Internet access is often presented as a gateway to wider markets, better prices and faster agricultural information. For many rural women involved in small-scale farming in South Africa, however, that gateway may still be difficult to reach from home.
A new peer-reviewed study published in Agriculture & Food Security on 28 September 2026 analysed Statistics South Africa’s 2019 General Household Survey to examine household internet access among rural women involved in agriculture. The headline result was stark: only 7.2% of the analysed population was recorded as having internet access in the household, while 92.8% did not.
The study also found pronounced differences across population group, education, income source, marital status, age and province. Yet the findings require careful interpretation. The analysis is cross-sectional, uses data collected in 2019 rather than current connectivity data, and the authors themselves report that a formal goodness-of-fit diagnostic indicated that their logistic regression model did not fit the data well. The paper is therefore most useful as a detailed picture of inequality in household connectivity within the dataset, rather than proof that any one demographic characteristic causes internet access.
Why household connectivity matters for small-scale agriculture
For a farmer, connectivity is not simply about browsing the web. It can affect access to information about prices, buyers, production practices, weather, inputs and market requirements. Digital communication can also reduce some of the geographic disadvantages faced by producers who live far from major commercial centres.
The researchers, F. S. Mbamba and P. Nsengiyumva, framed internet access as part of a broader rural livelihoods problem. Their central question was whether rural women farmers had internet access within their households and how that access varied alongside characteristics such as age, population group, marital status, education, income source and location.
They also considered the markets to which agricultural products were sold. These were grouped into local buyers from the district and buyers beyond the local district, including neighbouring cities and towns, formal South African markets and international export agencies.
The study drew on a large national household survey
The researchers used secondary data from the 2019 General Household Survey collected by Statistics South Africa. The underlying survey used a multistage sampling design. Primary sampling units were selected through a stratified design with probability proportional to size, followed by systematic sampling of dwelling units. The researchers then filtered the data to focus on rural women involved in agriculture.
The paper reports an analysed sample of 372,190 observations. Household internet access was treated as a binary outcome: access was coded as one and no access as zero. The explanatory variables included age group, population group, marital status, highest education level, income source and geographic coding. Descriptive statistics, chi-square tests, Phi and Cramer’s V association measures, and binary logistic regression were used.
The demographic profile reported in the study was broad. Black respondents made up 87.6% of the analysed population, White respondents 11.2% and Coloured respondents 1.2%. Around 59.7% had completed secondary education, 21.4% had primary education, 11.6% tertiary education and 7.3% no education. Grants were the most common reported income source at 44.5%, followed by salaries, wages or commission at 27.6% and business income at 18.5%.
Household internet access was reported for just 7.2%
Of the 372,190 observations reported in the paper, 26,672, or 7.2%, were classified as having household internet access. The remaining 345,518, or 92.8%, were classified as not having it.
That measure is narrower than general internet use. A person might still connect through a mobile phone away from home, use a public access point or rely on someone else’s connection. The study specifically focused on internet availability within the household, which makes the 7.2% figure a measure of household connectivity in this selected 2019 rural agricultural population rather than a current estimate of all internet use among South African women farmers.
Market participation was also heavily local. The researchers reported that 330,711 observations, or 88.9%, were linked to local buyers within the district. Only 41,479, or 11.1%, were linked to the combined group of buyers in neighbouring cities and towns, formal domestic markets or international export agencies.
This contrast matters because one of the promises of digital connectivity is that it can reduce the information and communication barriers that keep producers dependent on nearby markets. The study cannot show that internet access itself moves farmers into formal or export markets, but it demonstrates that both household connectivity and access to more distant markets were uncommon in the analysed population.
The digital divide was not evenly distributed
Population group showed one of the strongest associations reported in the bivariate analysis. Among farmers with household internet who reached local district buyers, 57.4% were White and 42.6% were Black. Among those without household internet who sold locally, 93.3% were Black, 5.2% White and 1.5% Coloured. The reported Cramer’s V value was 0.504.
The contrast became even larger in the subgroup reaching markets beyond the local district. Among farmers with household internet in this market group, 94.3% were White and 5.7% Black. These percentages describe the composition of specific subgroups, not the probability that an individual farmer of a particular population group will obtain internet access. That distinction is important when interpreting the size of the disparity.
Education was also associated with connectivity. Among farmers using household internet to reach local buyers, 58.0% had secondary education, 31.3% tertiary education and 10.7% primary education. No respondents without schooling were represented among those with household internet in this comparison. For the group using household internet to reach buyers beyond the district, 69.6% had secondary education and 30.4% tertiary education. The reported Phi and Cramer’s V statistic for education and household internet access was 0.195, with the chi-square association reported as statistically significant.
Income source showed another divide. Among connected farmers reaching local buyers, 49.4% reported salaries, wages or commission as an income source, 34.9% received grants and about 16% reported business income. The association measure reported for income source was 0.196.
Geography produced some of the largest differences
Connectivity also varied sharply by province. Among women farmers with household internet who reached local buyers, 42.8% were from the Western Cape and 42.6% from KwaZulu-Natal, while 8.0% were from Mpumalanga and 3.1% from the Northern Cape.
The geographic pattern differed for those reaching buyers beyond the local district. In that connected subgroup, 37.6% were from the Free State, 25.7% from the Eastern Cape and 10.4% from the Western Cape. The study reported a Cramer’s V of 0.556 for its geographic coding and household internet access comparison, one of the largest association statistics in the paper.
These percentages should not be read as a contemporary provincial ranking of rural broadband. They describe the composition of subgroups within a 2019 survey-derived analytical sample and are influenced by the number and characteristics of women farmers represented in each province.
Regression results pointed to large disparities, but the model raises a warning
The researchers then used binary logistic regression to examine the predictors jointly. They reported a Cox and Snell R-squared of 0.261 and a Nagelkerke R-squared of 0.648. In the model, White respondents had reported odds of household internet access 11.575 times those of the Black reference group. Tertiary education was associated with reported odds 7.141 times those of respondents with no education, while residence in the rural Free State was associated with odds 6.033 times the Western Cape reference category.
Those are large estimated differences, but they should not be treated as clean causal effects. The paper reports a Hosmer-Lemeshow test with p = 0.000 and explicitly concludes that the data do not fit the model well. That diagnostic weakens confidence in using the fitted regression as a precise description of individual probabilities. It is therefore safer to interpret the regression as further evidence of substantial inequality in the dataset rather than as a definitive prediction tool.
The authors also reported that unmarried women farmers were less likely to use household internet than married farmers, with an odds ratio of 0.457. Age was statistically significant overall, although the oldest group, aged 67 years and above, was not significant in the model.
What the findings mean for rural digital policy
The study’s strongest policy message is not that one demographic characteristic determines connectivity. It is that digital access sits inside a web of socioeconomic and geographic inequalities. Household internet was uncommon overall, while the connected population differed markedly by education, income, population group and province.
For agricultural policy, this means that simply placing market information online may not reach the farmers who could benefit most from it. Connectivity, affordability, digital skills and the relevance of information all matter. The authors recommend affordable or free internet access points, capacity-building initiatives and public Wi-Fi in community facilities, alongside stronger farmer organisations that can help circulate agricultural information.
There is also a market-access dimension. If digital tools are intended to help small producers compare prices or reach formal buyers, policy evaluation should ask whether farmers are merely connected or whether connectivity actually changes who they can sell to, the prices they receive and the profitability of their farms.
Important limitations keep the result in perspective
The most obvious limitation is time. The study was published in 2026 but analyses 2019 data. South Africa’s connectivity landscape has continued to change since then, so the 7.2% figure should not be presented as the current prevalence of household internet access among rural women farmers.
The design is also cross-sectional. It can identify associations but cannot establish that education, marital status, population group, income or location caused internet access. Unmeasured household wealth, infrastructure availability, device ownership, mobile coverage and other factors could contribute to the observed relationships.
The regression model deserves particular caution because its Hosmer-Lemeshow diagnostic was statistically significant and the authors state that the data did not fit the model well. In addition, several subgroup percentages are based on combinations of connectivity and market destination, making them unsuitable for simple population-wide comparisons.
Finally, household internet access is not equivalent to useful digital participation. Future research would be stronger if it measured connection quality, cost, device access, digital skills, actual online agricultural activities and whether internet use leads to measurable changes in sales, prices, productivity or farm income.
A large gap, and a need for newer evidence
The study captures a striking feature of South Africa’s rural digital divide at the end of the 2010s. In the analysed data, fewer than one in 14 observations had household internet access, and access was distributed very unevenly across social and geographic groups.
The next step is not merely to repeat the measurement. Researchers need to establish how much the picture has changed, whether improved connectivity is reaching historically underserved farming households and, most importantly, whether access is translating into better agricultural opportunities. A connection matters most when farmers can turn it into information, bargaining power and viable routes to market.
Source Information
Study: Mbamba, F. S. & Nsengiyumva, P. “Exploring factors influencing the use of internet in the household by women small-scale farmers in rural South Africa: evidence from GHS 2019.” Agriculture & Food Security, 15, Article 49 (2026).
Published: 28 September 2026.
DOI: 10.1186/s40066-026-00638-4.
Study design: Cross-sectional secondary analysis of Statistics South Africa’s 2019 General Household Survey, filtered to rural women involved in agriculture. The study reports N = 372,190 and uses descriptive statistics, chi-square tests, Phi and Cramer’s V measures, and binary logistic regression.
Data source: Statistics South Africa General Household Survey 2019.
Funding and competing interests: The article reports no competing interests. The study used anonymised secondary survey data collected by Statistics South Africa.
