (2)) for a log of real rural per capita income and rural poverty have been estimated.In all four models, we employed districtlevel data at three points in time (1981, 1998, and 2019).In eq.(1), the direct impact of RT1and RT2was explored on dependent variables (log of real rural income and rural poverty), while in eq.(2), we attempted to assess how different stages of rural transformation correlate with the dependent variable by adding multiplicative terms of stage dummies with RT1and RT2as explanatory variables.This allows us to further validate the hypothesis that higher stages of rural transformation are associated with a stronger positive correlation with increased per capita income and reduced poverty.The stage dummiesGDs2andGDs3of RT1and RT2are generated after classifying RT1and RT2into three stages.Stage one is considered as a base for comparison purposes.Where,λ1andλ2in eq.(1) are coefficients of RT1and RT2whileλ1,λ2,λ3, andλ4in eq.(2) are coefficients of multiplicative terms of stagedummies with RT1and RT2.Whereas,βnrepresents coefficients of control variables both in eqs.(1) and (2).The control variables in both equations are literacy rate,employment rate, and urbanization.As discussed above,the eqs.(1) and (2) can be written as below:

We used panel data at three points in time (1981,1998, and 2019).To avoid the issues of heterogeneity,we employed the panel estimation technique.The separate models were run for two dependent variables:rural per capita income and rural poverty.The fixed effect was applied for the former, and the random effect for the latter, as identified by the Hausman test.There might be a two-way relationship, where not only RT1and RT2affect the outcome variables (income and poverty), but RT1and RT2are determined by income.Since our objective is to examine how rural transformation may increase rural income and decrease rural poverty, we only conducted the regression that fulfills the study objective and excluded the other channel from our analysis.

A statistically significant linear relationship of RT with the rural per capita income and rural poverty is observed.In model 1, both RT1and RT2are statistically significant,indicating that a 1% increase in RT1and RT2leads to enhanced real per capita income by 1.4 and 1.0%,respectively.Similarly, a 1% increase in RT1and RT2causes a reduction in poverty by 2 and 2.6%, respectively(model 3).This implies that rural transformation not only increases the real rural per capita income but also alleviates rural poverty.

The impact of stages 2 and 3 of rural transformation has a positive and significant relationship with rural per capita income, indicating that both group dummies of RT1and RT2(in interactive form) are associated with an increase in rural per capita income in stage 2 and stage 3 compared to stage 1 (Table 4).Also, stage 3 shows a slightly stronger correlation than stage 2 with dependent variables in terms of both RT1and RT2(Table 4) (Timmer 2017; Fan 2019; Otsuka and Fan 2021).

Table 4 Impact of rural transformation on income and poverty - Panel analysis

By comparing the coefficient of stages 2nd and 3rd of RT1with the corresponding stages of RT2in model 2, our empirical results reveal that the magnitude of stages 2nd and 3rd of RT1is larger than the corresponding stages of RT2.These results clearly demonstrate that each stage of RT1is more strongly associated with an increase in rural per capita income than the corresponding stage of RT2.Hence, it leads to the conclusion that investments aimed at boosting RT1may have a stronger correlation with increasing rural per capita income compared to investments in RT2.

Our empirical analysis further reveals that a 1%increase in female literacy rate and female employment rate contribute to an increase in rural per capita income by 0.012 and 0.009%, respectively.The findings are consistent with the earlier studies (Chaudhry and Rahman 2009).Contrary to expectations, an increase in the male employment rate leads to a decline in per capita rural income.Probably, a negative impact of male employment on rural per capita income is due to self-employment at their own farm.The increase in the male employment rate does contribute to an increase in income because working at their own farm does not cause monetary transactions.Hence, the definition of employment needs to be revisited.Urbanization is also found to have apositive and significant impact on rural per capita income.The results are consistent with the general perception that urbanization increases employment opportunities and,thus, per capita income (Zhang 2016).This suggests that urbanization can play a key role in increasing per capita rural income through a series of structural transformations,such as attracting unemployed labor from rural areas and promoting rural industry and services sector.

The rural transformation (both RT1and RT2) is found to have a negative and significant impact on rural poverty.However, comparing the size of coefficients of RT1and RT2in model 3 reveals that RT1dominates the impact on rural poverty.This implies that increasing the share of high-value crops can play a vital role in alleviating rural poverty.When we examine the correlations between stages of rural transformation and rural poverty in model 4, it becomes evident that stages 2 and 3 of both RT1and RT2has negative and significant coefficients, implying that stages 2 and 3 exhibit stronger correlations with rural poverty reduction as compared to stage 1.Furthermore,upon comparing the coefficients of stage 3 with those of stage 2 in model 4, it becomes evident that the magnitude of the coefficient for stage 3 is greater than that of stage 2, providing empirical evidence that higher stages of rural transformation exhibit a stronger correlation with poverty alleviation (model 4, Table 4).On the other hand, comparing the magnitude of the coefficients of RT1and RT2at stage 2 in model 4 reveals that RT1has a stronger correlation with reducing rural poverty.However, the comparison of coefficients at the 3rd stage indicates that RT2exhibits a stronger correlation with reducing rural poverty.This implies that in districts that are at the 2nd stage of rural transformation, there is a need to promote high-value agricultural crops to alleviate poverty.However, in those districts at the 3rd stage of rural transformation, there is a need to promote non-farm employment by promoting processing and other small industries to alleviate poverty.

In general, it can be concluded that achieving higher stages of rural transformation can sharply lead to alleviating rural poverty.Among other controlling factors, the female literacy rate significantly alleviates rural poverty (models 3 and 4).However, contrary to the expectation, the male employment rate increases rural poverty.It is probably because of male self-employment at their own farm that does not allow them to earn money by working on others’ farms.Self-employment does not allow the transfer of money from someone to the selfemployed person, but it restricts them from working elsewhere.Hence, self-employment leads to an increase in rural poverty.Hence, the definition of male employment needs to revisit where self-employment is included in the employment schedule.

Overall, the analysis reveals that rural transformation acts as a driving force for increasing rural per capita income in the country where the contribution of RT1in promoting rural per capita income is comparatively higher than RT2, suggesting that the country must focus on increasing the share of high-value crops to achieve the objective of high rural per capita income.Keeping in view the high land inequality and stagnant productivity of highvalue crops, employment in the non-farm sector has a vast potential to reap certain benefits from globalization and liberalization policies, remittances, and urbanization.The district-level poverty evidence suggests that pure agricultural zones (i.e., south Punjab and interior Sindh)have the highest poverty rates as these regions mainly focus on food grain crops by ignoring the importance of high-value crops in the cropping system.Moreover, these regions lack vibrant urbanization, industry, and market connectivity.On the other hand, regions like north Punjab and central Khyber Pakhtunkhwa have the least poverty as the region provides resource diversification where both farm and non-farm activities prevail.Even within the farming system, there is a high level of diversification towards high-value crops and livestock production.All this suggests that efficient utilization of resources in the agriculture sector and an increase in employment in the non-agriculture sector by establishing small processing units could provide a strong basis to speed up the rural transformation process.

4.Conclusion

The study employs district-level data from the last four decades at three points: 1981, 1998, and 2019; hence,we employed panel data techniques.The study attempts to quantify the relationship between rural transformation,stages of rural transformation with per capita income,and rural poverty.Unlike China, the rural transformation is found to have a linear relationship with rural per capita income and rural poverty reduction.

A descriptive analysis reveals that all districts have witnessed a rural transformation, but speed and pace vary.Most districts fall in stage 2 of rural transformation in terms of RT1and RT2, but some fall in stage 3 but only in terms of RT1.However, no district is observed to fall in stage 3 in terms of RT2.

Our analysis based on the last four decades provides substantial empirical evidence that rural transformation(RT) has a positive impact on rural per capita income but a negative impact on rural poverty, implying that rural transformation can be used as a tool to enhance per capita income and to alleviate rural poverty.

Comparing the contribution of RT1and RT2in enhancing rural per capita income reveals that the contribution of RT1is more than double that of RT2.This suggests that policies, institutions, and investments need to be redirected to increase the share of high-value crops to achieve the objective of high rural per capita income.Similarly, the larger magnitude of the coefficient of RT1compared to RT2in model 3 demonstrates that efforts to increase the share of high-value crops could play a vital role in alleviating rural poverty.

Better education and women’s economic inclusion would effectively address the challenges of rural poverty.Similarly, urbanization in the region helps in generating more economic opportunities in rural areas.

The study draws some policy implications from our empirical findings.First, more district-level data would help portray a clear picture of rural transformation, i.e.,information on infrastructure, market linkages, and valuechain of agricultural products.Second, the estimates of RT1appear high, primarily due to the livestock component,whereas the share of high-value crops within the RT1is comparatively smaller than livestock.The country requires a massive investment in R&D to enhance the productivity of high-value crops (fruits and vegetables) and an augmentation in the product’s value chain to promote exports.On the other hand, market reforms are essential to promote efficiency and equity by avoiding market fragmentation and a host of crop shocks, including price failure, lack of crop insurance mechanisms, and domination of few buyers in the product market.A market’s easy,smooth, and equitable functioning can be facilitated through supporting institutional mechanisms that promote economic activity by reducing transaction costs and other hurdles.Increasing competition requires institutions for quality control, capacity building, research, and development,reducing disputes, defining property rights and contracts,and increasing healthy market competition.Third, the non-farm sector has a great potential to absorb surplus labor and produce backward and forward linkages with agriculture and rural development.In this regard, public investment, along with technical training, is required to improve the productivity and size of this sector, especially to improve the poor manufacturing base.Targeted policies are required to overcome the regional disparities by diverting resources toward the deprived and remote areas.It will ultimately improve livelihood opportunities, especially for poor households that lack an agricultural base.For rural development, dynamic labor-intensive agriculture along with a modern non-agriculture sector can provide better employment and income to rural households, with egalitarian income distribution and elimination of rural poverty.Policy intervention to promote rural non-farm employment can help stop migration to cities and boost the rural transformation process.

Acknowledgements

We would like to thank all the Government of Pakistan(GoP) officials, particularly the Pakistan Bureau of Statistics, the Planning Commission, the Pakistan Agriculture and Research Centre, and the provincial government agriculture departments for assisting us in compiling broad-based data used in this research.We highly acknowledge the financial support of the Australian Centre for International Agricultural Research (ACIAR),Australia (ADP/2017/024).

Declaration of competing interest

The authors declare that they have no conflict of interest.