Public and Private Grains Breeding

From the nineteenth to the late twentieth century, most new crop varieties in the United States came from public breeding programs run by universities or the USDA. Patenting new asexually propagated crop varieties first became legal in 1930 with the passage of the Plant Patent Act (Shelton and Tracy, Reference Shelton and Tracy2017). Intellectual property protections for new crop varieties produced from seed (as opposed to asexually propagated crops) were subsequently introduced by the Plant Variety Protection Act of 1970 (PVPA), making it the relevant legislation for protection of new wheat and barley varieties (Klotz-Ingram and Day-Rubenstein, Reference Klotz-Ingram and Day-Rubenstein1999). However, the protection offered by the PVPA was relatively limited in scope, as crop breeders had the right to use protected varieties for breeding and other research purposes. Additionally, the “farmer exemption” in the original PVPA allowed farmers to save, replant, and even sell seeds saved from protected varieties, although a 1994 amendment prevented farmers from selling (but not using) saved seeds protected under the PVPA without permission. Because of these weaknesses, a study of the effects of the PVPA on wheat breeding found that the PVPA alone did not contribute strongly to increased private investment in the wheat breeding sector (Alston and Venner, Reference Alston and Venner2002).Footnote ¹

Further developments in the institutional and scientific arenas strengthened incentives for private investment in plant breeding. The Bayh-Dole Act of 1980 played a key role in shaping the germplasm industry by allowing universities to patent technologies resulting from federally funded research and to issue exclusive licenses for patented material (Alston and Venner, Reference Alston and Venner2002). Patenting allowed universities to transfer intellectual property rights to private partners, as is done with exclusive varieties released by VT (Graff et al., Reference Graff, Cullen, Bradford, Zilberman and Bennett2003). These legal changes increased incentives for private participation in plant breeding, causing the size and number of private breeding programs to swell while those of the public sector dwindled (Shelton and Tracy, Reference Shelton and Tracy2017). As the regulatory landscape changed to better accommodate private breeding companies, advances in research methods and biotechnology contributed to increased private investment in plant breeding. It is argued that the surge in private patents beginning in the mid-1980s had more to do with changes in research methods, such as the emergence of modern biotechnology, than with the regulatory environment in the United States (Kortum and Lerner, Reference Kortum and Lerner1999).

Although the private breeding sector grew quickly to eclipse the public breeding sector in terms of total R&D investment, the two sectors focus on different types of breeding projects. In general, private breeders are more likely to focus on hybrid varieties for commercially successful crops, such as corn and soybeans, whereas public breeders are more willing to develop non-hybrid varieties for a wider range of crops, including small grains (Klotz-Ingram and Day-Rubenstein, Reference Klotz-Ingram and Day-Rubenstein1999). However, as investment becomes more heavily concentrated in private breeding programs, especially those utilizing modern genetic engineering techniques, higher-cost public breeding programs using conventional methods often receive less resources as a result, even in the case of less commercialized crops such as wheat (Knight, Reference Knight2003). As the plant breeding landscape continues to change, public breeding programs face the question of what their role should be.

Public and private plant breeding programs are not direct substitutes. A fundamental difference is that universities generally (though not always) share their germplasm with other universities and public programs (Shelton and Tracy, Reference Shelton and Tracy2017). Private breeders rarely share proprietary germplasm. All breeders are interested in improving varieties for the market, but private breeders are more likely to safeguard their successes from others in an attempt to maximize their own returns. This safeguarding may entail forgoing potential varietal improvements from collaboration and may limit farmer access to some varieties based on financial or geographic barriers. Public breeding programs may be more likely to balance royalties (which support their programs) with broad access to varieties. Shelton and Tracy (Reference Shelton and Tracy2017) surveyed public plant breeders and found that grants (which often have broad access as a priority) were more important than royalty funding in dictating the focus of their breeding.

Small Grains Breeding and Distribution at VT

The university-run breeding program in Virginia is supported by the Commonwealth of Virginia, USDA grants, the Virginia Crop Improvement Association (VCIA), and royalties. Roughly three quarters of the royalties collected are from exclusive crop varieties (developed by VT and licensed to private companies), with the remainder from public varieties released by the program (Santantonio and Hardiman, Reference Santantonio and Hardiman2021). Breeding objectives have changed over time based on producer demands, but generally include disease resistance and quality improvement, especially for specialty varieties such as malting barley and bread wheat (USDA National Institute of Food and Agriculture, n.d.). Varieties with highly specialized end uses or specific geographic adaptations are more likely to be licensed as exclusive varieties in order to meet specific client demands. Since 1990, the program has released 12 public wheat varieties and 11 public barley varieties. It has also licensed and released one exclusive barley variety and 67 exclusive wheat varieties.

The program uses field trials at agricultural experiment stations around the state. The role of these trials is twofold. The first is to compare yields of varieties to provide information to producers on how well the lines are adapted to area-specific production conditions. The second role is to generate germplasm that can be crossed and/or linebred to contribute to development of public and exclusive experimental varieties. Although improved lines (especially public ones) may be used as parents for later lines, much of the experimental germplasm created by the program changes from year to year as new lines are developed and older lines are discarded. Most germplasm used in these experiments is from VT, while a small but significant portion is from other universities or produced in collaboration with other universities.

The role played by exclusive varieties in the program is significant (Santantonio and Hardiman, Reference Santantonio and Hardiman2021). Licensing varieties for exclusive release provides two advantages. First, licensing allows the variety to have marketing support that private entities can provide but VT cannot. Consequently, although these varieties are marketed for less time by the private sector, they generally enjoy greater uptake by farmers compared to what they would have if they were marketed as part of a larger public portfolio (Thomason, Reference Thomason2020). Because the resources available to the VT breeding program to market public varieties are finite, the addition of private resources allows more varieties to reach more markets in more places. More varieties available in more markets increase the gross benefits from yield improvement generated by VT germplasm relative to what they would be in the absence of exclusive varieties. The second advantage is that royalties from exclusive varieties provide revenue to support breeding and fund the release of exclusive and public varieties.

Royalties from exclusive varieties might incentivize breeders to emphasize exclusive releases over public ones for their own benefit or to expand the program. However, the royalties are divided among the programs, administrative partners such as VCIA, and university administration, meaning that breeders and their programs do not reap the full benefits of these increased royalties. In fact, Shelton and Tracy (Reference Shelton and Tracy2017) report that breeders and their programs often receive less than half of the royalty funds they generate. Thus, incentives to maximize royalties are more attenuated than they may appear.

Distribution of seeds and administration of royalties from public varieties are undertaken in partnership with VCIA, which acts as the seed certification agency in Virginia. Varieties are marketed within Virginia under the administration of VCIA and in other states with the cooperation of their seed certification agencies. VCIA is also responsible for the production of “Foundation” seed, which is distributed to selected farmers for production of certified seed which is then sold to farmers (Thomason, Reference Thomason2021b). Royalties are collected based on certified seed production and split between VT and VCIA, with smaller royalties collected for out-of-state production (Hardiman, Reference Hardiman2021).Footnote ² The process of creation, administration, and distribution of public varieties is shown in Figure 1. Exclusive varieties are produced and distributed by the entity to which it is licensed, and royalties are paid to Virginia Tech Intellectual Properties (VTIP), with payments disbursed to VCIA, other administrative funds within VT, and the breeding program itself. The process for exclusive varieties is shown in Figure 2.

Figure 1. The creation, licensing, and distribution of public varieties created by the Virginia Tech small grains breeding program and the movement of revenues, royalties, and funding from these varieties.

Figure 2. The creation, licensing, and distribution of exclusive varieties created by the Virginia Tech small grains breeding program and the movement of revenues, royalties, and funding from these varieties.

Conceptual Model

It is assumed that farmers choose to plant the bundle of crop varieties that maximizes expected utility. In general, expected utility increases as farm profits increase, although farmers may also derive utility from traits such as environmental impacts and yield risk reduction due to drought resistance, among others. Adoption may also be impacted by barriers such as incomplete information on the part of farmers or unequal availability across locations. In expected utility theory, mean profits and their variability are assumed to matter. Farmers may be reluctant to adopt new varieties quickly because of uncertainty, but new varieties can be associated with lower risks of crop loss due to drought, insect pests, and plant diseases. Farmers may plant one or a mix of varieties depending on their preferences (Useche, Barham, and Foltz, Reference Useche, Barham and Foltz2013). As varieties age, they become more susceptible to insect pests and diseases (Brennan and Murray, Reference Brennan and Murray1995), so farmers constantly adopt new varieties for resistance and to take advantage of other new traits.

We assess benefits to farmers from the VT breeding program using methods applied by Pardey et al. (Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2004, Pardey et al., Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2006). The approach allows for trial data that do not include all varieties in all locations in all years, as is the case for the VT program, where new varieties were released over time. It also accounts for adoption and disadoption over time. Although we use the methods used by Pardey et al. (Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2004, Pardey et al., Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2006) to estimate changes in yield and their associated benefits, we do not apportion the source of these benefits between VT and partner institutions as do Pardey et al. We lack information on the genetic lineage of the varieties in the trials or the contribution of work and resources between VT and partners institutions.

Following Pardey et al. (Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2004, Pardey et al., Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2006), we take proportional yield improvements from research (k) and apply them to the value of total production in Virginia (shown here as the product of commodity price P and total production Q) to assess total benefits (B) from small grains research in a given year t:

(1) $$B_{t}=k_{t}P_{t}Q_{t}$$

It is assumed that inputs are held constant across varieties in an experimental setting and that differences among varieties in a specific place at a specific time result from varietal differences and differences in environment rather than management. This assumption allows regression of experimental yields on variables reflecting variety, trial location, time, and weather. Experimental yields for observation i of variety j in each year t and test site s are estimated as follows:

(2) $$Y_{ijst}=\alpha _{0}+\alpha _{j}+\boldsymbol{\beta }\boldsymbol{x}_{ijst}+\delta _{S}+\gamma _{t}+\boldsymbol{\phi }\boldsymbol{W}_{st}+\varepsilon _{ijst}$$

where Y _ijst is the experimental yield in bushels for observation i of variety j at site s in year t; α _j is a variety effect; x is a vector of four possible combinations of tillage and management treatments applied to an observation in a given year;Footnote ³ δ _S is a site effect; γ _t is a year effect; W _st is an index of weather during the wheat growing season for site s in year t; Footnote ⁴ and ϵ _ijst is the model residual. For selected regression results, please see Tables A1 and A2.

Using the parameters estimated in equation (2), we calculate fitted experimental yields (in bushels per acre) for each variety at each site in each year as:

(3) $$\hat{Y}_{jst}=\alpha _{0}+\widehat{\alpha _{j}}+\widehat{\delta _{S}}+\widehat{\gamma _{t}}$$

where α _j is the impact of variety selection on yield, δ _S is the impact of site conditions on yield, and $\widehat{\gamma _{t}}$ is the impact of growing conditions in each year (after controlling for weather). In equation (3), the effects of management ( $\widehat{\beta x}$ ) and weather variations ( $\widehat{\phi _{s}}W_{st}$ ) on yield are set equal to zero.Footnote ⁵

Experimental yield indices are constructed for varieties released by the program (represented here by J’ to denote a subset of J, the set of all varieties enrolled in the field trials) using the fitted yields averaged across all sites for each year combined with the observed area adopted by variety and year as:

(4) $$Y_{t}^{a}=\sum _{j=1}^{J'}\sum _{s}\hat{Y}_{jst}\pi _{jt}$$

where Y _t ^a represents the experimental yield given current adoption patterns, π _jt represents the area of adoption of variety j (including Virginia and other states where the variety is grown) as a proportion of total area A of all wheat/barley varieties released by the program ( $\pi _{jt}={A_{jt} \over A_{t}}$ ).Footnote ⁶ In this case, A _t refers the total area planted to varieties released by the VT breeding program, while A _jt refers to the area planted to a specific variety j released by the program. Variety-specific yield indices are summed across all varieties to create a weighted average that is proportional to variety acreage in the region and year. Although π _jt contains acreage from both inside and outside Virginia, trial data for these varieties are available only from the Virginia trials, so the same fitted yields are used for in- and out-of-state acreage.

The actual-acreage yield index shown in equation (4) is compared to counterfactual yields given adoption patterns of varieties in the base year. This counterfactual represents what yield would be if no new varieties had been released and farmers continued to use only those varieties available in the base year. The counterfactual takes the form

(5) $$Y_{t}^{b}=\sum _{j=1}^{J}\sum _{s}\hat{Y}_{jst}\pi _{jb}$$

The counterfactual is the same as the actual-acreage yield index except that proportional acreage for each variety is determined by acreage planted to the variety in the base year divided by total area in the base period, π _jb . Only varieties available in the base year are used to calculate the counterfactual yield index and their proportions of total yields are fixed, while the actual-acreage yield index allows the weights to change as new varieties become available. While yields change over time for the counterfactual yield index, the varieties used and the weights attached to the yield index do not. The counterfactual shows what yields would have been in the absence of improved varieties from the program, while the actual-acreage yield index shows the fitted yields from new varieties released by the program and account for adoption of new varieties over time.

Proportional yield improvements attributable to varietal improvement ( $k_{t}$ ) are obtained by comparing actual yield and adoption (in terms of acres planted) to projected yield and adoption rates in the absence of crop improvement research ( $k_{t}=({Y_{t}^{a}-Y_{t}^{b} \over Y_{t}^{a}})$ ). We assume that k _t is the same for in-state and out-of-state areas. In reality, different varieties are likely to be better-adapted to some states than they are to Virginia and less well-adapted in others. However, the weighted indices of actual and counterfactual yields include both in-and out-of-state areas, so it is assumed that the yield indices, and thus, k _t are the same for both in-state and out-of-state acreage. This allows for a rough estimate of out-of-state benefits in the absence of trial data for these areas. Because out-of-state acreage is such a large portion of total acreage planted to VT varieties, out-of-state benefits are an important portion of program benefits, so even a rough estimate of these benefits is important in understanding the scope of the program.

k _t is then combined with estimates of the fitted value of production for acres planted using VT varieties in a given year (Pardey et al., Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2004, Pardey et al., Reference Pardey, Alston, Chan-Kang, Magalhães and Vosti2006) to find total benefits for that year as B _t :

(6) $$B_{t}=k_{t}P_{t}(\sum _{j=1}^{J}(\hat{Y}_{jt}A_{jt}))$$

In equation (6), fitted production of varieties released by the program is substituted for Q in equation (1). Fitted production for a variety is defined as fitted yield for a variety averaged across all sites (Ⓨ _jt ) multiplied by acreage planted in that variety (A _jt ).Footnote ⁷ Because the area of each variety grown in the area surrounding each site is not known, fitted yields are averaged evenly across all sites in the field trial data. Fitted production is then summed across all varieties. Because sales by private entities located in Virginia are counted as Virginia acreage and because acreage for exclusive varieties cannot be attributed by state, actual yields and production cannot adequately be disaggregated by location, so fitted yields are used. Although use of fitted yields and fitted production may cause benefits to be biased upwards, it allows for some estimation of benefits, which would not otherwise be possible due to attribution issues.Footnote ⁸

Finally, total benefits from the program put into are 2018 dollars using the Producer Price Index for grains (U.S. Bureau of Labor Statistics, 2021) and then compounded:

(7) $${\rm PV}(B)_{t}=\sum _{t=0}^{T}B_{t}*(1+r)^{t}$$

where t runs from 2018 (t = 0) to 2000 (T = 18). We assume a discount rate (r) of 3% to account for the riskless opportunity cost of research funding. Benefits discussed in the paper are in 2018 dollars.

Because Virginia is a small producer of wheat and barley and VT varieties are small proportions of out-of-state areas, we assume that the program will not impact market prices.

Description of Data

Experimental data are obtained from field trials conducted by VT breeders from 1991 to 2018 (Table 1). These data cover nine test sites across the state and include over a thousand varieties, with more than 20,000 observations for barley trials and 67,000 observations for wheat. Barley yields in the trials averaged 101.56 bushels per acre with a standard deviation of 27.02, while wheat yields averaged 74.95 bushels per acre with a standard deviation of 17.77. Summary statistics of yield and selected other variables are shown in Table 2. Because the trial data available to us did not contain information on crop quality characteristics and because representative data sets containing such information are not publicly available, we assess varietal improvement purely in terms of yield.

Table 1. Description of variables used in regression and other analysis

Source: Temperature and precipitation data are taken from National Climatic Data Center historical data. All other data are from Virginia Tech wheat and barley field trials.

Table 2. Summary statistics of selected variables ^a

^a In this table, yield is presented in bushels. The variables “No-till,” “Treated,” “VT,” “Private,” “Public,” and “Exclusive” are presented as proportions; for instance, the mean of 0.717 for VT in the barley data set indicates that 71.7% of the trials used VT germplasm.

Source: Virginia Tech wheat and barley field trials.

Varieties in the trial may be public, private, exclusive, or experimental lines. The data set also identifies public, exclusive, and experimental lines by source. Many lines tested in Virginia are experimental lines from other universities, privately developed germplasm from seed companies, or public varieties from the USDA or other universities (Figure A1). The barley trials include relatively more public germplasm (13.18% of observations) with no exclusive varieties and very few private varieties (0.24%). The wheat trials are more diverse, with observations consisting of 6.32% public, 5.30% exclusive, and 19.53% private germplasm. The remainder consists of experimental germplasm from VT or other universities.

Area planted by year to each variety, shown in Table 3 is estimated for public varieties released by the program since 1990 based on acreage data from 2000 to 2018 and is used to estimate the proportion of total area cultivated in a given variety (π _jt ). For public varieties, acreage estimates can be separated between Virginia and out-of-state acreage. Acreage data for public varieties are constructed based on royalty receipts by state provided by VCIA, which are used to estimate certified seed sold using the royalty rate per unit of seed. We assume that new certified seed acreage accounts for roughly 60% of total acreage, while saved seed accounts for roughly 40% based on estimates from industry experts (Thomason, Reference Thomason2021c). Public variety royalties paid by seed producers based in Virginia are counted towards Virginia acreage, whereas royalties paid by universities and crop improvement organizations in other states are counted toward out-of-state acreage. Historical data for wheat and barley production and prices in Virginia (for Virginia acreage) and the United States (for acreage in other states) are used to construct the value of production estimates for each year (National Agricultural Statistics Service, 2020). These value of production estimates are then reported as in-state or out-of-state to allow comparison of the impact of public varieties in Virginia with their impact in other states.

For exclusive varieties, royalty receipts from VTIP are used to estimate acreage using rates per unit as above. However, because licensees for exclusive varieties generally operate in more than one state and royalties are generally paid by one or a small number of multi-state licensees, acreage estimates could not be separated by state for these varieties. Therefore, US prices were used for these varieties.

Exclusive varieties that were omitted from the variety trials were not included in this analysis. Amaze 10, the only exclusive barley variety licensed by VT during the study period, was omitted for this reason, meaning that all exclusive varieties analyzed are wheat varieties. Exclusive varieties were also omitted if royalty rates per unit were unavailable, making it impossible to estimate acres planted in the variety from royalty data. Benefits were estimated for 33 exclusive wheat varieties in total. For Amaze 10 and 13 of the missing wheat varieties, enough royalty data were available to make rough estimates of acreage and yield impacts possible. These estimates are presented in the results section but are not formally considered part of the analysis due to the lack of concrete data. For the remaining 21 exclusive wheat varieties, no royalty data are available. This makes reliable estimation impossible and may also suggest that these 21 varieties were licensed but never actually made it to market.

Weather data were compiled by the National Climatic Data Center (National Climatic Data Center, n.d.). Total precipitation and mean temperature by month during the main wheat and barley seasons in Virginia (October to March) were collected for each trial test site from 1991 to 2018. Data were taken from historical records for the National Weather Service station in the ZIP code nearest to the test site. In cases where the nearest weather station lacked complete records, the closest station with complete records was used. In cases where data from neither of the two closest stations were complete, the more complete station was used as the primary data sources, with the less complete station used to fill in where applicable. Precipitation data were summed, and temperature data were averaged over the growing season. Deviations from a time trend were entered in the regression.

Empirical Procedures

Yield is regressed on variety, year, location, management practices, and weather to provide fitted yields for each variety. Although the regression model was used to calculate fitted yields for both public and exclusive varieties, the actual and counterfactual yield indices (Y _st ^a and Y _st ^b ) were calculated separately. This separate calculation preserves the ability to report in- and out-of-state benefits separately for public varieties, where it is possible to separate the two, while disaggregation is impossible for exclusive varieties.