The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1095

63 Farmers’ preferences for varietal trait improvements: The case of rice farmers in Nueva Ecija, Philippines Farmers’ preferences for varietal trait improvements: The case of rice farmers in Nueva Ecija, Philippines Rio Maligalig and Matty Demont Farmers’ preferences for varietal trait improvements: The case of rice farmers in Nueva Ecija, Philippines Rio Maligalig*1 and Matty Demont**1

Abstract

Farmers have their own preferences for agricultural technology attributes, which have been found to significantly influence adoption decisions. However, these are not always known nor do they always match with the objectives of the researchers. To understand farmers’ preferences for rice varietal trait improvements (VTIs), we conducted a framed field experiment. The experiment provided the farmers the opportunity to participate early in rice breeding research by expressing their need for trait improvements. In the experiment, farmers were given an endowment fund of 100 Philippine pesos and were asked to invest it among the VTIs they prefer and need using the Investment Game Application (IGA), a newly developed application for eliciting preferences. Farmers were sampled from randomly selected villages in three municipalities in Nueva Ecija, a major rice producing province in the Philippines. In total, 122 households joined the experiment, with both husband and wife participating. We use the fractional multinomial logit model to examine the relationship of the proportion invested to VTIs with various factors that may influence farmers’ preferences. Results indicate that market and climate change information, wet season cropping, hybrid varieties, and farm size are among the factors that influence farmers to invest in trait improvements. Moreover, results of the gender-specific analysis indicate that there are differences in the factors that influence husband and wife in investing in trait improvements. Overall, information from this study can assist breeders in their efforts to make rice breeding more resource efficient and client-oriented, which could help facilitate the adoption of new and improved varieties. Keywords: Farmer preferences, investment games, field experiment, trait improvements, rice JEL codes: Q12, Q16

1 The first and second author contributed equally to this work. * Corresponding author, Centre for Global Food and Resources, The University of Adelaide, Australia, [email protected] ** Corresponding author, Social Sciences Division, International Rice Research Institute, Philippines, [email protected]

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1096

Introduction

Agricultural researchers are constantly faced with decisions on how to best allocate scarce research resources. It is important that these limited resources be allocated efficiently to generate outputs that will benefit the end-users, which are the farmers. Priority-setting exercises are usually done to guide resource allocation decisions (Fox, 1987; Gollin, 2000). The main idea in priority setting is “to achieve allocative efficiency which requires that the products or outputs of the organization are those of maximum value to the organization’s clientele, given the organization’s comparative advantage in a larger system of organizations” (Evenson, Herdt, & Hossain, 1996). The common features of the approaches used in priority setting include: (i) the identification of criteria for assessing research activities or themes, which usually include economic impact or efficiency, poverty elevation, sustainability, and capacity building; and (ii) estimation of likely adoption and probability of success. In most of the assessment and estimation needs in priority setting, it is the scientific staff and senior management who are involved and consulted. However, as end-users of agricultural technologies, farmers have their own preferences for technology traits and have specific production and marketing requirements that they need to consider (Pingali, Rozelle, & Gerpacio, 2001). For example, farmers need to make trade- offs in choosing which varieties to grow to satisfy both the demand of their production environment and consumers. These preferences have been found to significantly influence adoption decisions, aside from socio-economic, demographic, and institutional factors (Adesina & Baidu-Forson, 1995; Adesina & Zinnah, 1993; Kshirsagar, Pandey, & Bellon, 2002; Pingali et al., 2001; Sall, Norman, & Featherstone, 2000). But these preferences are not always known nor do they always match with the objectives of the researchers (Hellin, Bellon, Badstue, Dixon, & La Rovere, 2008). Thus, there are technologies which have not been adopted or for which adoption is limited. In the recent years, several studies have been conducted to better understand adoption decisions in the context of farmers’ perceptions or preferences for variety traits (Adesina & Baidu-Forson, 1995; Adesina & Zinnah, 1993; Fisher & Snapp, 2014; Ghimire, Wen-chi, & Shrestha, 2015; Hintze, Renkow, & Sain, 2003; Joshi & Bauer, 2006; Kshirsagar et al., 2002; Lunduka, Fisher, & Snapp, 2012; Sall et al., 2000; Smale, Bellon, & Aguirre Gomez, 2001). These trait-based studies conducted household surveys to elicit farmers’ subjective assessment of variety traits. Both production and consumption traits were assessed since most farming households in the developing countries both consume and sell their produce. Results of most of these studies revealed that both production and consumption attributes significantly influence adoption decisions. This is contrary to the traditional notion that only yields condition variety choice. Stated preference techniques are also used to elicit farmers’ preferences for variety traits. Among the family of stated preference techniques, the choice modeling or the multi- attribute valuation (MAV) is the more efficient method for examining preferences for multiple attributes simultaneously and the trade-offs between them (Merino-Castello, 2003). Ward,

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1097

Ortega, Spielman, and Singh (2014) used choice experiment, a choice modeling technique, to examine preferences of rice farmers from Bihar, India for variety traits embodied in hybrid and non-hybrid varieties. They specifically examined preferences for drought tolerant trait as drought is a major constraint to rice production in most of India. A similar study was done by Arora, Bansal, and Ward (2015) in Odisha, India. Dalton, Yesuf, and Muhammad (2011), on the other hand, examined farmers’ preferences for drought tolerance in maize varieties in Kenya using choice experiment as well. Horna, Smale, and Oppen (2007) and Baidu-Forson, Ntare, and Waliyar (1997) used contingent ranking to examine farmers’ preferences for variety traits. These choice modeling basically assess the relative value (e.g. willingness-to-pay) that farmers place on different crop variety traits. In this study, we used an experimental methodology based on investment games to elicit rice farmers’ preferences for varietal trait improvements (VTIs). Investment games have been traditionally used to measure trust and trustworthiness (Berg, Dickhaut, & McCabe, 1995). It involves a series of exchange wherein participants are endowed with a certain amount and assigned a particular role. At the end of the game, participants earn based on their decisions and decisions of other participants. To our knowledge, investment games have not been used to elicit preferences. One exemption is the study of Paris et al. (2001) who used a game to elicit farmers’ preferences for rice variety traits in eastern India. Farmers were asked to allocate a particular amount among different variety attributes. Unlike Paris et al. (2001) who used a hypothetical amount, our experiment involves real budget constraints and incentives. Our study contributes to the existing preference elicitation methods used for crop variety traits through the following. First, our approach is forward-looking such that we elicit preferences not for the variety traits per se, but for trait improvements. In most of the existing elicitation methods, preferences are obtained from the traits themselves. But farmers are able to identify and suggest improvements in the agricultural technologies they use to make these more suitable to their needs (Pingali et al., 2001). Moreover, since an aim of this study is to provide feedback to breeders to help them set research priorities and allocate resources efficiently, the focus is on trait improvements. Preferences for trait improvements are not commonly elicited for crop varieties, but these were the focus in two studies in animal breeding. Byrne, Amer, Fennessy, Hansen, and Wickham (2012) examined preferences of experts and farmers from the Irish sheep industry for sheep trait improvements to contribute in setting breeding objectives. In another study, Martin-Collado et al. (2015) analyzed dairy farmers’ preferences for improvements in dairy cow traits to provide feedback to a review of national breeding objectives for the Australian dairy industry. The second contribution of our study is that we confront farmers with the same resource constraint and risk faced by breeders, which allow farmers not only to express their preferences but also compel them to prioritize trait improvements they need. Third, as with other economic experiments, our approach involves real money and real returns, which are based on farmers’ choices. Stated preference methods rely on hypothetical scenarios, which make them prone to hypothetical bias (Hensher, 2010; List & Gallet, 2001; Little & Berrens, 2004; Murphy, Allen, Stevens, & Weatherhead, 2005).

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1098

Although it can be argued that the empirical evidence for this is mixed, we address this right away by making our approach incentive-compatible. Thus, the objective of this study is to understand farmers’ preferences for VTIs by examining the investment allocation patterns across trait improvements and the factors that may influence these allocation patterns. We also examine whether there are gender differences in the preferences and the factors that influence these. To achieve our objective, we conducted a framed field experiment with selected rice farming households in Nueva Ecija, Philippines. The experiment was carried out using the Investment Game Application (IGA), a newly developed application for eliciting preferences for rice VTIs (Demont, Custodio, Villanueva, & Ynion, 2015). To examine if there are gender differences in preferences, we invited both husband and wife to participate in the experiment. The Philippines, specifically the province of Nueva Ecija, is an ideal setting for studying rice farmers’ preferences for VTIs. Rice continues to be an important crop in the country and Nueva Ecija remains to be a major rice producing province. Adoption of innovations in agricultural technologies has played an important role in increasing production and improving the livelihoods of rice farming households (Mariano, Villano, & Fleming, 2012; Villano, Bravo-Ureta, Solís, & Fleming, 2015). A major contributor to gains in production is the adoption of modern varieties. Several studies like that of Estudillo and Otsuka (2006); Herdt and Capule (1983); Launio, Redondo, Beltran, and Morooka (2008) have documented the adoption of modern varieties in the Philippines since the beginning of the Green Revolution. A recent study by Laborte et al. (2015) analyzed the variety traits important to rice farmers in Central Luzon, Philippines based on the agronomic, grain quality, pest and disease resistance traits of the adopted varieties. The study found that farmers consider high yield, good grain quality, and resistance to pests and diseases important when selecting for rice varieties to plant. However, it would also be interesting to get farmers’ insights on the traits they want to be improved to better adapt the varieties to their environment and consumer needs. It would also be useful to understand the trade-offs in variety traits they are willing to make to have their most ideal variety. The remainder of the paper is organized as follows. The next section describes the experimental approach, while section three describes the econometric approach. In section four, the empirical results are presented and discussed. The last section provides the summary, conclusions, and recommendations.

Experimental Approach

Experimental Design

The experiment was framed around a hypothetical situation wherein a breeding program received a large grant from a donor. The grant was then distributed in small shares among farmers. As shareholders in the breeding program, farmers were given the opportunity to allocate their share, an endowment fund, to several alternative breeding programs for

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1099 improving varietal traits. This was done through the use of IGA. In IGA, farmers selected their preferred traits to be improved by pulling the VTI bars to the level that they wanted a particular trait to be improved using the up and down spin buttons (Figure 1). Farmers’ investments in the VTIs yielded an immediate return, subject to risk, which is defined as the probability that the level of improvement they selected will be achieved. Returns to investment in breeding research would normally be realized only after a new variety is released and adopted. This process will take about six years. In our study, breeding investment is framed as a single-period investment such that returns will be calculated and given right after playing the game. At the start of the experiment, farmers were asked to identify a model variety, which is the basis to improve upon to obtain their ideal variety. The model variety can be farmers’ most preferred or popular variety, which they may have or may not have grown yet. Farmers selected from among 10 VTIs that they prefer to be improved. These VTIs can be broadly categorized into (i) grain quality traits – slenderness, aroma, stickiness, and head rice recovery; (ii) loss reducing traits – lodging tolerance, disease resistance, insect resistance, abiotic stress tolerance, and reduction in shattering; and (iii) agronomic – earliness. The specific baseline and target metrics on which the IGA is calibrated is shown in Table 1. The experiment was comprised of four information treatments to test whether there would be differences in farmers’ preferences when given particular information. The first information treatment is the control, where no information was provided. The second is the market information to which information on the most preferred rice traits of urban (Metro Manila) consumers was given. The third treatment is the climate change information. The information provided in this treatment includes increasing climate variability and the rise in frequency of extreme weather events, which can bring more frequent droughts, floods, and more uncertainty in rainy/wet season onset. The fourth information treatment combines both market information and climate change information. The IGA was repeated over six rounds by each household. Husband (H) and wife (W) played the IGA for two seasons – wet (WS) and dry (DS) – independently, and then jointly (J) for two seasons as well. In each round, participants had an available endowment fund amounting to 100 Philippine pesos (PHP hereafter) (around USD2.10) to invest in VTIs.1 This amount, however, was not given in cash as the final pay-off was given at the end of the experiment and was based on one of the six rounds. To determine which among these six rounds will be the basis (binding round) for payment of returns, they were assigned a number in a dice: 1 – H/WS, 2 – H/DS, 3 – W/WS, 4 – W/DS, 5 – J/WS, 6 – J/DS. The dice was rolled after all the participants have finished the IGA.

1 At the time of the experiment (February 2016), one US dollar was equivalent to approximately PHP48.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1100

Sampling

Study Sites

We purposively selected Nueva Ecija to be the study site. Nueva Ecija is a major rice producing province in the Philippines and predominantly irrigated, which allowed us to capture farmers’ preferences for VTIs in both wet and dry seasons. Our sample consists of 122 rice producing households, with both husband and wife participating.

Sampling Approach

We used a multi-stage sampling approach to form our survey sample. In the first stage, we purposely selected three municipalities, where we sampled the participants: Muñoz, Talavera, and Guimba. In the second stage, we randomly selected four villages in each municipality. In the final stage, we randomly selected 10 households per village. Several steps were carried out in the random selection of the villages and rice producing households. First, we approached the Municipal Agriculture Office (MAO) in each of the municipality to obtain a master list of rice farming households. The master lists include information on the names of the farmers, their respective rice areas classified either under irrigated or rainfed. Second, we approached the local officials of the villages selected and asked them to check and verify the names included in the master list to determine who among in the list meet the criteria for participant selection. The selection criteria are as follows: (i) both husband and wife are involved in rice production and marketing activities; (ii) the household is planting rice on both wet and dry seasons; (iii) the household is selling a portion of their rice production. Once the list was verified and checked, a new list per village was created to include only those households that satisfy the selection criteria. We used a spread-sheet program to randomly select from these lists 10 households to be invited to participate in the experiment. We also randomly selected another set of 10 households to serve as the back-up list just in case those in the original list are unsure or will not be available on the schedule of the experiment.

Recruitment of Participants

The randomly selected households were invited through the designated local field coordinators in each of the selected villages. The local field coordinators are village official in- charge of the Agriculture Committee in their village. The households were invited to participate through a letter, which explains the details of the research, and the schedule of the experiment. The invitation letters were given two weeks before the scheduled experiment. Invited households were then reminded of the schedule two days before the actual experiment.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1101

Implementation and Procedures

The first few experimental sessions were held in Philippine Rice Research Institute (PhilRice) Training Hall in the municipality of Muñoz. Succeeding sessions were held in local village halls. Two kinds of venue set-up were used depending on the available facilities and resources in the villages. A classroom set-up was implemented for bigger halls that can accommodate at least 10 tables arranged vertically. For smaller halls, a drive-thru system was used. The explanation and presentation of the experiment were done in one area of the hall. Tables and chairs were set-up in another area of the hall for playing of the IGA. The experiment comprised of 12 sessions – one for each of the village selected. The sessions were conducted over the course of six days: one morning and one afternoon session on each day. Of the 12 sessions, three followed the control treatment, and three sessions for each of the three other information treatment. The assignment of the information treatment was randomly drawn prior to the start of the experimental sessions. Each session consisted of registration, introduction of the team and to the experiment, presentation and explanation of the IGA and VTIs, training on IGA, six consecutive rounds of IGA, short post experiment survey, and payment of returns. The sessions were conducted using the local language Filipino. A household survey questionnaire was also administered to gather data on socio-demographic, rice varieties grown, constraints in rice production and marketing, and marketing practices. The farmers were trained in the methodology of investing with budget constraints by using the “Training on Investment Game Application” (TIGA). In TIGA, farmers invested in their optimal dish by adding to a fixed amount of rice a vegetable or meat dish, using a budget amounting to PHP50 (Figure 2).The purpose of the training is for farmers to get familiarized with the application, particularly in terms of the budget constraint involved and the use of spin buttons (or up and down arrows) in the tablet. It is important that the participants be given the chance to use the tablet before the actual game as most of them are not familiar and have not used a tablet before. The IGA was played first by the husband and wife independently and simultaneously. Each of them was assigned an agent who facilitated the IGA and the post-experiment survey. The post-experiment survey was also answered independently by husbands and wives. This survey includes questions on the motivations behind their allocation decisions in IGA and a short quiz (two questions) to see how well the participants understood the experiment. After playing the IGA independently, husband and wife were asked to play the IGA jointly. One agent was assigned per couple for the joint round. To provide equal opportunity in answering the IGA during the joint round, husband and wife were given separate stylus pens and the tablet was placed in the middle of their table. Due to a limited number of agents who can facilitate the IGA, only two couples were playing the IGA at the same time. Other couples not yet doing the IGA were either approached for registration or for the household survey. This strategy helped in reducing the total time for the whole experiment and minimized the idle time of participants not yet doing the IGA.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1102

After all the households completed the six rounds of IGA, one participant was requested to roll the dice to randomly draw the binding round. Computation of the returns was based on experts’ estimate that the total cost invested in developing a new rice variety through rice breeding program can have a return of investment up to 10 times the total investment cost. Therefore, a maximum investment of PHP100 may earn around PHP1,000, depending on the chosen VTIs. The resulting cash returns were placed in an envelope and distributed to couples one at a time. A single-blind payment protocol was used where the research team knows the participants’ earnings but the participants did not know other participants’ earnings. On average, the participants earned PHP1,210 (around USD25), which is roughly equivalent to four daily wages for agricultural labor. 1 This return is on top of the fixed show-up fee of PHP250 (around USD5).

Estimation Strategy

Econometric Approach

We are interested in analyzing the relationship of the proportion of the endowment fund invested to trait improvements to various factors that may influence farmers’ allocation decisions. Estimation techniques such as ordinary least squares or general linear models are not applicable as these do not take into account the fractional nature of our dependent variables, that is, the investment allocations fall between zero and one (Musumba, Mjelde, & Adusumilli, 2015; Wollni & Fischer, 2015). Papke and Wooldridge (1996) proposed a quasi-likelihood method to estimate the conditional mean of the dependent variable in a single fractional response model. Their approach does not need to specify a particular distribution, and so no special adjustments or transformations are needed to the extreme values of zero and one. One possible distribution of the fractional observations is the dirichlet distribution, which is an extension of the beta distribution to multiple proportions. A limitation of the beta distribution, and consequently of the dirichlet distribution, is that it is difficult to justify if a considerable portion of the proportional or fractional observations has an extreme value of zero or one (Papke & Wooldridge, 1996). In our case, 65% of the observations on investment preferences for the individual trait improvements are zeros. Since we are interested in analyzing the 10 proportional shares corresponding to 10 VTIs, we used a fractional multinomial logit model, a multivariate generalization of the fractional logit model proposed by Papke and Wooldridge (1996). We let E as the total endowment fund and T as the investment allocation outcomes based on the 10 trait improvements and one outcome representing the uninvested portion. We then let 푦푖푘 = 푒푖푘/퐸, where yik is the proportional share for the kth investment allocation by the ith participant

(husband, wife, or joint) and k = 1, …,K, be the marginal outcomes of interest such that 푦푖푘 ∈

1 At the time of the experiment, the typical daily wage rate for agricultural labor in the province was around PHP334.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1103

퐾 (0,1) and ∑푘=1 푦푖푘 = 1. From here, two restrictions are enforced in the estimation. First is that 퐾 퐸[푦푖푘|푥푖] ∈ (0,1) for all i; and second is that ∑푘=1 퐸[푦푖푘|푥푖] = 1 for all i; xi represents all relevant explanatory variables (Mullahy & Robert, 2010).

The assumed model for investment allocations is 퐸[푦푖푘|푥푖] = 퐺푘(훽, 푥푖), where β are the parameters to be estimated. Keeping the investment allocations within the unit interval of zero and one is done by using a multinomial logit functional form for 퐺푘(훽, 푥푖), such that the conditional distribution of investment shares among v trait improvements is

(푥푖훽푘) 퐸[푦푖푘|푥푖] = 퐾 ∑푣=1 푒푥푝 (푥푖훽푣)

(푥푖훽푘) = 퐾 1 + ∑푣=2 푒푥푝 (푥푖훽푣) In estimating a fractional multinomial logit model using the 10 VTIs including the uninvested portion as the dependent variables, we used the command fmlogit (Buis, 2008) in STATA 13. In estimating the model, we use the proportion of the uninvested or the amount left from the endowment fund as the base outcome. Moreover, since each of the respondent (i.e. husband, wife, joint) provided their investment preferences for two cropping seasons, cluster- robust standard errors are calculated to account for the fact that standard errors across the two seasons for a respondent may be correlated.

Dependent Variables

Our dependent variables are the proportions representing farmer preferences for research budget allocation among the VTIs. Table 2 shows the overall average investment share of each of the trait improvement by season. The table shows that for wet season, farmers on the average allocated 22% of the endowment fund to improvement in lodging tolerance. They also allocated a relatively large share of the fund to improvements in disease and insect resistance, and abiotic stress tolerance. These traits with most of the allocations are traits that can potentially reduce crop losses. We find that respondents also allocated a major portion of the endowment fund to loss reducing traits during the dry season, with insect and disease resistance getting the highest proportions. However, we also see that there is an increase in the allocation to two grain quality traits: slenderness and head rice recovery. Table 3, on the other hand, shows gender-specific average investment share for each VTI. We find that more than 60% of the endowment fund was allocated by husbands and wives individually and jointly towards improvements in loss reducing traits. This is the case for both wet and dry seasons.

Independent Variables

The independent variables are factors that may have influence on farmers’ investment allocation decisions. Tables 4 & 5 show the descriptive statistics of the variables used for

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1104 pooled data analysis and gender-specific analysis, respectively. We include variables related to: (i) information treatment, (ii) characteristics of the model varieties, (iii) constraints in rice production experienced in the last two years, (iv) farm and farmer characteristics, (v) farmers’ behavior and attitudes, and (vi) rice production and marketing. These variables were tested for multicollinearity through the estimation of variance inflation factors (VIFs) and correlation coefficients. We did not find any major problems as the maximum VIF is 2.31 and maximum correlation is 0.71. The information treatment refers to the information provided to the participants during the experiment. About half of the participants received market information while around 52% received climate change information. We expect that respondents provided with market information will invest more in grain quality traits, while those who received climate change information will invest in loss reducing traits. For the model variety traits, we characterized the model varieties based on which season they are grown (wet or dry). We also characterized them based on the variety type (hybrid or inbred). Exactly half of the model varieties identified are for wet season and the other half for dry season since all respondents were able to answer the IGA for two cropping seasons. Hybrid varieties, which are more commonly grown in the dry season, comprise 37% of the model varieties and the remaining is inbreds. We expect that farmers are more likely to invest in loss reducing traits during the wet season, for which there is high risks of pests and diseases and more uncertainties in weather conditions (David, 2006). On the other hand, we expect farmers to invest more in grain quality traits if the model variety is hybrid as grain quality of earlier hybrid varieties were found to be inferior compared to inbred varieties (Casiwan et al., 2003). The third group of independent variables relates to the constraints in rice production experienced by the farmers. We focused on constraints related to abiotic stress, diseases, and grain shattering. We find that 34% of the sample households said they experienced abiotic stress, specifically submergence, in the last two years. On the other hand, 30% mentioned rice diseases as one of the constraints in rice production they experienced. Of the specific diseases, tungro and blast were cited the most by the respondents. These two are among the common diseases of rice in the Philippines (Laborte et al., 2015). Shattering of grains was also mentioned as another constraint by 43% of the households. Shattering of grains can reduce harvest due to shedding of mature grains from the panicle caused by birds, wind, rats, and handling (IRRI, 2016). We expect that farmers who experienced these problems will invest more in improving tolerances and resistances as well as reduction in shattering. We then include variables representing farm and farmer characteristics. Our respondents on the average are 49 years old, with wives younger at 47 years as compared to husbands who are 51 years old. Average years of schooling is eight for both husband and wife. The average annual family income is around PHP74,000 (around USD1,500). Overall, only about 33% have attended agricultural training in the past. Around 70% of the husbands have agricultural training, while this is only true for 17% of the wives. On average, the sample households have 1.30 hectares of farm land, which include own and leased area. Of the total

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1105 farm size, about 45% is leased. Previous studies on factors affecting adoption of improved varieties were used to guide the choice of farm and farm characteristics variables. Age can have both positive and negative effect. Older farmers have more experience in cultivation, thus they can better assess trait improvements they need. It could also be that older farmers are more risk averse and would rather save their endowment fund than invest it. We expect both education and attendance to training will have a positive effect. More educated farmers and those exposed to training can better process information and identify trait improvements they need. Moreover, we expect that those with higher income will be less risk averse and thus will invest their endowment fund to trait improvements, instead of keeping or saving them. We also expect land access to have a positive effect. We expect those with larger farm size to invest more as improved varieties could help increase the quantity and quality of the harvest, which could lead to higher incomes. On the other hand, we expect that the proportion of leased area would have both positive and negative effect. Positive such that leaseholders will invest in improvements to have higher production and higher incomes. But it could also have negative effect such that they rather save their endowment fund since in the case of most leaseholders, it is the land owners who decide which varieties to grow. We also include variables that relate to respondents’ behavior and attitudes. Overall, we find that 43% of the respondents considered their past and current farm experience in prioritizing trait improvements. However, we find a significant difference between husbands and wives. Results show that 60% of the husbands considered their past and current experience, while 57% of the wives based their decisions on future events. On the other hand, the joint variable shows that only 26% of the couples considered both their past and current experience. We also measured farmers’ time preference through a discount rate, which we estimated from a series of hypothetical questions relating to their preference of receiving a specific amount of cash now or a higher amount in a month’s time. On average, we find that the sample’s discount rate is 1.52%, with husbands having a higher discount rate at 1.63% as compared to wives’ 1.41%. Most of the respondents prefer to receive the cash amount immediately, instead of waiting for one month to receive a higher amount. This is consistent with Cardenas and Carpenter (2008) who argued that “people in less developed countries have high discount rates and are risk averse enough so that it is impossible for them to save and take risks necessary to begin to accumulate capital”. Respondents were also asked to assess their willingness to take risks in investing in rice farming on a Likert scale. On a scale of one to five, with five representing “extremely likely”, results show that the respondents are very much willing to take the risk in investing in rice farming with an average rating of 4.87. However, we find a significant difference between the rating of husband and wife. We find that on average, husbands rate their willingness to invest in rice farming at 4.94, which is significantly different from the wives’ rating of 4.80. Farmers mentioned that they are willing to take the risk in investing in rice farming since this is their main source of livelihood. It will then be interesting to see if this willingness will also translate to a willingness to invest in trait improvements given real money with real returns.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1106

We also include variables related to rice production and marketing. These are based on a specific variety planted per cropping season. We matched the model varieties identified in IGA with planted varieties provided through the household survey questionnaire. We find that only 74% of the model varieties matched with one of the varieties planted by the household in a particular season. We use the production and marketing information on the ‘matched’ planted variety for the said cases. As for the rest, we use available information on the specific variety planted per season as reported by the households. We find that 98% of total farm size was planted to one specific variety. This can be explained by the fact that most of our sample households only plant one variety per season. On average, proportion of the total production sold is 64%. The rest is for home consumption. The average price received from paddy selling is around PHP16 per kilogram while the average distance of the farm to the market is about 4.13 kilometers. In relation to marketing of their produce, 61% of the sample households said that their buyer/s require certain quality standards in terms of moisture content and cleanliness of their paddy. Lastly, for the pooled data analysis, we include a dummy variable representing husband as the respondent and another dummy variable for the wife as the respondent. To better capture gender differences in the preferences, we carry out the analysis using a gender-disaggregated data. For the gender-specific analysis, we only included a sub-set of the independent variables used in the pooled data analysis. We focus more on variables that have an individual response from husbands and wives.

Results and Discussion

Pooled Data Analysis

Results of the fractional multinomial regression of the pooled data set are shown in Tables 6 and 7. Table 6 shows the parameter estimates, while Table 7 shows the marginal effects. We focus the discussion here on variables with both parameter estimates and marginal effects significant. In terms of the information treatment, we find that market information is positively associated with investments in stickiness and aroma. Relative to those who did not receive market information during the experiment, farmers who received such information invested more by 2.71% in slenderness and 1.46 % in aroma. On the other hand, we find that climate change information is negatively related to investments in earliness. Compared to those who did not receive such information, farmers who received climate change information invested less in earliness by 1.38%. As for the model variety traits, wet season and hybrid varieties have positive effect on investments in lodging tolerance and earliness. Relative to dry season, farmers invested more in lodging tolerance by 18% and in earliness by 1.45% during the wet season. On the other hand, if the model variety is a hybrid, farmers invested more in lodging tolerance by 6.56% and in earliness by 1.94%.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1107

With respect to rice production constraints, results show that experience with abiotic stress has both positive and negative effects on investments in trait improvements. Specifically, farmers invested more by 6.52% in abiotic stress tolerance, but they invested less by 2.15% in stickiness. On the other hand, experience with grain shattering has a negative effect on investments in disease resistance and abiotic stress tolerance. Farmers invested less by 5.79% in disease resistance and by 3.80% in abiotic stress tolerance when they experienced grain shattering. Among the farm and farmer characteristics, farm size and percent leased area are significantly associated with investments in trait improvements. Results show that farm size has a positive effect on investments in lodging tolerance and disease resistance. Based on the results, a unit increase in farm size is associated with am increase in investments in lodging tolerance by 2.59% and by 1.53% in disease resistance. On the other hand, the proportion of leased area has a positive effect on investments in aroma and earliness. Specifically, an increase in the proportion of leased area is associated with increase in investments in aroma by 2.12% and in earliness by 2.08%. In terms of farmers’ attitudes, we find that the past and current experience have a positive relationship to allocation in aroma and lodging tolerance improvement. Farmers who considered their past and current experience in prioritizing trait improvements invested more in aroma by 3.28% and in lodging tolerance by 4.67%. On the other hand, farmers with higher discount rate invested more in stickiness, disease resistance, insect resistance, and earliness. An increase in discount rate is associated with an increase in investments by 0.28% in stickiness, 0.44% in disease resistance, 0.80% in insect resistance, and 0.23% in earliness. With respect to rice production and marketing variables, we find that the proportion of area planted to the model variety or any specific variety has both positive and negative effect on investments in trait improvements. We find that this has a positive effect on investments in lodging tolerance. An increase in the proportion of area planted leads to increase in investments in lodging tolerance by 19.31%. On the other hand, the proportion of area planted has negative effect on investments in stickiness, aroma, and head rice recovery. An increase in the proportion of area planted is associated with a decrease in investments by 6.85% in stickiness, 7.05% in aroma, and 11.58% in head rice recovery. These suggest that farmers were making trade-offs by investing less in quality attributes in favor of a loss reducing trait. Relative to the wives and the joint round, we find that husbands invested significantly less in insect resistance by 4.29%. On the other hand, relative to husbands and the joint round, we find that wives invested significantly less in insect resistance and abiotic stress tolerance by 5.46% and 3.46%, respectively. The results imply that farmers had the tendency to invest less and save a portion of their endowment fund during the individual rounds of the IGA, and maximize allocation of the endowment fund during the joint round.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1108

Gender-Specific Analysis

Tables 8 and 9 show the results of regression for husbands’ investment allocations. Analysis of husbands’ preferences for trait improvements show that climate change information has a negative effect on investments in earliness. Those that received climate change information invested 2.99% less in earliness. We also find that attendance to training has a negative effect on investments in abiotic stress tolerance. Husbands who attended training in the past invested less in abiotic stress tolerance by 5.32%. On the other hand, husbands’ past and current farm experience are positively related to their investment in slenderness. They invested more in slenderness by 5.89% when they based their investment decisions on their past and current experience. These findings indicate that husbands were influenced more by the experience they have acquired over the years in prioritizing variety traits for improvements, instead of using information acquired through formal means such as through their education, which we did not find significant in influencing allocation decisions. Husband’s age has a positive effect on investments in stickiness. An increase in the husband’s age is associated with 0.23% increase in investments in stickiness. Results also show that wet season has a positive effect on husbands’ investment in lodging tolerance. Husbands invested more by 15.52% in lodging tolerance during the wet season. Regression results for wives’ investment allocations are shown in Tables 10 and 11. We find that market information given during the experiment and attendance to training have a positive association with investments in grain quality improvement. According to the results, wives who received market information invested more by 2.76% and 4.10% in stickiness and aroma, respectively. They also invested more in slenderness by 4.26% if they have attended agricultural training in the past. These suggest that wives were able to make use of the information they acquired through these channels in prioritizing traits for improvements. Results also show that wet season has a positive effect on wives’ investments in lodging tolerance and earliness. Specifically, wives invested more by 15.86% in lodging tolerance and by 2.09% in earliness during the wet season. A hybrid model variety has also a positive effect on investment in earliness. Wives invested more by 2.78% in earliness when the model variety is a hybrid. Moreover, we find that wives’ past and current farming experience have a positive association with investments in aroma. Specifically, wives invested more by 4.61% in aroma if they have considered their past and current experience in prioritizing trait improvements. Tables 12 and 13 show the regression results for the joint decision. Results show that market information has a positive association with investments in slenderness. Specifically, couples who received market information invested more by 4.97% in slenderness. Results also show that wet season has both positive and negative effects. Couples invested less by 5.05% and 4.07% in slenderness and stickiness, respectively if the model variety is for wet season. On the other hand, wet season has a positive association with investments in lodging tolerance, with couples investing more by 24.56% for model varieties during the wet season. These results imply that the couples were making trade-offs in terms of which trait improvements to prioritize. Since there are more uncertainties in weather conditions during the wet season, they

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1109 prefer that a loss reducing trait be improved, rather than quality traits. Similarly, we find that couples were making trade-offs in trait improvements of hybrid varieties. Results show that a hybrid model variety has a negative association with investments in stickiness, with farmers investing less by 7.15%. On the other hand, hybrids have a positive effect in investments in lodging tolerance. Farmers invested more by 11.68% in lodging tolerance of hybrid varieties. Although there are fewer uncertainties in the weather conditions during the dry season when most of the hybrids are planted, couples still invested more in a loss reducing trait. This could be explained by the fact that hybrid seeds are more expensive as compared to inbreds, and so farmers place more importance on trait improvements that could lessen the risk of crop failure. In terms of farm and farmer characteristics, results show that the average age of the couple has a positive effect on investments in aroma, but a negative effect on investments in lodging tolerance, abiotic stress tolerance, and reduction in shattering. Specifically, results show that an increase in the average age of the couple is associated with a 0.25% increase in investments in aroma. On the other hand, an increase in the average age leads to a decrease in investments in lodging tolerance by 0.35%, in abiotic stress tolerance by 0.30%, and in reduction in shattering by 0.20%. These results suggest that couples were making trade-offs such that older couples are more interested in improvements in grain quality traits, which is probably for own consumption purposes, and less priority was given to loss reducing traits . With the proportion of leased area, we only find that it has a positive relationship with investments in aroma. An increase in percent leased area is associated with a 2.89% increase in investments in aroma. We also find that having higher discount rate affect positively joint investments in stickiness and earliness. Specifically, couples invested more by 0.54% and 0.45% in stickiness and earliness improvement, respectively. Moreover, education and past and current farm experience have negative effect on investment in abiotic stress tolerance. The more educated couples invested less by 1.21% in this trait. They also invested less by 7.45% in abiotic stress tolerance if they have both considered their past and current experience in prioritizing trait improvements.

Summary, Conclusions, and Recommendations

In this paper, we used data from a framed field experiment to examine farmers’ preferences for the allocation of breeding research budget among rice varietal trait improvements. Existing trait-based elicitation methods obtain preferences from variety traits per se. In this study, we focused on trait improvements, which farmers can identify and provide given the extent of their farming experience. Randomly selected rice farming households in Nueva Ecija, Philippines participated in the experiment, where they were asked to allocate a certain amount of endowment fund among 10 VTIs. Based on farmers’ allocation decisions, their investments generated real returns, which were given at the end of the experiment. By making the experiment incentive-compatible, we were able to motivate farmers to reveal their true preferences.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1110

Our findings show that farmers prefer to have the research budget allocated to improving loss reducing traits, rather than to quality traits. More than 60% of the endowment fund was allocated to lodging tolerance, disease and insect resistance, abiotic stress tolerance, and reduction in shattering. This is the case for both wet and dry seasons. We also find a similar trend for gender-specific investment allocations. Husbands and wives individually and jointly allocated more than 60% of their endowment fund to loss reducing traits, both for wet and dry seasons. These results suggest that the farmers are more concerned with ensuring the quantity of their production, and not so much of the quality. One possible explanation is that since most of them rely on rice farming for their livelihoods, higher production would mean higher income. Moreover, improvements in resistances and tolerances are critical since most of the model varieties as well farmers’ current varieties were released after 2005, where more than 80% of the varieties released during this period have no resistance to several insects and diseases (Laborte et al., 2015). We then used a fractional multinomial logit model to examine the factors that influence investment allocation decisions. We included factors that are related to the varieties, production constraints, farm and farmer characteristics, and personal values and attitudes. We carried out the analysis using pooled data and gender-specific data to see whether there are differences in the allocation decisions and the factors that condition these among husbands and wives individually and jointly. We find that market and climate change information, variety characteristics, access to land, and personal values and attitudes significantly affect farmers’ investment allocation decisions. Farmers responded to the market information provided during the experiment by investing more in several quality attributes. They also responded to climate change information by investing less in earliness. On the other hand, farmers invested more in loss reducing traits and earliness if the model variety is for wet season and if it is a hybrid. The results imply that although hybrids are commonly grown in the dry season, and inbreds are more widely used in the wet season, preferred trait improvements are similar across seasons and variety types. Access to land is also an important consideration in prioritizing trait improvements. Our results show that the larger the farm size and the larger the area devoted to the model variety or any specific rice variety, the more the farmers invested in loss reducing traits and less in grain quality traits. Farm size is one of the determinants of household income in rice growing villages in the Philippines (Estudillo, Sawada, & Otsuka, 2008; Takahashi & Otsuka, 2009). As such, growing rice varieties that have improved tolerances and resistances to different biotic and abiotic factors can ensure harvest and consequently guarantee farmers’ income. On the other hand, a higher proportion of leased area influenced the farmers to invest more in earliness. This could be due to the fact the growing early maturing varieties can be a means to escape severe weather conditions (Laborte et al., 2015), thus ensuring leaseholders’ share in the production. Although leaseholders are less likely to invest in productivity-enhancing activities (Abdulai, Owusu, & Goetz, 2011), early maturing varieties can provide them opportunity to plant high-value crops in between seasons or plant the next season rice earlier.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1111

Our results also show that there are gender differences in investment preferences for trait improvements, but there are also similarities. Both husband and wife displayed preferences toward improvements of loss reducing traits, which imply that both are involved in ensuring their production. However, we find that wives responded to the market information provided during the experiment by allocating a portion of the endowment fund to grain quality traits such as stickiness and aroma. This suggests that wives recognized the importance of having good quality grains not only for own consumption purposes but also to meet consumer demands. Overall, our results show that farmers were making trade-offs in terms of trait improvements to be prioritized. Investment preferences were mostly for loss reducing traits, but improvements of quality attributes should also be considered in breeding research since farmers, especially women, are likely to place importance on grain quality once they become aware of what the market or consumers demand. These trade-offs and preferences can then guide breeders and donors to make rice breeding programs more resource efficient and client- oriented, which could help facilitate the adoption of new and improved varieties. But beyond these, the novel approach of our methodology can transform the way preferences for variety traits are elicited and can provide a new opportunity for farmers to be truly involved in the agricultural research process through participation in resource allocation and priority setting. Our study, however, has a limitation with regard to its sample selection. This means that generalization of the findings to a larger population is also limited. Variety choice is site- specific and it is recommended that similar research is done to other major rice-growing areas to help in the development of rice variety product profiles that are better targeted. Preferences and trade-offs of other actors in the rice value chain can also be elicited using our methodology. Their preferences are also important and so it may be worthwhile to also engage them in this kind of research. Through this, it will be possible to have portfolios of trait improvements for rice breeding research that capture and integrate preferences across the rice value chain.

Acknowledgements

We acknowledge funding support from the Lee Foundation Rice Scholarship Program and the International Rice Research Institute (IRRI) in survey data collection. We are also grateful for the funding support from the Centre for Global Food and Resources (GFAR), The University of Adelaide. We want to thank the first author’s supervisory panel at GFAR – Wendy Umberger, Alexandra Peralta, and Risti Permani – for their helpful feedback. We also wish to acknowledge the assistance provided by Jhoanne Ynion and Donald Villanueva from IRRI and all six agents/enumerators from Nueva Ecija.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1112

References

Abdulai, A., Owusu, V., & Goetz, R. (2011). Land tenure differences and investment in land improvement measures: Theoretical and empirical analyses. Journal of Development Economics, 96(1), 66-78. doi: http://dx.doi.org/10.1016/j.jdeveco.2010.08.002 Adesina, A. A., & Baidu-Forson, J. (1995). Farmers' perceptions and adoption of new agricultural technology: evidence from analysis in Burkina Faso and Guinea, West Africa. Agricultural Economics, 13(1), 1-9. doi: http://dx.doi.org/10.1016/0169- 5150(95)01142-8 Adesina, A. A., & Zinnah, M. M. (1993). Technology characteristics, farmers' perceptions and adoption decisions: A Tobit model application in Sierra Leone. Agricultural Economics, 9(4), 297-311. doi: http://dx.doi.org/10.1016/0169-5150(93)90019-9 Arora, A., Bansal, S., & Ward, P. S. (2015). Eliciting farmers' valuation for abiotic stress- tolerant rice in India. Washington, DC: International Food Policy Research Institute. Baidu-Forson, J., Ntare, B. R., & Waliyar, F. (1997). Utilizing conjoint analysis to design modern crop varieties: Empirical example for groundnut in Niger. Agricultural Economics, 16(3), 219-226. doi: http://dx.doi.org/10.1016/S0169-5150(97)00009-1 Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, Reciprocity, and Social History. Games and Economic Behavior, 10(1), 122-142. doi: http://dx.doi.org/10.1006/game.1995.1027 Buis, M. L. (2008). FMLOGIT: Stata module fitting a fractional multinomial logit model by quasi maximum likelihood: Statistical Software Components, Boston College of Department of Economics. Retrieved from https://ideas.repec.org/c/boc/bocode/s456976.html Byrne, T. J., Amer, P. R., Fennessy, P. F., Hansen, P., & Wickham, B. W. (2012). A preference-based approach to deriving breeding objectives: applied to sheep breeding. animal, 6(Special Issue 05), 778-788. doi: doi:10.1017/S1751731111002060 Cardenas, J. C., & Carpenter, J. (2008). Behavioural Development Economics: Lessons from Field Labs in the Developing World. Journal of Development Studies, 44(3), 311-338. doi: http://www.tandfonline.com/loi/fjds20 Casiwan, C., B., Janaiah, A., Francisco, S., R., Hossain, M., Narciso, J., Cabrera, E., & Hidalgo, F., C. (2003). Hybrid Rice Cultivation in the Philippines: Early Farm-Level Experiences. Economic and Political Weekly, 38(25), 2502-2508. Dalton, T. J., Yesuf, M., & Muhammad, L. (2011). Demand for drought tolerance in Africa: selection of drought tolerant maize seed using framed field experiments. Paper presented at the 2011 AAEA & NAREA Joint Annual Meeting, July 24-26, 2011, Pittsburgh, Pennsylvania. David, C. C. (2006). The Philippine Hybrid Rice Program: A Case for Redesign and Scaling Down. Research Paper Series No. 2006-03. Makati, Philippines: Philippine Institute for Development Studies. Demont, M., Custodio, C., Villanueva, D., & Ynion, J. (2015). Investment Game Application (IGA): A new app for prioritizing rice farmers' preferences for varietal trait

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1113

improvements - Report on the IGA pre-test conducted in Banca-Banca, Victoria, Laguna, Philippines. Laguna, Philippines: International Rice Research Institute (IRRI). Estudillo, J. P., & Otsuka, K. (2006). Lessons from three decades of green revolution in the Philippines. The Developing Economies, 44(2), 123-148. doi: 10.1111/j.1746- 1049.2006.00010.x Estudillo, J. P., Sawada, Y., & Otsuka, K. (2008). Poverty and Income Dynamics in Philippine Villages, 1985–2004. Review of Development Economics, 12(4), 877-890. doi: 10.1111/j.1467-9361.2008.00458.x Evenson, R. E., Herdt, R. W., & Hossain, M. (1996). Priorities for Rice Research: Introduction. In R. E. Evenson, R. W. Herdt, & M. Hossain (Eds.), Rice Research in Asia: Progress and Priorities (pp. 3-15). Wallingford, UK: CAB International. Fisher, M., & Snapp, S. (2014). Smallholder farmers' perceptions of drought risk and adoption of modern maize in Southern Malawi. Experimental Agriculture, 50(04), 533-548. doi: doi:10.1017/S0014479714000027 Fox, G. (1987). Models of Resource Allocation in Public Agricultural Research: A Survey. Journal of Agricultural Economics, 38(3), 449-462. doi: 10.1111/j.1477- 9552.1987.tb01063.x Ghimire, R., Wen-chi, H., & Shrestha, R. B. (2015). Factors affecting adoption of improved rice varieties among rural farm households in Central Nepal. Rice Science, 22(1), 35- 43. Gollin, D. (2000). Setting Priorities for Agricultural Research: Theory and Experience. Paper presented at the Challenge of plant and agricultural sciences to the crisis of biosphere on the earth in the 21st century. Hellin, J., Bellon, M. R., Badstue, L., Dixon, J., & La Rovere, R. (2008). Increasing the Impacts of Participatory Research. Experimental Agriculture, 44(01), 81-95. doi: doi:10.1017/S0014479707005935 Hensher, D. A. (2010). Hypothetical bias, choice experiments and willingness to pay. Transportation Research Part B: Methodological, 44(6), 735-752. doi: http://dx.doi.org/10.1016/j.trb.2009.12.012 Herdt, R. W., & Capule, C. (1983). Adoption, Spread, and Production Impact of Modern Rice Varieties in Asia. Los Baños, Philippines: International Rice Research Institute. Hintze, L. H., Renkow, M., & Sain, G. (2003). Variety characteristics and maize adoption in Honduras. Agricultural Economics, 29(3), 307-317. doi: 10.1111/j.1574- 0862.2003.tb00167.x Horna, J. D., Smale, M., & Oppen, M. V. (2007). Farmer willingness to pay for seed-related information: rice varieties in Nigeria and Benin. Environment and Development Economics, 12(06), 799-825. doi: doi:10.1017/S1355770X07003956 IRRI. (2016). Rice Knowledge Bank. Retrieved August 16, 2016, from http://www.knowledgebank.irri.org

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1114

Joshi, G., & Bauer, S. (2006). Farmers' Choice of the Modern Rice Varieties in the Rainfed Ecosystem of Nepal. Journal of Agriculture and Rural Development in the Tropics and Subtropics, 107(2), 129-138. Kshirsagar, K. G., Pandey, S., & Bellon, M. R. (2002). Farmer Perceptions, Varietal Characteristics and Technology Adoption: A Rainfed Rice Village in Orissa. Economic and Political Weekly, 37(13), 1239-1246. doi: 10.2307/4411927 Laborte, A. G., Paguirigan, N. C., Moya, P. F., Nelson, A., Sparks, A. H., & Gregorio, G. B. (2015). Farmers’ Preference for Rice Traits: Insights from Farm Surveys in Central Luzon, Philippines, 1966-2012. PLoS ONE, 10(8), 1-18. doi: 10.1371/journal.pone.0136562 Launio, C., Redondo, G., Beltran, J., & Morooka, Y. (2008). Adoption and spatial diversity of later generation modern rice varieties in the Philippines. Agronomy journal, 100(5), 1380-1389. List, J. A., & Gallet, C. A. (2001). What Experimental Protocol Influence Disparities Between Actual and Hypothetical Stated Values? Environmental and Resource Economics, 20(3), 241-254. doi: 10.1023/a:1012791822804 Little, J., & Berrens, R. (2004). Explaining Disparities between Actual and Hypothetical Stated Values: Further Investigation Using Meta-Analysis. Economics Bulletin, 3(6), 1-13. Lunduka, R., Fisher, M., & Snapp, S. (2012). Could farmer interest in a diversity of seed attributes explain adoption plateaus for modern maize varieties in Malawi? Food Policy, 37(5), 504-510. doi: http://dx.doi.org/10.1016/j.foodpol.2012.05.001 Mariano, M. J., Villano, R., & Fleming, E. (2012). Factors influencing farmers’ adoption of modern rice technologies and good management practices in the Philippines. Agricultural Systems, 110, 41-53. doi: http://dx.doi.org/10.1016/j.agsy.2012.03.010 Martin-Collado, D., Byrne, T. J., Amer, P. R., Santos, B. F. S., Axford, M., & Pryce, J. E. (2015). Analyzing the heterogeneity of farmers’ preferences for improvements in dairy cow traits using farmer typologies. Journal of Dairy Science, 98(6), 4148-4161. doi: http://dx.doi.org/10.3168/jds.2014-9194 Merino-Castello, A. (2003). Eliciting consumers preferences using stated preference discrete choice models: contingent ranking versus choice experiment. UPF economics and business working paper(705). Mullahy, J., & Robert, S. A. (2010). No time to lose: time constraints and physical activity in the production of health. Review of Economics of the Household, 8(4), 409-432. doi: 10.1007/s11150-010-9091-4 Murphy, J. J., Allen, P. G., Stevens, T. H., & Weatherhead, D. (2005). A Meta-analysis of Hypothetical Bias in Stated Preference Valuation. Environmental and Resource Economics, 30(3), 313-325. doi: 10.1007/s10640-004-3332-z Musumba, M., Mjelde, J. W., & Adusumilli, N. C. (2015). Remittance receipts and allocation: a study of three African countries. Applied Economics, 47(59), 6375-6389. doi: 10.1080/00036846.2015.1071472

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1115

Papke, L. E., & Wooldridge, J. M. (1996). Econometric methods for fractional response variables with an application to 401(k) plan participation rates. Journal of Applied Econometrics, 11(6), 619-632. doi: 10.1002/(SICI)1099- 1255(199611)11:6<619::AID-JAE418>3.0.CO;2-1 Paris, T. R., Singh, A., Luis, J., Hossain, M., Singh, H. N., Singh, S., & Singh, O. N. (2001). Incorporating gender concerns in participatory rice plant breeding and varietal selection: preliminary results from eastern India. In N. Lilja, J. A. Ashby, & L. Sperling (Eds.), Assesing the impact of participatory research and gender analysis (pp. 109-121). Cali, Colombia: International Center for Tropical Agriculture. Pingali, P. L., Rozelle, S. D., & Gerpacio, R. V. (2001). The Farmer's Voice in Priority Setting: A Cross-Country Experiment in Eliciting Technological Preferences. Economic Development and Cultural Change, 49(3), 591-609. doi: http://www.jstor.org/action/showPublication?journalCode=econdevecultchan Sall, S., Norman, D., & Featherstone, A. M. (2000). Quantitative assessment of improved rice variety adoption: the farmer’s perspective. Agricultural Systems, 66(2), 129-144. doi: http://dx.doi.org/10.1016/S0308-521X(00)00040-8 Smale, M., Bellon, M. R., & Aguirre Gomez, J. A. (2001). Maize Diversity, Variety Attributes, and Farmers' Choices in Southeastern Guanajuato, Mexico. Economic Development and Cultural Change, 50(1), 201-225. doi: http://www.jstor.org/action/showPublication?journalCode=econdevecultchan Takahashi, K., & Otsuka, K. (2009). The increasing importance of nonfarm income and the changing use of labor and capital in rice farming: the case of Central Luzon, 1979– 2003. Agricultural Economics, 40(2), 231-242. doi: 10.1111/j.1574- 0862.2009.00372.x Villano, R., Bravo-Ureta, B., Solís, D., & Fleming, E. (2015). Modern Rice Technologies and Productivity in the Philippines: Disentangling Technology from Managerial Gaps. Journal of Agricultural Economics, 66(1), 129-154. doi: 10.1111/1477-9552.12081 Ward, P. S., Ortega, D. L., Spielman, D. J., & Singh, V. (2014). Heterogeneous Demand for Drought-Tolerant Rice: Evidence from Bihar, India. World Development, 64(0), 125- 139. doi: http://dx.doi.org/10.1016/j.worlddev.2014.05.017 Wollni, M., & Fischer, E. (2015). Member deliveries in collective marketing relationships: evidence from coffee cooperatives in Costa Rica. European Review of Agricultural Economics, 42(2), 287-314. doi: 10.1093/erae/jbu023

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1116

Tables and Figures

Table 1 Traits and trait-specific metrics on which the IGA is calibrated Trait Metric Baseline Target Slenderness Length/width ratio 2.4 3.2 Stickiness Amylose content (%) 27% 22% Aroma Price premium (%) (market benchmark = 100%) 0% 100% Head rice recovery % head rice obtained from a sample of paddy 45% 60% Lodging tolerance Crop losses eliminated (%) 20% 80% Disease resistance Crop losses eliminated (%) 50% 90% Insect resistance Crop losses eliminated (%) 80% 95% Abiotic stress tolerance Crop losses eliminated (%) 0% 90% Reduction in shattering Crop losses eliminated (%) 80% 95% Earliness Number of days the duration is shortened 0% 14 Source: Demont, Custodio, Villanueva, & Ynion, 2015

Table 2 Average investment shares per trait improvement and by season Wet season Dry season Trait Mean Std. Dev. Min Max Mean Std. Dev. Min Max Slenderness 0.07 0.15 0.0 0.70 0.10 0.18 0.0 0.81 Stickiness 0.04 0.11 0.0 0.55 0.03 0.10 0.0 0.55 Aroma 0.04 0.11 0.0 0.54 0.05 0.13 0.0 0.56 Head rice recovery 0.05 0.13 0.0 0.65 0.10 0.18 0.0 1.00 Lodging tolerance 0.22 0.22 0.0 1.00 0.09 0.17 0.0 1.00 Disease resistance 0.17 0.21 0.0 0.70 0.16 0.21 0.0 1.00 Insect resistance 0.16 0.20 0.0 1.00 0.18 0.21 0.0 1.00 Abiotic stress tolerance 0.12 0.19 0.0 0.80 0.12 0.21 0.0 1.00 Reduction in shattering 0.09 0.12 0.0 0.53 0.12 0.15 0.0 0.54 Earliness 0.04 0.09 0.0 0.48 0.03 0.09 0.0 0.50 Uninvested 0.01 0.02 0.0 0.24 0.01 0.03 0.0 0.39 Observations 366 366

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1117

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1118

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1119

Table 5 Descriptive statistics for the independent variables used in the gender-specific analysis Husband Wife Joint Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Market information 0.49 0.50 0.49 0.50 0.49 0.50 Climate change information 0.52 0.50 0.52 0.50 0.52 0.50 Wet season 0.50 0.50 0.50 0.50 0.50 0.50 Hybrid 0.37 0.48 0.35 0.48 0.40 0.49 Age 50.71 10.56 47.66 10.73 49.47 10.35 Education 8.42 2.60 8.16 2.38 8.39 2.03 Attendance to training 0.70 0.46 0.17 0.38 0.13 0.34 Past & current experience 0.60 0.49 0.43 0.50 0.26 0.44 Time preference 1.63 4.66 1.41 1.99 1.52 2.62 Percent leased area 0.45 0.49 0.45 0.49 0.45 0.49 Observations 244 244 244 Note: Variable definitions are the same with that for the pooled data analysis (Table 4) except for the joint definition of: (i) Age: average age of husband and wife; (ii) Education: average years in school of husband and wife; (iii) Attendance to training: 1 – if both husband and wife attended training in the past, 0 – otherwise; and (iv) Past and current experience: 1 – if both husband and wife selected past/current experience, 0 – otherwise; (v) Time preference: average of the discount rate of husband and wife.

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1120

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1121

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1122

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1123

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1124

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1125

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1126

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1127

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1128

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1129

The 9th ASAE International Conference: Transformation in agricultural and food economy in Asia 11-13 January 2017 Bangkok, Thailand 1130

Figure 1 Investment Game Application (IGA) with example allocations in stickiness, lodging tolerance, and earliness. The blue horizontal bar at the bottom shows the status of the endowment fund, while the pie charts below the VTI bars indicate the riskiness of each investment – green segments represent the probability that the target VTI will be achieved; the red segments represent the odds of achieving a random VTI somewhere between zero and the target VTI.

Figure 2 Training on Investment Game Application (TIGA) with example allocation of PHP15 to Pakbet (vegetable dish) and PHP20 to adobo (meat dish). The blue horizontal bar at the bottom shows the remaining budget.