1 Introduction

Many scholars discuss that acute labor shortage due to shrinking and aging of farmers has become one of the critical constraints of agricultural development in Japan. According to census by the Ministry of Agriculture, Forestry and Fisheries (MAFF), the labor force primarily engaged in agriculture has decreased from 1.76 million in 2015 to 1.36 million in 2020. Alarmingly, more than 70% of farmers were above the age of 65 years in 2020, compared with 65% in 2015 (MAFF, 2020). Under these circumstances, the Japanese government has encouraged the vigorous development of smart agriculture to overcome the disadvantages of agricultural labor shortage, improve agricultural production efficiency, and revitalize the progress of agriculture and rural areas (MAFF, 2022). Moreover, the widespread application of information and communication technology (ICT) in agriculture has proven crucial for optimizing the market activities, promoting the succession of agricultural skills, and boosting the development of agricultural informatization in Japan.

Meanwhile, a structural change toward consolidation is ongoing in Japanese agriculture, with the decline of agricultural households but the rise of large-scale farming and agricultural corporations in recent decades (EU-JAPAN CENTRE FOR INDUSTRIAL COOPERATION ECOS GmbH, 2021; Nanseki, 2021). The emergence of agricultural corporations has become the backbone of realizing large-scale production, heightening the strategic management of agribusiness and accelerating industrial clusters. Intensive adoption of ICT and smart farming (SF) by corporations is anticipated to allow for the technological optimization of agricultural production systems and food value chains, ultimately contributing positively to agricultural development. Ogata et al. (2019) analyzed the cost-effectiveness of ICTs for agricultural corporations using factor analysis and observed that the factors for production and accounting visualization are related to human resource development. Their factor scores comparisons by farm characteristics revealed three points: (1) ICT cost-effectiveness is greater for livestock farms than for farms producing other goods in terms of enhancing the profitability factor; (2) farms with higher sales place a greater value on production and accounting visualization factors than those with lower sales; and (3) farms with more employees place a higher value on production visualization factors than those with fewer employees. Nanseki (2019) and Nanseki et al. (2016) reported on interdisciplinary aspects based on ICT and smart farming technology by focusing on rice farming. Bucci et al. (2019) discussed factors affecting ICT adoption in Italian agriculture and reported Internet access, web pages, production standards, age, and educational background as the factors affecting successful adoption of management information systems on farms. However, the determinants of ICT and smart farming (ICT&SF) technology adoption by agricultural corporations in Japan remain unclear.

To this end, the objective of this chapter was to identify the determining factors of ICT&SF technologies adoption by Japanese agricultural corporations. Section 2 outlines empirical models, followed by a description of data sources and variables used in econometric analysis. Section 3 discusses the empirical results, and Sect. 4 presents the key conclusions.

2 Methodology and Data

2.1 Methodology

Previous studies have analyzed the adoption of a particular or several agricultural technologies by applying ordered probit models, multinomial logit regressions, and double-hurdle models (Knowler & Bradshaw, 2007; Zhang et al., 2020). In this chapter, we investigated the intensity of ICT&SF technologies adopted by agricultural corporations. Accordingly, the dependent variable is a count variable taking a non-negative integer value from 0 to 21. Thus, count data models were deemed appropriate to estimate the effect of potential influencing factors on the number of technologies adopted (Cameron & Trivedi, 1986; Isgin et al., 2008; Rahelizatovo & Gillespie, 2004). Count integer values were assumed to follow a compound Poisson regression, in which the number of technologies adopted and the probability density function of Y can be given as follows:

$$f(y_{i} |x_{i} ) = P\left( {Y_{i} = y_{i} } \right)\, = \,\frac{{e^{{\lambda_{o} }} \lambda_{i}^{{y_{i} }} }}{{y_{i} }},\,y_{i} = 0,\,1,\,2,\,3 \ldots$$
(1)

where \({y}_{i}\) is the total number of ICT&SF technologies adopted by the agricultural corporation i and \({x}_{i}\) is the expected determinant of ICT&SF technology adoption. The expected mean parameter (\(\lambda\)) of this function is defined as \(\lambda_{i} \, = \,\exp \,\left( {x^{\prime}_{i} \beta } \right)\), the \(\beta\) can be estimated using the maximum likelihood.

The Poisson model assumes the mean and variance of dependent variable are equal; that is, \({\lambda }_{i}\)=mean \(\left({y}_{i}|{x}_{i}\right)\)= variance \(\left({y}_{i}|{x}_{i}\right)\). However, when the conditional variance is greater than the conditional mean, overdispersion is the most likely situation (Ehiakpor et al., 2021). Thus, a negative binomial (of which Poisson is a special case) may be an appropriate count data handling procedure to accommodate the overdispersion issue by modeling variance as a function of mean. The variance in negative binomial model is given as follows:

$$Var\left( {Y_{i} |x_{i} } \right)\, = \,\lambda_{i} + \alpha \lambda_{i}^{2}$$
(2)

where \(\alpha\) is the dispersion parameter to be estimated. If α is zero, the negative binomial model is the same as the Poisson regression model, and the corresponding log-likelihood is log L = \(\sum_{i}log\left[\mathrm{Pr}({y}_{i})\right]\). In this chapter, the test indicated the presence of overdispersion, which led to the selection of a negative binomial model.Footnote 1

2.2 Data

2.2.1 Data Collection

The data used in this chapter were obtained from the “Business Development and Innovation in Agricultural Corporation Management” survey conducted by the Laboratory of Farm and Management at Kyushu University in 2019 (Nanseki, 2021). Information was gathered through mail questionnaires sent to agricultural corporations across Japan. The names of agricultural corporations were collected from the relevant publications, reports, and website of the Japan Agricultural Corporations Association (https://hojin.or.jp/).

In the survey, respondents were asked questions covering six parts: (1) basic information and operating policy of the corporation, such as corporate form, location, establishment year, development stage, annual sales/profit margin, operating targets in the next 5 years, and so on; (2) innovative realization of corporations within the past 3 years; (3) current status of ICT&SF technologies adoption; (4) detailed business content, management strategy, and self-evaluation; (5) social contribution and perception of the Free Trade Agreement (FTA); and (6) profile of corporate representatives, such as age and education.

The questionnaires were sent to 2885 corporations, and 505 corporations provided valid answers, resulting in the effective response rate of 18% (Nanseki, 2021). The outline and basic survey results is shown in Nanseki (2021). In this study, we eliminated the observations without sufficient supporting information on questions of technology adoption and deleted the missing data of corporate and representative attributes. After screening for the missing data of all variables, most respondents made a single selection for the indicators of corporate attributes, and only one respondent made multiple selections for corporation’s establishment background. Finally, 183 valid observations were used for further analyses.Footnote 2

2.2.2 Variable Description

The dependent variable used in this chapter was the number of technologies adopted by an agricultural corporation. It is a count variable that can be used to estimate the intensity of technology adoption. Specifically, we counted the number of combined technology categories involved in both ICT and SF technologies. According to the Food and Agriculture Organization of the United Nations (FAO), ICT is defined as “a broader term for Information Technology (IT), which refers to all communication technologies, including the internet, wireless networks, cell phones, computers, software, middleware, video-conferencing, social networking, and other media applications and services enabling users to access, retrieve, store, transmit, and manipulate information in a digital form.Footnote 3” According to MAFF (2022), “smart agriculture” or “smart farming” refers to the utilization of cutting-edge technologies, such as robots, artificial intelligence (AI), and the Internet of Things (IoT), in agricultural or farm management. Recent studies have distinguished SF technologies into the following types: (1) recording and mapping technologies, which collect precise data for subsequent site-specific application; (2) tractor GPS and connected tools, which use real-time kinetics to appropriately apply variable rates of inputs and accurately guide tractors; (3) apps and farm management and information systems, which integrate and connect mobile devices for easier monitoring and management; and (4) autonomously operating machines, such as weeding and harvesting robots (Fountas et al., 2015; Knierim et al., 2019). In this study, the ICT&SF technologies adopted by Japanese agricultural corporations are tentatively identified as two types. One refers to the smart farming technologies (SFTs) contained ICT and (2) common ICTs applied in SF.

The definitions and adoption rates of each technology categories are shown in Table 1. Three aspects including data monitoring and collection, operation automatization, and robotization, and business management, were involved, and 21 ICT&SF technology categories were described. The most frequently adopted technology category was financial management systems, such as bookkeeping and accounting, with an adoption rate of 84.2%. Advertisement for companies and products was a relatively frequently used technology category with an adoption rate of 65.0%. The third most frequently adopted technology category was sales information management, with an adoption rate of 61.7%. In contrast, technologies with relatively low adoption rates included “automation of crop cultivation machines/robots”, “automatic measurement of product harvest”, and “measurement of crop growth using drones and artificial satellites”, with adoption rates of 8.2, 7.7, and 5.5%, respectively. These trends are consistent with the statistics reported by Nanseki (2021).

Table 1 Definition and adoption rates of ICT&SF technologies
Full size table

The independent variables in our count data modelling covered wide range of corporation attributes and representatives characteristics, classified into the following 17 groups: (1) corporate form; (2) eligibility to own farmland; (3) location of corporations; (4) age of corporations; (5) establishment background; (6) human capital; (7) annual sales; (8) profit margin, (9) development stage of the corporations; (10) sales target for the next 5 years; (11) profit target for the next 5 years; (12) major product; (13) self-evaluation of ICT utilization and information management; (14) perception of FTA participation of Japan; (15) age of representatives; (16) educational background of representatives; (17) non-agricultural experience of representatives. The definition, along with the unit and expected signs, are listed in Table 2.

Table 2 Definition of the variables in estimation (Nanseki, 2021)
Full size table

3 Results and Discussion

3.1 Descriptive Results

Distribution of ICT&SF Technology Adoption. Figure 1 presents the distribution of the ICT&SF technology adoption rates by Japanese agricultural corporations. Of the 183, 175 corporations had adopted at least one ICT&SF technology category until 2019, indicating an overall adoption rate of 95.6%. In contrast, 4.4% corporations implemented none of these technologies. Majority (82.0%) of the corporations adopted 10 or fewer technologies, and only 18.0% adopted 11 or more technologies. Moreover, the observed Japanese agricultural corporations adopted nearly 6.6 technologies on average.

Fig. 1
A bar graph represents the distribution of the I C T and S F technology adoption rates by Japanese agricultural corporations. There is 95.6% of the overall adoption rate of I C T and S F technology by corporations until 2019, and 4.4% of the corporations did not implement any of these technologies.

Distribution of technology adoption frequency of agricultural corporations (N = 183) (The Questionnaire Survey on Business Development and Innovation in Agricultural Corporation Management in 2019)

Full size image

Summary of the Descriptive Statistics. Table 3 depicts the summary of descriptive statistics for all variables. Majority (84.7%) of the corporations are limited and stock companies. Approximately 86.9% corporations are judicially qualified to own farmland. Nearly 24.6% corporations are located in Tohoku, 23.5% are located in Kyushu and Okinawa, and only 1.6% are located in Hokkaido. The average age of the sampled corporations is approximately 19.0 years. Regarding establishment background, approximately 47.5% are solely owned corporation, established by a farmer and 26.8% are joint corporations founded by several farmers. Regarding human capital, the number of board members is approximately 3.6 on average, and the number of regular employees is approximately 11 on average. Nearly half of the corporations have a profit margin between 1 and 10%, while 20.8% are running in financial deficit. Regarding development stage, approximately 40.4% corporations are at the “growing stage,” compared with 16.4 and 6.0% corporations at the “mature” and “recession” stages, respectively. Regarding the operating target, the largest proportion of companies (approximately 29.5%) have set the target of 1.5 times sales growth in the next 5 years. Moreover, 83.6% corporations have set the target of 1–20% profit growth, compared with 10.4% corporations with a target of over 20% profit growth in the next 5 years. Regarding the major product, the corporations with major products as ‘paddy rice’ account for the largest proportion (18.0%), whereas the ‘beans and coarse cereals’ accounted the least, only for 1.1%. Moreover, approximately 8.7% corporations follow multiple crop farming. Regarding the profile of corporate representatives, over half of the representatives (54.6%) graduated from high schools and 36.6% from universities. Of the corporate representatives, 2.7% held a postgraduate degree.

Table 3 Result of descriptive statistics
Full size table

3.2 Empirical Results

We applied a negative binomial model to identify the potential determinants of ICT&SF technologies adoption by Japanese agricultural corporations. We tested two non-nested forms of the negative binomial model denoted NB1 (which is a negative binomial model with constant dispersion) and NB2 (which is a negative binomial model with no constant dispersion) and compared their estimates according to Akaike’s information criterion (AIC) and Bayesian information criterion (BIC). The results are presented in Table 4.

Table 4 Result of negative binomial regression model
Full size table

The result of NB1 revealed corporate form, eligibility to own farmland, sales targets, profit target, major product, self-evaluation of ICT utilization and information management, and educational background of representatives as the potential determinants of ICT&SF technologies adoption by Japanese agricultural corporations. Here we mainly discuss these indicators with parameters at 1 and 5% significance levels. First, the marginal effect of CFORM_3 on ICT&SF technology adoption was −2.431 at 5% significance level, indicating that cooperative agricultural corporations tend to adopt fewer technologies than limited companies. Second, the coefficient of FARML was positive and statistically significant at 5% level, indicating that corporations eligible to own farmland were likely to adopt two more technologies. Third, the self-evaluation of ICT utilization and information management significantly and positively affected technology adoption (p < 0.01). It demonstrated that corporations with a higher self-evaluation of ICT utilization and information management tended to use more ICT&SF technologies. Finally, the marginal effects of EDU_2 and EDU_3 are both positive statistically significant at 5% level, indicating the representatives who graduated from specialized schools and vocational colleges were more likely to adopt ICT&SF technologies. These results differ from the finding of Carrer et al. (2017), who demonstrated that university-level education positively affected the likelihood of technology adoption in farm management. This discrepancy may be explained by the fact that representatives who graduate from specialized schools and vocational colleges have more opportunities to receive specific agricultural knowledge and training lessons on farming skills and are, therefore, more willing to adopt technologies.

With regard to the empirical results at 10% significance level, first, the marginal effect of TSALE_2 was 1.637, indicating that corporations targeting 1.2 times sales growth in the next 5 years were likely to use two more technologies than corporations aiming to maintain the current sales. Second, the marginal effect of TPROF_5 was 3.443, indicating that corporations targeting 15–20% profit growth in the next 5 years were likely to use three more technologies than corporations that aimed to maintain the profit. Finally, the marginal effects of PROD_6 and PROD_12 were −3.144 and 2.493, respectively. Compared with the benchmark major product “paddy rice”, corporations operating “flowers and foliage plants” were likely to use three less technologies, whereas corporations operating “poultry” were likely to use two more technologies.

In particular, indicators with estimated parameters at 10% significance level were slightly different from the previous results, which based on 193 samples (see Table 5 in Appendix). Some variables with 10% significance level in the previous version, such as the number of board members and representatives’ age, were altered. As shown in Table 14.4, the number of board members promoted ICT&SF technologies adoption even the marginal effect is not significant. Similarly, the coefficient of AGE_R was insignificant as well, but still, it revealed a negative sign. This is also consistent with a previously reported finding from the adoption literature, which demonstrated a negative association between the age of decision-makers and technology adoption (Simmons et al., 2005).

4 Conclusion

Through a national questionnaire survey of “Business Development and Innovation in Agricultural Corporation Management”, this study identified the determinants of ICT&SF technology adoption by Japanese agricultural corporations. Negative binomial models were employed to examine the relevant corporate attributes and representative characteristics potentially affecting the technology adoption by agricultural corporations.

The results revealed that, of the 183 sampled corporations, 175 had adopted at least one ICT&SF technology until 2019, indicating an overall adoption rate of 95.6%. Among the 21 ICT&SF technologies, the most frequently adopted component was financial management systems, such as bookkeeping and accounting, with an adoption rate of 84.2%, whereas the least frequently adopted technology was the measurement of crop growth using drones and artificial satellites, with an adoption rate of 5.5%. Regarding the attributes of sampled corporations, majority (84.7%) of the corporations were limited and stock companies and 86.9% were qualified to own farmlands. In addition, 18.0% corporations operated paddy rice as major product and only 1.1% mainly operated beans and coarse cereals. Regarding the profile of corporate representatives, over half of the representatives (54.6%) graduated from high schools and 36.6% from universities.

The results of empirical models revealed corporate form, eligibility to own farmland, sales target, profit target, major product, self-evaluation of ICT utilization and information management, and educational background of representatives as the potential determinants of technologies adoption by Japanese agricultural corporations. Specifically, regarding corporate form, cooperative agricultural corporations tended to adopt fewer technologies than limited companies. Moreover, corporations eligible to own farmland were likely to adopt two more technologies. Regarding sales and profit targets, corporations aiming to increase their sales by 1.2 times the current value or raise their profits by 15–20% of the current margin in the next 5 years were likely to adopt more technologies than those aiming to maintain the current status. Compared with corporations operating paddy rice as the major product, those mainly operating flowers and foliage plants were likely to use less technologies, whereas those targeting poultry were likely to adopt more technologies. Moreover, the self-valuation of ICT utilization and information management positively affected technology implementation. Finally, in terms of corporate representatives’ characteristics, those who graduated from specialized schools and vocational colleges were more likely to adopt the technologies.