Information and communication technology (ICT) has seen significant advancements in recent decades across various fields. Due to its potential to transform the economy and society, extensive research has been conducted to examine its impact on various aspects (Chandio et al., 2023). Early studies concentrated on production, economic growth, and poverty reduction, and ICT was considered part of the production function alongside land, capital, and labor (Chandio et al., 2023). Numerous studies indicate that ICT positively affects employment (Atasoy, 2013), family income, and labor mobility (Hartje and Hübler, 2017). Some scholars suggest that ICT may help reshape rural economies and narrow the global development gap (Ma et al., 2020). Research on ICT has expanded to encompass various topics, including gender gap reduction (Ojo et al., 2013), entrepreneurship promotion (Afutu-Kotey et al., 2017), and financial empowerment. ICT benefits these factors by enhancing the efficiency of information generation, transmission, and access, reducing search and transaction costs, and enabling more efficient production and management systems.

Several factors influencing agricultural productivity have been identified (Issahaku et al., 2018). Over the last two decades, a significant body of literature has emphasized the role of ICT. Lio and Liu (2006) initially demonstrated the role of ICT in enhancing agricultural output in 81 countries between 1995 and 2000. Subsequent research by various scholars supported these findings. Ogutu et al. (2014) showed that widespread ICT use improves production in small-scale agriculture by addressing information asymmetry. Internet connectivity significantly boosted food production in Vietnam (Kaila and Tarp, 2019), reducing poverty in rural areas (Twumasi et al., 2021). Another study in Pakistan found that mobile phone and Internet usage increased wheat growers’ income, indicating improved marketing and sales efficiency that enhances crop profitability (Khan et al., 2022). Deng et al. (2019) also reported that Internet use enhances resource efficiency and lessens agricultural waste.

Extensive literature evidence supports the Internet’s role in agricultural production, prompting investigations into its causes. These studies highlight increased human capital and information access as key factors. Access to technical agricultural information aids growers in diversifying crops, allocating land and inputs more efficiently (Leng et al., 2020), and expanding their land holdings (Hou et al., 2019), leading to higher productivity. Additionally, addressing challenges like knowledge asymmetry and adverse selection helps farmers make better decisions and exhibit more effective management. Enhancing farmer communication and providing learning opportunities can significantly bolster social capital and information literacy and ultimately influence farmer behavior toward the adoption of more productive agricultural practices. For instance, when farmers have access to Internet-based resources, they tend to exercise greater discernment in using chemical inputs such as fertilizers and pesticides. Additionally, IT platforms have the potential to broaden the social capital of households, fostering an environment conducive to the dissemination and application of production technology (Fu and Zhu, 2023). Similarly, Deng et al. (2022) indicated that Internet use influenced rural growers’ perceptions of ecological contamination in China, suggesting that Internet access can be a valuable tool for promoting environmentally friendly agricultural expansion and mitigating environmental issues.

Furthermore, studies have examined the agriculture industry from various perspectives, analyzing the impact of technological advancements on the incomes of the agricultural sector and rural families. For instance, research suggests that using ICT could effectively reduce income inequality in rural areas (Deng et al., 2022). Moreover, Min et al. (2020), utilizing empirical data from 2008 to 2015, concluded that ICT plays a significant role in driving economic expansion and growth. The assumption that IT usage positively influences rural growers’ well-being is supported by Ma et al. (2020). Nie et al. (2021) provide support for their conclusions. Existing literature commonly acknowledges the positive impact of technology on the agriculture industry. Technology holds promise for improving the economic feasibility of biochar in conventional agriculture while also fostering contributions to the circular economy (Maroušek et al., 2023). Based on the results, this research investigates the impacts of IT on wheat crop output. While some research has focused on different Pakistani crops, most have examined the long-term consequences of climate change. Research centered on crop production differs from ours in several fundamental ways. The most recent study by Lin et al. (2022) is comparable to ours regarding topic choice. Investigators look at the key elements that will boost agricultural output, but, in contrast to our findings, they place a greater emphasis on cooperative participation.

The findings of the study suggest that cooperative contributions positively impact the overall factor efficiency of small- and medium-sized businesses. Specifically, in Pakistan, a limited number of studies have explored the influence of IT usage on crop yield, particularly among those investigating the impact of ICT on the agricultural industry. The information collected in this regard will aid in understanding how the agricultural sector, facing pressure from both demand and supply due to population growth and climate change, may address this issue.

In the following section, we elaborate on key concepts within the productivity framework to clarify the potential pathways through which IT usage can impact the TE of farms. We begin with a simple representation of the production frontier, which signifies the maximum output attainable at each input level. Productivity is quantified by the ratio of aggregated output over aggregated input (Coelli, 1995). Consequently, farms situated on the frontier are deemed technically efficient, while those below are not, as a greater output can be achieved with the same input level, or inputs can be conserved without compromising the output level. Therefore, achieving elevated TE requires either increasing the output with the current inputs or reducing the inputs without compromising the prevailing output.

Internet technology could act as a factor influencing TE for several reasons. First, IT can assist farmers in making informed decisions and guide them toward adopting suitable farming methods. Crop and vegetable farmers in less developed countries face challenges such as a lack of education and experience, restricted access to inputs, and inferior agricultural extension services. By facilitating direct, fast, and global information and idea exchange between farmers and experts and addressing these issues (Hobbs, 1996; Bozoğlu and Ceyhan, 2007; Aker, 2008; Schmidt and Wagner, 2019; Quintana-García et al., 2021; Kang et al., 2023), ICT enhances the transmission of information. Producers may also gain greater access to advice and instruction from reliable professionals.

Second, IT promotes the availability of agricultural inputs with greater quality or cheaper cost and gives information on products and services (Zhu et al., 2021). Farmers may currently acquire the up-to-date market report and are no longer confined to the few options they had previously for keeping up with factor markets. Third, IT can assist rural families in distributing labor and capital more effectively by connecting them with suppliers and consumers and enabling communication (Zanello and Srinivasan, 2014; Hou et al., 2019). Farmers may identify market trends and quickly modify production methods to account for potential risks and losses when they have immediate access to market and pricing information related to agriculture.

Balochistan, the largest province in southwestern Pakistan, covers a vast area of 347,190 square kilometers. Despite being the least populous province, it constitutes 44% of the country’s total land area. The agricultural sector in Balochistan holds significant economic potential (Shami et al., 2016; Abdullah and Ahmed, 2018). Many areas in the province are conducive to agricultural production, but the true potential has not been fully realized due to various challenges. Over 81% of farmers across the province express concerns about issues such as power and water shortages negatively impacting agriculture (Ashraf and Routray, 2013). Provinces serve as the highest administrative units in Pakistan, each with its provincial government. Districts operate as second-level administrative units within a province, while tehsils are sub-district administrative units within a district. Union councils (UCs) represent the smallest rural administrative units within a tehsil.

To investigate the impact of IT on TE, we employ a multi-step approach designed to progressively mitigate potential bias arising from both observable and unobservable factors. Initially, we present SF model results on the original (unmatched) sample, recognizing potential selection bias. Subsequently, PSM is utilized to construct a balanced sample of IT users and non-users, addressing bias related to observed characteristics. Then, Greene’s (2010) sample selection model is applied to the matched sample to rectify potential bias stemming from unobserved factors. We then compare TE scores of IT users and non-users resulting from different combinations of these correction procedures, with the most reliable outcomes derived from the sample selection SF model on the matched sample. Finally, QTE can be calculated using observed data, effectively correcting for selection bias by comparing quantiles of the outcome distribution among individuals with varying treatment values. This analytical approach facilitates an understanding of whether the influence of IT varies depending on the efficiency level within the agricultural sector.

Table 1 presents the means and standard deviations for both the combined treatment (IT users) and control (IT non-users) groups. The IT group includes 310 observations, while the NIT group comprises 290. On average, household heads in the IT group are approximately 48 years old, compared to 53 years old in the NIT group. This age difference aligns with research indicating that older individuals are less likely to innovate and utilize the Internet for entertainment (Nguyen et al., 2023). Additionally, Internet users tend to have higher levels of education, possibly due to lower-educated individuals lacking IT skills or facing difficulties with comprehensive texts (Močnik and Širec, 2010). Penard et al. (2015) also demonstrate that younger and better educated individuals are more inclined to the Internet. Regarding gender, approximately 84% of households in the IT group are male-headed, which is approximately 5% higher than in the NIT group. Moreover, farmers affiliated with cooperatives are more likely to use the Internet, possibly because growers find it easier to understand and utilize modern technologies. Furthermore, farmers with greater farming experience are more inclined to adopt IT. These findings collectively emphasize the need for targeted interventions and support mechanisms to promote the widespread adoption of digital tools in agriculture, ultimately fostering enhanced TE and productivity in food production.

The results of the selection-corrected SF and conventional SF models are presented in Table 2 for both the IT users and non-users groups, utilizing the entire samples. One exhibits statistical significance at the 1 % level, indicating the need to consider the SF method with corrective selection. The conventional SF model is also refuted by the likelihood ratio (LR) tests conducted under both regimes. The first-order constants can be interpreted as outcome elasticities computed at the sample mean by categorizing all variables based on their geometric values before calculation. This interpretation holds because all variables are segmented using their geometric means, as estimated previously (Orea, 2002). IT users display an output elasticity of 0.30, signifying that a 1% increase in fertilizer usage will result in a 0.3% boost in output. For growers utilizing IT, land has the most significant impact on agricultural output, with an output elasticity of 0.48, as per findings from earlier research on vegetable productivity in Sri Lanka (Padmajani et al., 2014). Reduced yields in vegetable cultivation may be attributed to growers using excessive amounts of chemicals to mitigate the risk of crop loss due to illness and pests. In the case of IT non-users, land size has the largest elasticity (0.30), while fertilizer and other inputs contribute approximately 0.26 and 0.23, respectively, to production. Compared to labor, pesticides exhibit a lower production elasticity of 0.1%. Our estimates align with previous investigations (Dong et al., 2019).

The cumulative fractional productivity elasticities for the IT user and non-user groups sum to approximately 1, indicating a consistently sized regression that remains robust to the subsequent results (Shrestha et al., 2016). Standard TE scores for the unmatched are presented in Table 3 for both the conventional and selection-corrected SF approaches. In the conventional SF model, IT users exhibit an average TE score of 0.62, while the non-users group has a score of 0.57. The selection-corrected SF method is anticipated to yield slightly higher TE scores. When measuring unobservable bias, the TE value for the non-users group increases by 0.03, whereas it only increases by 0.01 for IT users. The assessment of unobserved bias reveals that the two-group frontier TE values in the selection-corrected SF demonstrate a positive effect of utilizing IT on crop producers’ TE.

This study used PSM to decrease the observed bias between IT users and non-users. After calculating P-scores and matching, Table 4 shows that the standardized biases of variables greatly decrease after matching, with all absolute values lowered to less than 10%, demonstrating the success of the matching procedure. To verify the trustworthiness of matching, we must analyze the covariate balance between IT and non-IT groups. Table 4 shows how the unmatched sample fails to attain a covariate balance. However, following matching, standardized bias is significantly decreased, with absolute levels. Furthermore, the t-test findings fail to reject the null hypothesis that there are no systematic distributional differences between the two groups, confirming the efficacy of the matching procedure.

The Probit model marginal effects and constants in Table 5 elucidate how various factors impact growers’ decisions to use IT for information gathering. Gender exerts a strong and favorable influence on IT use, suggesting that male growers are more inclined to use IT for information gathering than their female counterparts. Conversely, the age of respondents has a substantial and adverse effect on the choice to utilize IT, indicating that older growers are less likely to adopt IT, consistent with the belief that senior farmers may possess lower IT abilities. Despite the general trend, experienced farmers, while more knowledgeable about technology adoption, are also more likely to employ IT for agriculture-related information (Okello et al., 2012; Paustian and Theuvsen, 2017). Family members residing near government facilities or with access to IT-related training in town exhibit a greater motivation to adopt IT, as they tend to be more open to positive initiatives to improve farming and assist producers (Kiiza and Pederson, 2012). Furthermore, participation in agricultural cooperatives, which often promote IT adoption and disseminate information through IT channels, increases the likelihood of IT utilization (Abdul-Rahaman and Abdulai, 2018). As highlighted in earlier research (Aker, 2011), a higher information literacy score is crucial for IT adoption and optimizing available resources. In particular, cooperative membership and certificate ownership can influence TE, potentially biasing the outcome. To address this issue of endogeneity, Wooldridge’s (2015) two-stage control function model is employed. The coefficients of the generalized residuals for the certificate and cooperative variables, presented in Table 4 as predictions from the initial phase of the control function, indicate that cooperation and certification are indeed endogenous in the IT selection model, with both associated coefficients being statistically significant.

The parameter estimate findings for the selection-corrected SF and conventional methods for the matched samples are presented in Table 6. The returns to scale and output elasticities of both models do not differ significantly from the unsampled dataset. The results from the selection-corrected SF method indicate that the coefficient of the sample selection bias variable ρ₁ for the IT group is statistically different from zero, consistent with a random sample. The current research, which examines traditional SF for IT users, once again highlights questions related to selection bias. The significant value of ρ₂, indicating a selection bias of IT non-users in SF, is unsupported by any empirical evidence.

According to the findings in Table 3, IT users had average TE values after matching 0.61 in conventional SF and 0.62 in selection-adjusted SF, respectively, compared to 0.55 for IT non-users in both indicators. In the matched sample, our analysis reveals that the TE variance among the IT users’ and non-users’ groups is larger than in the mismatched group, increasing from 0.03 to 0.07. Consequently, if selection bias induced by both apparent and unobserved factors is disregarded, the mean TE variance between users and non-users may be understated. This result aligns with research conducted in past studies. There are a few factors to consider regarding the TE mean score. First, existing results are consistent with recent studies (Dong et al., 2019; Liang et al., 2019), with an average value of approximately 0.062. However, compared with neighboring countries’ crop growers globally, including Vietnam, where growers had an average productivity value of 0.74 (Nguyen et al., 2021, 2023), or India, where the score is 0.77 (Murthy et al., 2009), Pakistani crop growers seem to have inferior TE values. One probable explanation for this disparity is Pakistan’s land tenure structure, which may not be as favorable to efficient crop-growing techniques as in other nations. Farmers’ capacity to make investments in land resources and increase TE is constrained by land utilization or transfer limitations.

Second, crop growers often have fewer effective scores than other crop producers in Pakistan. For instance, the TE values for fruit growers and crop farms were determined to be 0.83 and 0.9, respectively. One of the causes of this mismatch is the labor-intensive nature of farming operations, which includes activities such as hand weeding, several harvests, and various insect management (Stringer et al., 2009). Additionally, compared to certain other crops, vegetables are more sensitive to environmental conditions such as temperature variations, water availability, and soil health (Tripathi et al., 2016). Furthermore, the progress of farmers’ TE may be hampered by a lack of institutional and socioeconomic assistance, including cooperative help and extension services (Hongyun et al., 2020; Zheng et al., 2021). These outcomes imply a requirement for more empirical research as the efficiency impacts can be country or crop-specific. It is also important to keep in mind that the efficiency score amount can be impacted by various productivity evaluation techniques and variable settings (Madau, 2012, 2015).

Understanding the diverse impact of IT on TE is crucial for developing effective agricultural development strategies. To achieve this, we utilize the residualized quantile regression (RQR) model as suggested (Nascimento et al., 2019; Borgen et al., 2021; Korkmaz et al., 2021), providing a flexible method to assess treatment effects across the distribution of results. In the existing study, the RQR model is calculated in two steps. First, to decompose the variation in the treatment variable into two different mechanisms, one that can be described through the examined control variable and one that is orthogonal to a control variable the treatment variable (IT) is adjusted for the control variable using ordinary least squares. The residualized treatment variable is regressed in the second phase using the minimal absolute deviation approach. Finally, QTE can be calculated using observed data while correcting for selection bias by comparing quantiles (τ) of the outcome distribution for individuals with different treatment values of equation QTE.

The outcomes in Table 7 shed light on this investigation. Apart from the 90th quantile, coefficients demonstrate a positive and statistically significant correlation between IT usage and TE, mirroring our prior findings. In particular, there is a marginal uptick in the coefficient for IT treatment from the 10th to the 25th percentile. The most substantial impact manifests at the 25th percentile, showcasing a coefficient of 0.116, suggesting that IT adoption notably enhances farm efficiency at lower distribution quantiles (Zheng et al., 2021). These results imply that embracing IT offers more significant advantages to farms initially operating at lower efficiency levels, as they possess greater potential for enhancement. Conversely, at the 90th percentile, the effect is statistically insignificant, hinting that IT utilization holds less sway over the most efficient farms. This may be attributed to their already optimized production processes, possibly extensively leveraging other information channels.

Enhancing the TE of agricultural production remains a pressing concern in Pakistan, reflecting challenges encountered by numerous developing nations. A recent study delves into this issue by drawing insights from a sample of 600 farmers in rural Pakistan, aiming to discern the impact of IT utilization on the TE of crop production. The study employs the SF and PSM models to mitigate biases stemming from observed and unobserved factors. The research findings underscore that, when accounting for these biases, the disparity in TE between IT users and non-users holds both financial and scientific significance. This suggests that integrating IT into wheat crop production can yield positive outcomes for rural areas. Further exploration through the QTE method reveals a nuanced relationship between IT adoption and TE. The most pronounced effects are observed among the least efficient farmhouses, gradually diminishing in significance toward the median and ultimately becoming non-significant for farmhouses achieving maximum yield. This nuanced perspective highlights the varying impacts of IT on TE across different efficiency levels in the context of wheat crop production in rural areas.

The findings of the study have some important policy implications. First, emphasizing the positive impact of IT usage on crop production efficiency underscores the need for policymakers to invest in rural IT infrastructure and reduce access costs to promote technology adoption in rural areas. Second, tailored policies promoting emerging technologies should consider the diverse characteristics of smallholder farmers, with efforts to enhance access to information through various channels, including traditional agricultural services, farmers’ organizations, and digital platforms. Third, policymakers can establish technical training centers to provide advisory services, improve rural education opportunities, and facilitate technology adoption among farmers. Finally, while IT adoption is crucial, policymakers should diversify support mechanisms by collaborating with financial institutions, research bodies, cooperatives, and agricultural enterprises to offer financial, technological, and production support.

This study has some limitations. First, its focus only on wheat production may restrict the applicability of findings to other crops due to differences in agricultural extension services. Second, the small sample size limited to one province may compromise the representativeness of the results. Finally, using cross-sectional data prevents exploration of the dynamic impact of IT usage on TE over time. Future research should aim to address the limitations of the study. First, expanding the scope of crops studied will provide a more comprehensive understanding of the impact of IT on agriculture beyond wheat production. Second, increasing the sample size and considering multiple provinces can enhance the representativeness and generalizability of the results. Finally, utilizing longitudinal or panel data analysis techniques can facilitate the exploration of the dynamic effects of IT usage on TE.