Expert workshops showed that about 22% of 587,000 ha of potatoes in Yunnan were planted to C88 in all seasons in 2014. In the winter season, when few potatoes are available for the processing industry in China,Footnote 9 56% of Yunnan’s roughly 60,000 ha of potatoes is covered by C88. C88 has an advantage in the winter season due to its resistance to late blight, which flourishes under the cool, wet conditions of Yunnan’s winter.
The potato value chain
Heavy involvement of the state on the input side is a defining characteristic of the Chinese potato value chain. Potato variety research in China is conducted by some 30 national and regional research organizations, and universities develop varieties to increase yields and address production constraints, particularly late blight. As a result, before C88 most cultivars were slow in maturing and did not have acceptable processing characteristics (Jansky et al. 2009). Variety testing takes many years as potential germplasms are tested under different agroecological conditions for multiple years. Once a variety is approved for release, research centers provide germplasm to groups capable of seed multiplication.
Interviews with YNU faculty and experts in the Yunnan potato industry revealed that most potato seeds in Yunnan are provided through the formal sector led by potato research institutes and universities and coordinated by the Chinese agricultural extension system.Footnote 10 These groups include seed companies and extension/experiment stations. Government extension agents and other officials build relationships with seed multipliers to determine if the variety is appropriate for their farmers. This determination may include running regional trials to evaluate the agronomic properties of the variety in their targeted areas. Seed companies will sell the seed either in the private market or to government (local, county-level, or provincial).
County officials and extension agents identify broadly suitable areas for the variety (basing this decision on agroecological conditions, farm sizes, access to markets and other factors) and work with village leaders to determine if a variety is appropriate for a village. Many villages do not have regular access to clean seed, and extension agents and officials are the primary sources of new and replacement seed. Households located in villages targeted for diffusion of C88 have the option to adopt C88; our data indicate that many villages have no access to the variety. Households residing in villages without access to C88 seed are less likely to be aware of the new variety and thus, to adopt.
Market channels vary by season of production. In the spring and summer seasons, potatoes are produced throughout Yunnan and sold locally. In the winter months, potatoes are produced mainly in the counties near the southern Myanmar border and processing potatoes are produced and are shipped to processors in Kunming and coastal areas. During winter months, C88 accounts for more than 50% of the potato area and C88 area expansion is constrained by the availability of seed. Our interviewees concurred that seed producers do not cater to industry needs and only in the past couple of years has certified C88 rebounded in response to demands from processors.
The enabling institutional environment for C88 might be described as a “perfect storm” of success factors. While the country was experiencing strong macroeconomic growth leading to growing demands for processed potatoes, potato technology-specific factors were present in Yunnan. A durable inter-institutional collaboration during the research phase (1990–1998) helped develop and validate the new technology. Domestic research institutions (mainly YNU) took a leading role to influence government policy and promote adoption of the variety. C88 was particularly attractive to a rapidly growing chipping industry lacking winter season supplies of potatoes and a set of private–public linkages within the potato chip value chain helped coordinate government policy. These policies supported dissemination and adoption of clean seed for the variety.
Diffusion was also promoted by engagement of the extension systems (with presence of Experiment Stations relatively near the potato growing areas) and farmers’ participation during the adoption phase through involvement of local communities. The presence of champions (the Chinese and CIP breeders) promoting the technology across all phases of the institutional environment is consistent with the literature describing similar cases (e.g. Jansky et al. 2009).
Intensity of adoption: Village-level model
Adoption of potato varieties at the village level is modeled using a FML; dependent variables are the adoption intensityFootnote 11 for seven varieties (see Table 2); for each village, these intensities sum up to one. Although the formal seed system constrains seed availability, all villages have access, through seed saving and informal transfer to many varieties. Conditions in the seed system do not eliminate village or household choices. As expected, C88 is more intensively adopted in areas with more commercial potential (due to the presence of larger farms and more people). The C88 adoption rate increases by 4.5 percentage points for each additional 10 ha of potato area and is marginally higher for larger (in terms of population) villages. Paradoxically, however, adoption intensity is also higher for more distant villages. As the distance between the village and the county town (where the county government is located, generally the most populous and developed area of the county) increases by 10 km, the C88 adoption intensity increases by 3.7 percentage points. Better average village education is associated with higher intensity of adoption (an additional year of education is associated with about an 8-percentage point increase in adoption), but experience growing potato, another indicator of human capital is associated with lower village adoption rates.
An advantage of the FML is that additional information about the relationship between C88 adoption intensity and potato-growing characteristics of the village can be gleaned from regression coefficients for the intensity of adoption of alternative varieties. For example, the adoption intensity of Mira is higher if a village has more households, and is lower if the village has a larger potato area, suggesting that intensive adoption of C88 displaces Mira in more commercialized areas. Mira is also more intensively planted in areas where farmers have more potato-growing experience, suggesting that C88 substitutes for Mira in villages with younger farmers. Hui-2, a relatively new variety, is frequently planted when farmers abandon C88 and it substitutes for C88 under certain conditions. For example, higher average education is associated with higher adoption intensity of C88, and lower adoption intensity of Hui-2.
The results show a tradeoff between village-level production of C88 and Mira. Except for the number of households in the village, all the variables that significantly affect C-88 adoption have marginal effects of opposite signs to those for adoption of Mira. C88 is also preferred in villages with better-educated farmers, and this variable has a significant and positive marginal effect only in the regression for C88 adoption. To gain more insight into adoption dynamics, we turn to household-level determinants.
Determinants of adoption: Household-level model
At the household-level, determinants of adoption include household, farm and village characteristics (Table 3). Results are intuitive and consistent with the village-level findings and with expectations. In particular, a household is more likely to adopt C88 if it has more working-age farmers and education of working members is greater. While experience is negatively associated with village-level intensity of adoption (Table 1), age is positively associated with adoption at the household level. With an additional working-age farmer, a household is 6.4 percentage points more likely to adopt C88. The probability of adopting C88 increases by about 1% as the average age of working-age farmers increases by four years and average years of education increases by one year. Comparing these findings to the village-level results, we see that better-educated farmers have higher adoption intensities and higher adoption, but older farmers are more likely to adopt, but less likely to adopt intensively.
Farmers with more access to technical assistance are also more likely to adopt C88. A household with a member in a farmer organization or who has been visited by extension agents in the previous year is 13.2 and 11.2 percentage points more likely, respectively, to adopt C88. So, access to extension and information services has a positive effect on C88 adoption.
Village characteristics also affect household decisions. For example, if the village had not experienced serious diseases or pest problems with potato in the past five years, the household is 8.4 and 28.9 percentage points more likely to adopt C88. These results might appear to be counter-intuitive, but may reflect two factors. First, since C88 has more resistance to late blight than other varieties, higher adoption may be associated with lower disease prevalence, particularly when the village-wide adoption intensity is high. In this case, the causality is reversed and higher areas of C88 reflect lower disease prevalence. Second, because C88 is targeted toward areas where commercial production is higher, seed vendors (mainly extension agents) may avoid areas with more pest and disease problems.
C88 has also been more widely disseminated in villages where farm sizes and potato areas are greater, providing evidence that the variety may have bypassed some of the poorer areas of Yunnan. Households in villages that are nearest to the metropolitan areas (city towns) are more likely to adopt C88, providing further evidence that the variety’s spread was more concentrated in favored areas where access to processors is highest.Footnote 12 Consistent with the village-level estimates, distance to county towns (in contrast to city towns, which tend to be larger) is positively associated with C88 adoption, and the combined (city, county town) results show that C88 diffusion efforts were focused on commercial areas nearest the larger cities. These efforts bypassed smaller villages, likely because of market access in the larger cities.
In general, results from village and household models are consistent. With better education, larger village potato area, and longer distance between villages and county towns, both village adoption rate and the likelihood that a household adopt C88 rise. Glauben et al. (2012) show larger holding sizes to be associated with higher persistent poverty in China, but villages with a higher population density and those located closer to cities have a lower probability of being long-term poor. Reliance on cropping as a single household business increases poverty persistence—agricultural activities yield only modest returns. In Yunnan, commercial production of C88 substituted for proximity to urban suggesting an alternate escape route from poverty.
Determinants of ever planting and disadoption of C88
The first two columns in Table 4 show the determinants of whether a farm household ever planted C88 and last two show the determinants of disadoption, conditional on having grown C88. We include additional covariates in this regression because factors such as commercial orientation and the importance of prices received are likely to be related with experimentation with new varieties. The study of ever adopting and disadoption decisions explores the determinants of this experimentation.
The determinants of ever adopting C88 are similar in sign and significance to the determinants of current planting of the variety (shown in Table 3). Larger farm sizes with less fragmented land holdings (reflected by the number of potato plots, conditional on total area planted) are more likely to have ever planted C88. Larger farms are also significantly less likely to disadopt, with a one hectare increase in landholding associated with an eight-percentage point lower probability of abandoning the variety, conditional on its adoption.
Commercial orientation, captured by the dummy variable reflecting the importance of price, is a strong determinant of adoption. Households who said price is important (very important) are 14 (17) percentage points more likely to have planted C88 compared to households that say price is unimportant. The importance of price is, however, not significantly associated with disadoption.
Further evidence of the commercial importance of C88 is reflected in the significant marginal effect of the importance of market demand. For adopters who state that market demand is very important, the likelihood of disadopting is 18-percentage points lower than farmers who say market demand is unimportant.
As in the case for current adoption, the presence of diseases and pests in the village are negatively associated with ever adopting C88, but, conditional on adopting, experiencing a disease problem in the village is negatively associated with disadoption. The negative sign on the marginal effect for disadoption is consistent with expectations, but the estimate is not significant. In contrast to disease problems, the presence of pest problems is positively and strongly associated with disadoption. Conditional on adoption, the presence of pest problems leads to more disadoption, which is to be expected because purchasers of C88 have high expectations for tuber quality and are unlikely to purchase commercial potatoes in areas with large disease infestations.
While location of a farm with respect to cities and market towns has similar effects on ever adopted and currently growing C88 (compare Tables 3 and 4), once a farmer adopts, location has only very small effects on disadoption. Proximity to a city is not significantly associated with disadoption, while being farther from a county town is negatively associated with disadoption. Villages that are distant from county towns have limited access to new potato seeds as discussed above, so variety turnover in such areas tends to be lower.
Economic surplus and impacts of diffusion on market-level outcomes
Several parameters are needed to estimate surplus changes using the model described above: adoption rates by year, equilibrium quantity and price of potatoes in the market, size of the shift in supply caused by C88 adoption (the K-shift in Eq. 1), supply and demand elasticities and exogenous changes in income and population which would lead to shifts in demand over time. These exogenous shifts in demand must be accounted for to isolate the market-level effects of C88 diffusion.
The C88 adoption rate in Yunnan from 1996 to 2015 is based on the land area adoption rates and area estimates for specific years. Adoption started with informal release of the variety in 1996 and was assumed to grow linearly from that time forward until a peak was achieved in 2007. In 2007, about 30% of total area under potato in Yunnan was planted to C88. Lacking further data on annual area adoption, we assume the variety declined in the province linearly since 2007 to 2015, for which the expert panel has specific estimates (SIAC 2.12018). With evidence of disadoption from the household-level data and the SIAC 2.1 database, a linear interpolation of the adoption profile begins with zero percent in 1995, peaks at 30% in 2007, and falls to 22% in 2014 (Fig. 3).
The price of potatoes was obtained by the community survey. Community prices, base year 2015, were weighted by quantity produced by variety by village and quantity weighted at the provincial level. The per-ton prices are divided by the Purchasing Power Parity (PPP) exchange rate of 3.56 Yuan to the US Dollar (OECD 2016). The weighted average price for all potato varieties in Yunnan in 2015 was $375 per ton.
In a small open economy, prices have to be adjusted to reflect the variation in the exogenously determined price from year to year while maintaining 2015 values. To do so, the nominal price for potatoes in Yuan per ton and GDP deflator for China from 1995 to 2015 were obtained from FAOSTAT and the World Bank, respectively. FAOSTAT only had prices for potatoes in China for 1996 to 2013. The proportionate changes for 2013 to 2015 were assumed to be the same as those from 2012 to 2013.
In an open economy model, demand is infinitely elastic. An elasticity of supply of one is assumed. Estimates of the potato supply elasticity for China are not available in the literature, and Alston et al. (1995) suggest using an elasticity of one when in doubt for annual crops.
Quantity produced, land area and K-shift
The land area under potato in Yunnan in 2014 is estimated to be 587,000 ha (Gatto et al. 2016). To infer land area under potato from 1996 to 2013 and in 2015, the percentage change in potato area from 1996 to 2015 in Yunnan is assumed to be the same as the percentage change for all China. The latter was computed using data on area under potato production in China between 1996 and 2015 obtained from FAOSTAT. Using the area estimates from Gatto et al. 2016, an annual growth rate of 2% was applied to project the area under potato in Yunnan backwards from 1996 to 2013 and forwards to 2015.
Household and community interview respondents indicated that variety-specific difference in inputs applied were minimal. We used expected differences in yield to compute the K-shift. Yields were calculated at the plot-level using household data. The estimated potato yield of non-C88 variety was computed as a weighted average based on land area for each variety. The weighted average yield for all potato varieties in Yunnan, C88, and all varieties excluding C88 were 21, 24, and 19 tons per hectare, respectively. Thus, C88 yields are substantially higher than other potato varieties.
Economic surplus results
C88 adoption has led to significant economic benefits to producers in Yunnan County. Assuming a small open economy model, total benefits, base year 2015, from 1996 to 2015 equal $2.35 billion with a present value of $2.84 billion.Footnote 13 In an open-economy model, all benefits are assumed to accrue to Yunnan C88 potato producers. These results were compared to those assuming a closed economy and an elasticity of demand of -0.65 (Ahmadi-Esfahani and Stanmore 1997). Under a closed economy (full details on model inputs and assumptions are available in Myrick 2016), where prices fall, benefiting consumers, the total surplus over the same period was estimated at $3.08 billion with a present value of $3.73 billion. Under model assumptions about demand, consumers captured 61% of the benefits, while 39% of the benefits went to producers.