THE RELATIONSHIP BETWEEN MARKET INFORMATION AND FARM - GATE PRICES RECEIVED BY SMALL - SCALE FARMERS IN THE MAGWAY REGION OF MYANMAR By Soojung Ahn A THESIS Submitted to Michigan State University in partial fulfillment of the requirement s for the degree of Agriculture, Food, and Resource Economics Master of Science 2019 ABSTRACT THE RELATIONSHIP BETWEEN MARKET INFORMATION AND FARM - GATE PRICES RECEIVED BY SMALL - SCALE FARMERS IN THE MAGWAY REGION OF MYANMAR By Soojung Ahn This pa per focuses on the relationship between market inform ation and the level rice pric es received by small - scale farmers in the Magway region of Myanmar . Using household - level d ata , we estimate this relationship based on a search cost model . The study extends previous literature by distinguishing between price information and other type s of market information. T he relationship between dif ferent information types and farm - gate prices received are estimated econometrically. We find a positive relationship between non - price information received from TV sources and the rice price re ceived by far mers. However, other sources of information, including price information , show no statistically significant relationship with rice prices received by farmers . The results ind icate there is a relationship between access to TV information and higher rice prices received, but we were not able to determine whether the information allows farmers to gain higher prices or farmers that get higher prices were able to access TV services bette r. (i.e., the direction of causality could not be determined). iii A CKNOWLEDGEMENTS I would like to thank the faculty members who have contributed to making this study possible. I would like to thank my advisor, Dr. Robert Myers, for h is guidance an d support throughout the second year, and my committee members, Dr. Nicole Mason and Dr. Eduardo Nakasone for their insightful comments and advice on this paper. Furthermore, I would like to extend my gratitude to Nyein Nyein Kyaw who kindly provided me he r survey data from Myanmar. In addition, I would like to thank the many faculty members in the Department of Agricultural, Food, and Resource Economics , as well as Economics , at Michigan State University from whom I learned so much , Dr. Robert Myers, Dr. Vincenzina Caputo, Dr. Nicole Mason, Dr. Saweda Liverpool - Tasie, Dr. Eduardo Nakasone. Finally, I would like to thank God and my parents for supporting me and gi ving me all these opportunities . iv TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ......................... LIST OF FIGURES ................................ ................................ ................................ ....................... ................................ ................................ ................................ ...................... 1 ................................ ................................ ................................ ................................ ........ 8 A. The Study Area ................................ ................................ ................................ .................... 8 B. Data and Data Collection ................................ ................................ ................................ ..... 9 . EMPIRICAL METHODS ................................ ................................ ................................ ...... 15 ................................ ................................ ................................ .............................. 18 ................................ ................................ ................................ ..................... 23 APPENDICES ................................ ................................ ................................ ............................. 25 APPENDIX A: Tables ................................ ................................ ................................ ............. 26 APPENDIX B: Figures ................................ ................................ ................................ ............ 29 REFERENCES ................................ ................................ ................................ ............................ 3 3 v LIST OF TABLES Table 1 . Agricultural O verview of the V illages ................................ ................................ ........... 1 0 Table 2 . Summary Statistics for Survey Questionnaire ................................ ............................... 1 1 Table 3 . OLS E stimation R esults for the D eterminants of M onsoon P rices ............................... 19 Table A.1 . Correlation Mat rix of I ndependent V ariables ................................ ............................. 26 Table A.2 . Outcome of Variation Inflation Factors (VIF) ................................ ........................... 28 vi LIST OF FIGURES Figure 1 . Comparison of the National Farm - Gate Monsoon Rice Prices (2 013/14 ) ..................... 5 Figure 2 . Number of Farms by Price Variations of Mons oon Rice in the H ousehold S urvey ...... 6 Figure B.1 . GDP Shares of Agricultural and Non - agricultural Sector ................................ ........ 29 Figure B.2 . Population Ratio Between Rural and Urban Areas ................................ .................. 3 0 Figure B.3 . Employment Rate of Total Employment ................................ ................................ .. 3 0 Figure B.4 . Rice Map of My anmar (Total Sown Area) ................................ ............................... 3 1 Figure B.5 . Sown Acreage of Rice Cultivation Area in Myanmar ................................ .............. 3 2 1 . INTRODUCTION A griculture plays a vital economic role in most less developed countries and a large portion of agricultural output is produced by small - scale farmers. Indeed, over 60% of global agricultural output is being produced by small - scale farmers (Poole, 2017). No t only does small - scale agriculture contribute to economic output but it is also a major source of food consumption for developing countries, particularly in sub - Saharan Africa and Asia. Hence, small - scale agriculture is a vital source of income, sustenanc e, and a driving force for economic development in less developed countries. Agriculture is a knowledge - intensiv e industr y (Hall, 2011). Farming requires information on what inputs to use, how much of the inputs to apply, when and where t o sell farm output , and so on . Above all, f armers make decisions for their farming activities based , at least in part, on the information they obtain. Hence, access to business and market information might have a close relationship with agr icultural develo pment. D elivering useful information to small - scale farmers could promote agricultural and economic development as well as food security in less developed countries. Nevertheless , access to information is limited in many of these countries since a majority of the smallholders are illiterate , or not proficient in business transactions . Moreover, their lack of knowledge and information may sometimes be exploited by brokers and traders who are better informed about market conditions 1 (European Commission, 2012 ) . In this situation, information asymmetry occurs as more informed agents bargain with less informed farmers (Magesa, Michael and Ko, 2014; Svensson and Yanagizawa, 2008) . 1 Small - scale farmers in developing countries are vulnerable in terms of bargaining for prices as the y can hardly refuse price offered from their traders. 2 To narrow or eliminate the information gap between farmer s and buyer s , the Food an d Agriculture Organization (FAO) and other organizations have been develop ing Market Information S ystems (MIS) for many years. This movement was initiated at the time of market liberalization in many developing countries, especially those that ha d been und ergoing structural transformation. It was believed that improved information would help farmers to plan their production based on market demand, make decisions on where and when to sell, and negotiate more equally with traders . According to Elly and Silay o ( 2013 ) , information might improve the bargaining position of farmers with traders, reduce transaction costs, and provide additional market opportunities for farmers and small traders , especially in those countries where market s are being liberalized. Egg leston , Jensen, and Zeckhauser (200 2 ) stated that m arkets in developing countries, especially rural areas, are ofte n inefficient because information flows sluggishly. As a result, f armers produce a different mixture of crops, or they use inefficient techno logies, so consumers get products that do not meet their needs. This paper investigates the relationship between access to information and the level of output prices received by smallholder farmers in developing countries. There has been previous research on this issue. For example, Salasya and Burge r (2010) identified why there are large variations in the price offered at different market outlets for kale in Kenya, and used hedonic analysis to investigate what characteristics of kale influence the price. R esults indicated that kale p rice variations were expla ined primarily by the location of production and distance to a local market. English ( 2008) examined the factors explaining farm gate price s received for cocoa among Liberian farmers using Ordinary Leas t Squares (OLS) regression. She found that the world p rice of cocoa , transport costs and distance , access to mark et information and resources were statistically 3 significant , while farmer characteristics were found to have little connection to the level of farmgate prices. Svensson and Yanagizawa (2008) assess ed the relationship between improved access to market information and farm - gate maize prices by comparing two regions with and without access to radio, as well as the districts affected and not affected by a n MIS project in Uganda . They used the difference - in - difference approach. They found evidence that better - informed farmers managed to bargain for higher farm - gate prices for their products. Lee and Bellemare (2013) focused on whether use of a mobile p hone is connected to the lev el of onion prices received by farmers in the Philippines using two OLS regressions . First, they examined if mobile phone ownership at the household level is correlated with price levels. Second, they estimate d the relationship between mobile phone ownership and prices after controlling for the intra - household distribution of mobile phones . I t was observed that about 47% of the total household s owned cell phones and the proportions of mobile phone ownership were 27.6% (farmer), 1 3.4% ( s pouse) and 11.2% (children) respectively. The r e sults indicated that household ow nership of a mobile phone does not correlate with onion price received . However, when the farmer or spouse owns a mobile phone, onion prices received are 5 to 8 per cen t highe r . This implies that the information obtained by mobile phones may not be shared within the household, s . Tadesse and Bahiigwa (2015) also investigated whether mobile phone use is a ssociated with the level of barley and wheat prices received by farmers in rural Ethiopia. R esults showed that mobile phone ownership generally has no statistically significant relationship with the pr ice received. They mentioned three possibilities for th e insignificant result: (1) farmers may not engage in spatial arbitrage; (2) farmers may not use t heir mobile phone for accessing market 4 information; and/or (3) farmers may not obtain useful informatio n through mobile phones. Aker and Fafchamps (2010) also measured the impact of mobile phone coverage on millet and cowpea price s received by farmers in Niger , but using panel data . They used market pair fixed effects and found that price variation of cowpe a has decreased by 6 percent with higher mobile phone c overage. A common problem in much of the previous research (and the research reported in this paper) is distinguishing between correlation and causality (Singleton and Straits, 1999; Handy, Cao, and Mokhtarian, 2005). Although the current paper does not s olve the causality issue, it does build on previous literature by studying multiple sources and types of market infor mation in one study . Previous research has focused on one or at most two aggregate information types. Here we divide market informa tion int o price and other information but then each information type is further decomposed into its source (e.g. broker versus TV). This leads to a rich classification of information types that goes beyond mo st existing studies. Furthermore , we control for both ho usehold ( farm ) characteristics and product quality attributes in our analysis. While many previous studies have controlled for one or the other of these , we include both which may lead to more robust results. In our empirical analysis w e focus on the case of rice in Myanmar, which is one of the two lowest income countries in Asia and heavily dependent on agri culture. In Myanmar, rice is the most important agricultural product in terms of exports and do mestic consumption and is produced primarily by smallhol der farmers . As shown in Figure B.4 , rice is produced in a lmost all regions in Myanmar with Ayeyarwady being the main rice - producing area. 5 According to the World Bank (2016), Myanmar has the lowest farm - gate rice prices for monsoon rice among any of its n eighbors (Figure 1 ). The large gap between Thai rice prices and most oth er countries is caused by T rice pledging scheme, a subsidy policy through which farmers sell rice to the government at a particular price, which is about 50% higher than the market price (Poapongsakorn and Pantakua, 2014). But Myanmar still has the lowe st farm - gate prices in its area, which is typically explained by poor harvest quality, low volume s, high costs of milling, transport, and export , etc. (World Bank, 2016). As mos t rice farmers in Myanmar are small - scale farmer s, marketable rice usually has high moisture content and many impurities, with multiple varieties used on a given farm. This makes it hard to get large volumes of uniform variety. 6 More importantly from the perspective of the current paper, there is also price variabilit y across regions and across farms within regions in Myanmar. The World Bank (2016) indicated that the average monsoon rice price in Shan State was $340 per ton while the price in Ayeyarwady , w hich is the main rice - producing area , was $200 per ton. Farm - gat e rice prices vary across the farmers within a region and even within a village. In a household survey conducted in Myothit Township of the Magway region, which is one of the rural areas in My anmar, price variation from 5,000 Kyats/basket ($157/ton) to 7,5 00 Kyats/baskets ($236/ton) 2 is observed in monsoon rice farm data (Figure 2 ). The reasons for this farm - level price variation are not immediately clear. Spatial differen ces and transportation costs may explain some of the differences but the results in this paper show that there are many other factors at work also. It is important to know the sources o f th ese price variations so that farmer actions and government policies can be recommended or designed to help lagging farms receive higher prices. 2 In Myanmar, 1 basket equals to 20.68kg and Myanmar currency is calculated into USD as of the exchange rate on March 18, 2019. The number of cent digits was rounded. Note 1. Per basket = 20.68kg 2. 5000 Kyat = 3.25 USD as of the exchange rate on March 18, 2019. 7 The remainder of the paper is organized as follows. In the next section we give a brief explanation of the data used in this study. The third section outlines the empirical methods used to in the anal ysis . In t he fourth section, we present the results and discuss the main findings of the paper. The last section states the major findings and policy implications. 8 . D ATA A. The S tudy A rea This study is based on a rural household survey of small - scale rice farmers . The survey was carried out in in the Magway region 3 of Myanmar, in the central dry zone. The Magway region is located west of the capital city of Myanmar , Naypyidaw (Nay Pyi Taw ) and consists of 25 townships within 5 districts . Among the townships, the survey was conducted in Myothit township, where a population of 157,544 (as of 2016) is located and m ost of the households are engaged in agriculture , mai nly rice . Myanmar s new ye ar begins in April and ends in March. The timing of rice cultivation depends on the location of the production area and whether irrigation facilities are used. In Myanmar, both monsoon rice , which is rain - fed , and summer rice , whi ch uses irrigated water , are produced. In the survey area , most of the rice farmers are cultivating monsoon rice while only a few farmers grow summer rice. Thus, we focus on monsoon rice. In the Magway region, monsoon rice is cultivated from July to Novemb er, normally for 4 months , then summer rice is produced (USDA, 2018). After monsoon rice production, most farmers in this region plant peanut, corn, green gram, sunflower, sesame , etc. Myothit t ownship is located 37 miles away from Magway , the capital of Magway r egio n where market is located. Brokers who obtain agricultural produce from farmers in Myothit township sell to the Magway market. In Myothit township, there are 47 village tracts and 169 villages. The annual temperature range s from 5 6 8 rainfall is 81 2 mm . Most studies on rice production and marketing in Myanmar have focused on the 3 Magway region is also called as M agwe. 9 Ayeyarwady region, which is the largest production area (Figure B.5 ). Howe ver, the current st udy investigate s rice pr ice variations in areas where rice production is less concentrated and price variation across households is likely to be higher. B. Data and Data C ollection The survey was primarily conducted for a personal study on marketing activ ities of smallholder far mers in Myothit Township 4 . The survey contains household information, income, finance, production information, market participation , extension, market information, infrastructures and production constraints. It was co llected by five enumerators from March to April in 2018. Among the 169 villages within Myothit Tow nship, three villages ( Ko Yan Taung, Pha Lan Tine, and Nyaung Kaing ) were chosen for the survey. These three villages contain the most rice growers in Myothit Township. Summary data on the three vi llages is shown in Table 1 . Among the three villages, Ko Yan Taung has the largest population and the most agricultural land. All three villages produc e mainly rice but also cotton and sesame. Most marketed ric e goes to Myothit Township market through vari ous channels , including broker s and miller s , and among the three villages Nyaung Kaing is closest to the Township market . 4 This paper uses data collected and used for Kyaw, Ahn and Lee (2018) . Thanks to them for kindly sharing their data. 10 Unit Ko Yan Taung Pha Lan Tine Nyaung Kaing Total population No. of people 977 769 364 Male No. of people 478 380 184 Female N o. of people 499 389 180 Total agricultural land Acre 2,740 446 1,271 Lowland Acre 1,510 360 1,118 Upland Acre 1,230 86 153 Paddy land Acre 1,510 360 935 Agricultural crops Paddy, Cotton, Sesame, Chickpea, Green gram Paddy, Cotton, Sesame, Green gram , Chili, Tomato Paddy, Cotton, Chickpea, Sesame, Groundnut Area of village (farmer's house) Square feet 2,526,480 3,136,320 114,022,656 Distance to township (market) Kilometers 19.75 19.35 12.9 O ne hundred and f ifty small - scale farmers were randomly selected from the three villages and household interviews were conducted by five enumerators during March through April of 2018. In the sample, 46% of th e farmers are fr om Ko Yan Taung village, 27% are from Pha Lan Tine , and 27% from Nyaung Kaing. The survey questionnaires were mainly on rice production and sales after rice production in 2017 . 5 Among the 1 50 households, 1 18 farmers were found to sell some of their rice an d so could respond to the question about rice price received . T herefore we focu s on the 118 farmers with price data . The household survey also contains household and farm characteristics, information on crop cultivation methods, availability of market, inf rastructure, etc. Descriptive statistics on key survey questions are shown in Table 2 . The average price received for monsoon rice sold during December 2017 to March 2018 varies from 5,000 kyats to 7,500 kya ts per basket . 6 , 7 Farmer characteristics such as age, gender, education level, farming experience, 5 As mentioned above, monsoon rice is cultivated during July to November, so the data are the average figures of December 2017 to M a rch 2018. 6 5000 Kyat = 3.25 USD as of the exchange rate on March 18, 2019. 7 1 basket = 20.68 kg . 11 household size, and access to extension services could all potentially be related to the variation in prices received. Financial characteristics such as household income, landholdings, household consumption levels, and access credit may also be relevant fa ctors. Variable Obs Mean Std. dev. Min Max Explanation Dependent variable Price 118 6,654.24 536.15 5,000 7,500 Average price of monsoon rice lprice 118 8.80 0.08 8.52 8.92 Log (Price) Explanator y variables Farmer Characteristics Age 118 51.82 10.39 25 74 Gender 118 0.93 0.25 0 1 =1 if the farmer is male Education 118 5.75 2.99 0 15 Experience 118 28.38 10.73 5 55 Farming e xperience HH_size 118 4.98 1.58 1 9 Number of individuals in the household HH_income 118 4,793 5,611 294 54,100 Household income in 1,000 kyats lincome 118 15.06 0.79 12.59 17.81 Log (HH_income) Landhold 118 7.15 4.90 0 30 Household landholdings in acr e Consumption 118 90.37 56.44 0 300 Household consumption of rice Credits 118 0.83 0.38 0 1 Access to credits Extension 118 0.85 0.36 0 1 Access to extension services Farm characteristics Cultivated 118 4.56 3.21 0 25 Cultivated area for monsoon rice Variety_1 118 0.59 0.49 0 1 Ayeyarmin Variety_2 118 0.25 0.44 0 1 Ayearpadathar Variety_3 118 0.03 0.16 0 1 Manawhtikesa Variety_4 118 0.03 0.16 0 1 Htikesa Variety_5 118 0.01 0.09 0 1 Kayinma Variety_6 118 0.01 0.09 0 1 Shwe Bo Quality of rice Mil l 118 0.08 0.27 0 1 Rice milled before selling Store 118 0.42 0.50 0 1 Rice stored before selling Plan 118 0.51 0.50 0 1 Plan to sell Buyer Sell_consumer 118 0.02 0.13 0 1 Sold to consumer Sell_broker 118 0.96 0.20 0 1 Sold to broker Sell_miller 118 0.03 0.16 0 1 Sold to miller Place of sale Place_farm 118 0.05 0.22 0 1 Sold at farm Place_house 118 0.89 0.31 0 1 Sold at house Place_town 118 0.02 0.13 0 1 Sold at town Place_road 118 0.01 0.09 0 1 Sold at road Place_other 118 0.02 0.13 0 1 Sold at other places 12 Table 2 (cont d) M ost farmers in the s ample are male and their mean age is around 5 2. The average year s of education is about 6 indicating that most farmers are at an elementary school level. Their average years of farming experience is 28 and the mean household size is 5 members. Household in come varies from 294,000 kyats to 54.1 million kyats 8 and the area of landholdings ranges from zero to 30 acres. The average household consumption of rice is around 90 baskets 9 per household and most farmers are using extension and loan services. The mean cultivated area is 4. 56 acres. Among the six rice varieties, Ayeyarmin and Ayearpadathar are the most cultivated and about 60% of farmers grew Ayeyarmin. The rice price may be influenced by quality, so it is important to account for whether the rice was milled or stored and sold later. The data show that most rice sold is not milled and , on average , 42% of production is stored for at least some time and 51% is planned for sale versus own consumption. M arket - 8 USD 177,870 to USD 32.7 million as of the exchange rate on March 18, 2019. 9 1 basket = 20.68 kg, 90 baskets = 1.86 tons Transport Road 118 0.91 0.29 0 1 Road condition Village Village_1 118 0.47 0.50 0 1 Ko Yan Taun Village_2 118 0.29 0.45 0 1 Pha Lan Tine Village_3 118 0.25 0.43 0 1 Nyaung Kaing M arket information Info_price 118 0.93 0.25 0 1 Access to price information Neighbor_price 118 0.29 0.45 0 1 Price i nformation from neighbors Broker_price 118 0.72 0.45 0 1 Price i nformation from brokers Miller_price 118 0.05 0.22 0 1 Price i nformation from rice millers DOA_price 118 0.01 0.09 0 1 Price i nformation from governmental staff Media_price 118 0.21 0.41 0 1 Price i nformation from media Info_nprice 118 0.81 0.40 0 1 Access to non - price information Radio_nprice 118 0.15 0.36 0 1 Non - price in formation from radio Mobile_nprice 118 0.22 0.42 0 1 Non - price information from mobile TV_ nprice 118 0.17 0.38 0 1 Non - price information from TV Broker_nprice 118 0.53 0.50 0 1 Non - price information from broker 13 related variables such as the type of buyer, pla ce of sale, and transport method are also potentially relevant. In this region, there a re three types of buyers ( consumer s , broker s, and miller s) . Broker s (or middlemen) ar e small, village - based rice traders who own a small store or resell rice to a bigger market. Some farmers directly sell their rice to consumers, or to rice millers (processors) who polish and pack rice. Among different buyer types, brokers are overwhelmin gly the most common outlet ( 7 7 % of respondents sell their rice to brokers ). Other re spondents sell to miller s (2%) and directly to consumers (1%). For place of sale, the , followed by farm, road, town and other places. In addition, most of the respondents indicated that road conditions in the area are good. In this study our main interest is in the role of market information and we classify information as price and non - price . Price information relates to prices available in the market place. N on - price information includes who and where the buyers are, how they can be contacted, what their preferences for varieties , etc. Statistics show that about 93 % of the farmers have access to price information. However, farme rs obtain price information through different channels. The most frequent source of price i nformation is broker s with 72% of the farmers obtaining price information this way . Some farmers (30% of the sample) indicated they get price information from multip le sources. Most farmers (98%) received some non - price information but the source varies. A bout 50% of farmers get information from brokers. The next most frequent source was mobile phone (22%) followed by TV (17%) and Radio (15%). One possible problem in this study is that there can be measurement errors in our information variables. Measuremen t error is the difference between the value of characteristics 14 provided by respondent and the true value s. A household survey is generally co nsidered to have true values, assuming that the measured characteristics are accurately reported and obtained throu gh appropriate procedures (Kasprzyk, 20 05 ). However, there can be measurement errors in a household survey caused by questionnaire, data - coll ection method, interviewer, and respondent (Biemer et al., 2011 ). In particular, as the respondents have different experiences, knowledge and attitudes, it is possible that they interpret the meaning of questionnaire items differently. This is a problem fa ced by all primary data collection efforts. 15 . E MPIRICAL METHODS In this section, we develop a model of factors r elated to cross - sectional variation in farm - g ate rice prices for smallholder farmers in developing countries. We use a search cost model . Search cost models are usually based in the notion of a prices or collecting information about alternative suppliers (Baye, Morgan and Scholten, 200 6 ; Wilson, 2 012). From a , search cost s can be defined as the cost of visiting or contacting different markets to colle ct information to determine where to sell h is produce. Based on the information received and search cost s , farmers would choose the market where they can maximize their profit. If search costs are too high, farmers would sell agricultural produce at low prices to any market instead of investigating alternative markets (Nakasone, Torero , and Minten, 2014 ). For the farmer, sea rch cost is travel, time, and other cost s of collect ing market information. Traditionally, farmers have gained market information from many sources including individual visits, r adio, TV , etc. In more recent times , some researche r s argue that have been reduced with the introduction of modern Information Communication Technologies ( ICTs ) in the developing world . For instance, b y using a mobile phone, a farmer could search for the price of his /her crops so there is no need t o visit markets . J ensen (2007) applies the search cost model to estimate the relationship between information technologies and market performance and economic welfare of producers in the Indi an fisheries sector . His experimental study is focused on how the intr oduction of mobile phones interacts with price dis persion, waste and economics welfare . Tadesse and Bahiigwa (201 5 ) estimate the effect of mobile phone ownership on crop prices received by farmers using the search cost model . Aker and Fafchamps (2010) a nd Lee and Bellemare (2013) also apply the search cost framework to model to farmers searching information 16 through mobile phone use instead of visiting individual markets. The basic s earch cost model for our application can be represented as : (1 ) where is the natural logarithm of the price of monsoon rice received by farmer ; is a vector of search cost and other variab les that affect the price received ; is the assoc iated parameter vector; and is an error term assumed inde pendently and identica lly distributed with zero mean. We define the variable vector in more detail as : (2 ) where is a dummy variable for access to market information (=1 if farmer received market information; =0 otherwise); is a vector of household and farm character istics with associated parameter vector ; and is a vector of market - relat ed variables with associated parameter vector . To investigate the relationship between market information and the farm - gate prices in more detail , we divide market infor mati on into two types: price and non - price . We include separate dummy variables for each information type and source . We also investigate different information variable specifications to examine the robustness of results to including different information types and sources. There are some potential problems with the estimation approach . First, the explanatory variables might have multicollinearity problem s since household - related variables can be correlated with farm an d market - related variab les. For example, a highly educated farmer may choose a high quality rice variety . According to Dohoo et al. (1997), multicollinearity can be evaluated by computing a correlation matrix and Variation Inflation Factors (VIF s ). They noted that when a correlation coefficient is 0.9 or higher multicollinearity may be a concern. We also like to see VIF va lues below 10. We computed the correlation matrix and VIFs for our explanatory 17 variables and results are shown in Table s A. 1 . and Table A.2. Hence, multicollinearity does not appear to be a problem in this study. Another concern is potential endogeneity of explanatory variables . Better information may lead to higher prices but farmers with higher prices may have more resources to col lect better market information. There also may be household unobservable factors that are related to both prices and access to information. If t hose who would benefit from better information are the ones who look for this ty pe of information then estimates of informatio n effects may be biased upward . Furthermore , there are unobservable factors ( such as entrepreneurship, innate ab ility, technology , etc. ) which can also influence price farmers receive. This may also bias the information effect on prices . The final endogeneity concern is spillover effects. Among 118 farmers who sell their rice, about 93% and 81% get price and non - pri ce information respectively. Moreover, 85% of the farmers are engaged in agricultural extension programs . These figures imply that there might be potential spillover effects hat bias causal effects . According to Aker (2010), traditional agricultural extens ion programs generally have spillover effects within villages. Likewise, farmers with information may play important roles del ivering information to oth er farmers within the village. She argued that potential spillover effects are much greater if ICT is be ing used . Endogeneity problems are a concern for this study but data limitations preclude comprehensive identification of causal effects. However, we do investigate potential directions and magnitudes of endogeneity bias in estimated coefficients and cauti on the reader not to make direct causal interpretations for estimated coefficients on information variab les. 18 Table 3 is a summary that shows only the main results of interest. We would expect that a ccess to price information should be positive ly correlated with rice prices received . However, the result (column 1) indicates no statistically significant relationship . Similarly, results (column 2) show no statistically significant relationship between non - price information and rice price levels. T here are three possible explanations . First, it could be that alternative markets are very competitive and access to market inf ormation does not change the price received by farmers. Second, since almost all the farmers have received price information, it could be that it is just difficult to identify a price information effect with this data set. Third, it could be that each of the individual models (columns 1 and 2) are misspecified. To investigate the third explanation , we estimated a model that inclu des both price and non - price information variables (column 3). For this model we find that acces s to non - price information is positively correlated with rice price received by farmers . To investigate this further we estimated models that differentiate non - pri ce information by source and find that TV is the key information delivery source that is positively correlated with higher rice prices (columns 4 and 5). Results show that farmers with non - price information experience rice prices 2. 7 % higher on average tha n other farmers (column 3) , and when the source of non - price information is TV 10 those farmers experience prices 5 .0 % higher on average (column 4) . The comprehensive model which include s all information variables and their sources (column 5 ) indicate that farmers getting non - pric e information from TV receive 5.3% higher rice prices on average . 10 According to our sample, 6 6.1% of the farmers own at least one TV. 19 (1) (2) (3) Market information Access to price information (0.031) - - - - From neighbor - (0.015) - (0.015) From broker - (0.018) - (0.018) From rice miller - (0.026) - (0.026) From DOA (governmental staff) - (0.030) - (0.037) From media - (0.014) - (0.016) Access to non - price information - - (0.016) - - From radio - - (0.016) (0.017) From mobile phone - - (0.015) (0.017) From TV - - (0.013) (0.015) F rom broker - - (0.013) (0.014) (0.001) (0.001) (0.001) (0.021) (0.002) (0.001) (0.005) (0.003) (0.000) (0.016) (0.019) Note 1. *, ** and *** indicate significance levels at 0.1, 0.05, and 0.01 respectively. 2. (1) the relationship between access to p rice information and monsoon rice price (2) the relationship between the sources of price information and monsoon rice price (3) the relationship between access to non - price information and monsoon rice price (4) the relationship between the source of non - pric e information and monsoon rice price (5) the relationship between all sources of both information and monsoon rice price (6) the relationship between all sources of both information and monsoon rice price (farmers with both informa tion only) 20 Table 3 (cont d) Note 1. *, ** and *** indicate significance levels at 0.1, 0.05, and 0.01 respectively. 2. (1) the relationship between ac cess to p rice information and monsoon rice price (2) the relationship between the sources of price information an d monsoon rice price (3) the relationship between access to non - price information and monsoon rice price (4) the relationship between the source of non - price information and monsoon rice price (5) the relationship betwee n all sources of both information and monsoon rice price (6) the relationship between all sources of both informa tion and monsoon rice price (farmers with both information only) Household rice cultivated area (0.004) (0.004) (0.003) (0.004) (0.004) Rice variety 1: Ayeyarmin (0.029) (0.025) (0.027) (0.020) (0.022) (0.019) Rice variety 3: Manawhtikesa (0.028) (0.029) (0.029) (0.026) Rice variety 4: Htikesa (0.046) (0.045) (0.047) (0.040) Rice variety 6: Shwe Bo (0.032) (0.034) (0.026) (0.029) (0.035) Quality of rice Mill (0.022) (0.026) Store (0.015) (0.016) Plan (0.016) (0.017) Buyer Sell_ consumer (omitted) (omitted) (omi tted) (omitted) Sell_ broker Sell_ rice miller Place of sale Place _ farm (0.033) (0.038) (0.036) (0.030) Place _ house (0.032) (0.033) (0.034) Place _ town (0.062) (0.069) (0.064) Plac e _ road (0 .049) (0.053) (0.050) Place _ other (0.102) (0.095) (0.102) Village Village 1: Ko Yan Taung (omitted) (omitted) (omitted) (omitted) Village 2: Pha Lan Tine Village 3: Nyaung Kaing (0.026) (0.027) (0.027) (0.028) (0.028) Constant Observations R - squared 21 TV can be a good source of information about agricultural production and often provides opinions or recommendations to farm ers by quoting experts . Farmers can prepare for production and mar ket sales based on the information obtained through TV programs. TV is better at delivering visual effects than other mediums. The ages of farmers who got information from the TV were between 41 and 66, and their average education level was about 7 years. According to Rahman, Lalon and Surya (2016), f armers prefer television as a source of information because of the With the rapid development of ICT around the world, the number of mobile phone users has a lso increased in Myanmar. In 2017, the mobile phone usage rate in Myanmar was 110 .43% of the population , which is an increase of 22% over 2016 (Mizzima , 2018 ) . The internet usage rate has also skyrocketed since 2013. In the survey data used here 91.5% of h ouseholds have at least one mobile phone . The data even suggests that households are more likely to have mobile phones than a TV. However, our results do not find a significant relationship between mobile phones as a source of market information and rice prices received . This result is c onsistent with some other research (Fafchamps and Minten, 2012; Lee and Bellemare, 2013). Another possible outcome was that ac cess to information from broker might have a connection with the prices since 96% are sellin g their rice to broker, getting price (72%) and non - price information (53%) from them but it was found that there was no relationship between them. About 35.6% of t he farmers were counted as they obtained price and non - price information from brokers and so ld rice to brokers. On the other hand, only 1.7% of the respondents received both information from the broker but did not sell it to the broker. Among hou sehold cha racteristics, farmer education level was found to have a positive and 22 statistically signific ant relationship with rice prices received by farmers . More educated farmers may have more ability to negotiate based on their knowledge and training . Among farm c haracteristics, some rice varieties received higher prices on average than others . Results suggest higher prices for Ayeyarmin (column 1 and 4 ), Ayearpadaethar (column 1, 3 and 4 ), and Shwe Bo (column s 1, 2 and 3 ). These results are generally consistent wi th price premiums for quality and uniform ity . Higher prices for Shwe Bo appear coun ter intuitive but t here is only one farmer who grew the variety in our survey data so this result is spurious. Ayeyarmin and Ayearpadaethar are high quality and more commonl y planted. Results for marke t - related variables indicate negative relationships bet ween using broker s as a buyer and the level of farm - gate prices in all models . In addition , w e f ou nd that as place of sale, road has a negative relationship with price in s ome models . This may be explained by the fact that farmers who sell their rice at the road may find it difficult to bargain for high er prices. Village dummies are statist ically significant showing that Nyaung Kaing is associated with lower farm prices for rice. As shown in Table 1, Pha Lan Tine has the highest proportion of rice cultivation . A l arge proportion of rice farmers may reduce farmer bargaining power. Finally , results indicate that having a road in good condition is associated with low er farmer rice prices. This seems to be th e result of selection bias since 91% of the farmers responded they have road s in good condition. 23 Agricultural output price dispersion in underdeveloped countries has always been a development challenge in smallholder agriculture . Considerable previous research has found a close relationship between output price s received by farmers and market information ( Svensson and Yanagizawa , 2008; Salasya and Burger , 2010 ; Mather, Boughton and Jayne , 2011 ; Tadesse and Bahi i gwa , 2 01 5). In this paper , we extend this literature by differentiating market informa tion into price and non - price types , and a lso focus on the sources of each information type. We apply our model to survey data on smallholder rice farme rs in the Magway region of Myanmar. Our main focus was the relationship between information variables and the level of rice prices received by f armers. W e found that non - price information from TV was the only statistically significant information variable related to farm - gate rice prices . No source of price information was found to be statistically significant ly related to rice prices received, al though the re is some evidence that selling to brokers is associated with a lower average rice price. Mobile phone and TV use are widespread, even in rural Myanmar. In our survey data, 91.5% of households owned mobile phones and 66.1% owned a TV . Despite the prevalence of mobile phone s , our empirical results suggest that farmers who received market information via mobile phone do not receive higher prices on average . However, we found that non - price information obtained via TV had a statistically significa nt positive relationship with rice prices received by farmers. This suggests that TV programming may be a good medium for deliver ing agricultural information to smallholders in rural areas. W e also found that farmers who sell to brokers received lower ric e prices on average, suggesting that brokers may have an advantage in price negotiations with farmers. O ur empirical results have some limitations . The sample is relatively small with some 24 variables experiencing little variation across respondents . Also, t he data is from a spec ific region of rural Myanmar so the external validity of the results is low . We are also unable to identify causal effects . These limitations need to be addressed in future research. 25 APPENDICES 26 A PPENDIX A. TABLES Note 1. Variety 1 - 6 are the r ice varieties ; Ayeyarmin , Ayearpadaethar , Manawhtikesa , H tikesa , Kayinma , Shwe Bo in the order. 2. The village names are Ko Yan Taung , Pha Lan Tine , Nyaung Kaing in the order. Price information Non - price information Age Gender Education Experience Neighbor Broker Miller DOA Media Radio Mobile TV Broker Price info Neighbor 1.000 Broker - 0.479 1.000 Miller 0.023 0.144 1.000 DOA - 0.059 0.058 0.399 1.000 Media - 0.009 - 0.139 - 0.026 - 0.048 1.000 Non - price info Radio 0.094 - 0.156 0.009 - 0.039 0.299 1.000 Mobile - 0.022 0.058 - 0.030 0.174 0.225 0.002 1.000 TV 0.062 0.181 0.101 - 0.042 0.208 0.123 0.141 1.000 Broker 0.118 - 0.025 - 0.012 - 0.097 - 0.089 - 0.022 - 0.191 - 0.068 1.000 Age 0.031 0.050 0.034 0.002 - 0.039 - 0.011 - 0.011 - 0.014 0.089 1.000 Gender 0.097 - 0.018 0.062 0.025 - 0.025 0.114 - 0.182 0.122 0.149 0.064 1.000 Education 0.0 53 - 0.089 - 0.085 0.039 0.133 - 0.084 0.195 0.121 - 0.084 - 0.210 0.080 1.000 Experience 0.026 0.021 - 0.044 0.083 - 0.063 - 0.022 0.067 0.007 0.112 0.743 0.079 - 0.151 1.000 HH size - 0.172 0.161 - 0.022 0.001 - 0.087 - 0.040 0.084 0.091 - 0.032 0.091 0.040 - 0.037 0.136 Landhold 0.139 - 0.182 - 0.027 0.016 0.065 - 0.041 0.162 0.111 - 0.025 - 0.002 0.064 0.402 0.061 Cultivated land 0.114 - 0.121 0.050 0.100 0.039 - 0.089 0.089 0.147 - 0.049 - 0.039 0.079 0.404 0.052 Variety1 - 0.159 0.214 0.113 0.077 0.007 - 0.033 0.232 0.05 2 - 0.372 0.053 - 0.086 0.111 0.043 Variety2 0.058 0.017 - 0.047 - 0.054 - 0.065 - 0.085 - 0.217 - 0.108 0.204 - 0.099 0.080 - 0.056 - 0.152 Variety3 0.016 - 0.019 - 0.037 - 0.015 - 0.084 - 0.069 - 0.086 - 0.073 0.154 - 0.044 0.044 - 0.240 - 0.046 Variety4 - 0.103 - 0.019 - 0. 037 - 0.015 0.180 0.231 0.044 0.071 0.046 0.133 0.044 - 0.113 0.075 Variety5 - 0.059 0.058 - 0.021 - 0.009 - 0.048 - 0.039 - 0.049 - 0.042 0.088 - 0.034 0.025 - 0.117 0.031 Variety6 - 0.059 0.058 - 0.021 - 0.009 0.178 - 0.039 - 0.049 0.205 - 0.097 0.091 0.025 0.070 0.075 Consumption 0.077 0.008 - 0.067 0.016 - 0.061 - 0.099 - 0.029 0.152 0.045 0.110 - 0.022 0.235 0.225 Credits 0.038 0.021 0.105 0.042 0.013 0.003 0.131 0.144 - 0.113 - 0.058 0.148 0.092 0.020 Extension - 0.094 0.261 0.098 0.039 0.047 - 0.082 0.226 0.129 - 0.073 0. 029 0.073 0.115 0.101 Village1 - 0.069 0.090 0.093 0.099 - 0.027 - 0.066 0.200 0.076 - 0.439 0.031 - 0.086 0.185 0.056 Village2 0.008 0.146 0.023 - 0.059 0.037 - 0.010 - 0.022 0.012 0.267 - 0.110 0.023 - 0.017 - 0.063 Village3 0.072 - 0.258 - 0.132 - 0.053 - 0.007 0.0 86 - 0.209 - 0.101 0.227 0.080 0.076 - 0.197 0.002 Sell_consumer - 0.084 0.082 - 0.030 - 0.012 0.093 0.127 0.089 0.116 0.125 0.047 0.035 - 0.055 0.057 Sell_broker 0.134 - 0.131 0.049 0.019 - 0.097 - 0.145 - 0.091 - 0.129 0.053 0.110 - 0.057 - 0.032 0.074 Sell_miller - 0.103 0.101 - 0.037 - 0.015 0.048 0.081 0.044 0.071 - 0.170 - 0.179 0.044 0.086 - 0.142 Place_farm 0.194 - 0.200 0.122 - 0.021 - 0.026 0.009 - 0.030 - 0.002 0.143 - 0.097 - 0.091 - 0.046 - 0.034 Place_house - 0.015 0.082 - 0.288 0.033 - 0.083 - 0.152 0.056 0.015 - 0.009 0 .057 0.013 0.135 0.111 Place_town - 0.084 0.082 0.269 - 0.012 0.093 0.127 - 0.070 - 0.059 - 0.007 - 0.125 0.035 - 0.011 - 0.103 Place_road - 0.059 0.058 - 0.021 - 0.009 0.178 0.218 0.174 0.205 0.088 - 0.052 0.025 0.039 - 0.021 Place_other - 0.084 - 0.065 - 0.030 - 0.012 0.093 - 0.056 - 0.070 - 0.059 - 0.138 0.098 0.035 - 0.099 - 0.042 Mill - 0.112 - 0.034 0.079 - 0.027 0.242 0.056 0.078 - 0.045 0.017 - 0.245 - 0.050 0.088 - 0.157 Store - 0.015 - 0.039 - 0.120 - 0.079 0.143 0.066 0.041 - 0.067 0.025 - 0.048 - 0.042 0.249 0.079 Plan - 0.086 0.105 - 0.081 - 0.094 0.136 0.040 - 0.009 - 0.008 0.118 - 0.050 0.005 0.277 0.076 Road - 0.054 - 0.005 - 0.059 0.030 - 0.262 - 0.107 0.100 0.067 - 0.071 0.011 0.030 0.111 - 0.005 27 Table A.1 ( cont ) Note 1. Variety 1 - 6 are the r ice varieties ; Ayeyarmin , Ayearpadaethar , Manawhtikesa , Hti kesa , Kayinma , Shwe Bo in the order. 2. The village names are Ko Yan Taung , Pha Lan Tine , Nyaung Kaing in the order Table A.1 ( cont ) Note 1 . Variety 1 - 6 are the r ice varieties ; Ayeyarmin , Ayearpadaethar , Manawhtikesa , Htikesa , Kayinma , Shwe Bo in the order. 2. The village names are Ko Yan Taung , Pha Lan Tine , Nyaung Kaing in the order. HH size Landhold Cultivate d land Variety1 Variety2 Variety3 Variety4 Variety5 Variety6 Consump - tion Credits Exten - sion Village 1 Village2 Village3 HH size 1.000 Landhold 0.252 1.000 Cultivated land 0.173 0.790 1.000 Variety1 0.200 0.195 0.183 1.000 Variety2 - 0.142 - 0.183 - 0.166 - 0.705 1.000 Variety3 0.070 - 0.038 - 0.113 - 0.195 0.029 1.000 Variety4 - 0.067 0.100 - 0.011 - 0.085 0.029 - 0.026 1.000 Variety5 - 0.234 - 0.107 - 0.103 - 0.112 - 0.054 - 0.015 - 0.015 1.000 Variety6 - 0.058 - 0.098 - 0.074 - 0.112 - 0.054 - 0.015 - 0.015 - 0.009 1.000 Consumption 0.399 0.436 0.480 0.166 - 0.180 - 0.087 - 0.020 - 0.116 - 0.116 1.000 Credits 0.096 0.125 0.118 0.224 - 0.255 - 0.071 - 0.071 0.042 0.042 0.105 1.000 Extension 0.250 - 0.054 0.001 0.320 - 0.077 0.069 - 0.231 - 0.218 0.039 0.124 0.123 1.000 Village1 0.205 0.205 0.092 0.739 - 0.546 - 0.151 - 0.151 - 0.086 - 0.086 0.146 0.286 0.349 1.000 Village2 - 0.065 - 0.201 0.047 - 0.197 0.273 - 0.1 03 - 0.103 - 0.059 - 0.059 - 0.010 - 0.311 0.010 - 0.594 1.000 Village3 - 0.169 - 0.026 - 0.155 - 0.649 0.345 0.283 0.283 0.162 0.162 - 0.158 - 0.004 - 0.415 - 0.533 - 0.363 1.000 Sell_ consumer Sell_ broker Sell_ miller Place_ farm Place_ hous e Place_ town Place_ road Place_ other Mill Store Plan Road Sell_ consumer 1.000 Sell_broker - 0.624 1.000 Sell_miller - 0.021 - 0.768 1.000 Place_farm - 0.030 0.049 - 0.037 1.000 Place_house - 0.164 0.195 - 0.115 - 0.535 1.000 Place_town - 0.017 - 0.298 0.396 - 0.030 - 0.373 1.000 Place_road 0.704 - 0.440 - 0.015 - 0.021 - 0.263 - 0.012 1.000 Place_other - 0.017 0.028 - 0.021 - 0.030 - 0.373 - 0.017 - 0.012 1.000 Mill 0.210 - 0.257 0.156 - 0.067 - 0.205 0.457 0.322 - 0.038 1.000 Store 0.020 - 0.075 0.079 - 0.199 0.028 0.153 0.108 0.020 0.271 1.000 Plan - 0.002 - 0.039 0.051 - 0.158 0.087 0.129 0.091 - 0.002 0.219 0.603 1.000 Road 0.042 - 0.067 0.052 0.074 0.073 - 0.184 0.030 0.042 - 0.237 - 0.315 - 0.199 1.000 28 Variable VIF 1/VIF Variable VIF 1/VIF Variety_1 6.21 0.16 Education 2.09 0.48 Landhold 5.49 0.18 Mill 2.05 0.49 Sell_broker 5.41 0.18 Extension 2.01 0.50 Village_3 5.11 0.20 Broker_nonprice 1.87 0.53 Sell_miller 4.73 0.21 Mill er_price 1.84 0.54 Cultivated 4.72 0.21 HH_size 1.79 0.56 Place_house 4.42 0.23 Road 1.75 0.57 Variety_2 3.64 0.27 Mobile_nonprice 1.74 0.58 Village_2 3.25 0.31 Variety_3 1.73 0.58 Experience 3.21 0.31 Neighbor_price 1.72 0.58 Age 3.14 0.32 Me dia_price 1.69 0.59 Place_road 3.14 0.32 DOA_price 1.52 0.66 Place_farm 2.99 0.33 Radio_nonprice 1.49 0.67 Place_town 2.6 0.39 Variety_6 1.48 0.67 Store 2.31 0.43 TV_nonprice 1.46 0.68 Plan 2.26 0.44 Credits 1.42 0.71 Broker_price 2.26 0.44 Var iety_4 1.41 0.71 Place_other 2.16 0.46 Gender 1.33 0.75 Consumption 2.13 0.47 Variety_5 1.29 0.78 Mean VIF 2.65 29 APPENDIX B: FIGURES Source: World Bank (2018). 0 2 4 6 8 10 12 14 16 18 20 0 10 20 30 40 50 60 70 80 90 100 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Unit: USD billion Non-agriculture (% of GDP) Agriculture, forestry, and fishing, value added (% of GDP) GDP of agriculture, forestry and fishing (current US$) Unit: % 30 Source: World Bank (2018). Source: World Bank (2018). 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 1980 1985 1990 1995 2000 2005 2010 2015 2016 2017 Rural population (% of total) Urban population (% of total) Unit: % 0 10 20 30 40 50 60 70 80 90 100 Employment in agriculture Employment in industry Employment in services Unit: % 31 32 0 5 10 15 20 25 2009 2010 2011 2012 2013 2014 2015 2016 Unit: Million Acre Magway Ayeyarwady Total 33 REFERENCES 34 REFERENCES Aker, J. C. (2010). Information from markets n ear and far: Mobile phones and agricultural markets in Niger. American Economic Journal: Applied Economics , 2(3), 46 - 59. Aker, J. C., & Fafchamps, M. (2010). How does mobile phone coverage affect farm - gate pri ces? Evidence from West Africa. Department of Economics and the Fletcher School, Tufts University. Baye , M. R., Morgan, J., & Scholten, P. (2006). Information, search, and price dispersion. Handbook on economics and information systems , 1, 323 - 375. Biemer, P. P., Groves, R. M., Lyberg, L. E., Mathi owetz, N. A., & Sudman, S. (Eds.). (2011). Measurement err ors in surveys (Vol. 173). John Wiley & Sons. CSO (2019). CSO (Central Statistical Organization in Myanmar) Year Book 2019. Dohoo, I. R., Ducrot, C., Fourichon, C., Donald, A., & Hurnik, D. (1997) . An overview of techniques for dealing with large numbers of independent variables in epidemiologic studies. Preventive veterinary medicine , 29(3), 221 - 239. Eggleston, K., Jensen, R., & Zeckhauser, R. (2002). Information and communication technologies, m arkets, and economic development. The Global Information T echnology Report 2001 - 2002: Readiness for the Networked World. Elly, T., & Epafra Silayo, E. (2013). Agricultural information needs and sources of the rural farmers in Tanzania: A case of Iringa ru ral district. Library review, 62(8/9), 547 - 566. English, A. (2008). Determinants of Liberian farmgate cocoa prices. European Commission (2012). Agricultural markets and small - scale producers: access and risk management tools. Rural Development, Food Secu rity and Nutrition. Fafchamps, M., & Minten, B. (2012). Impact of SMS - based agricultural information on Indian farmers. The World Bank Economic Review , 26(3), 383 - 414. - , 2011. (https://www.ictworks.org/agriculture - kno wledge - based - industry/) [Accessed on May 11, 2019] Handy, S., Cao, X., & Mokhtarian, P. (2005). Correlation or causality between the built environment and travel behavior? Evidence from Northern California. Transportation Research Part D: Transport and En vironment , 10(6), 427 - 444. 35 Jensen, R. (2007). The digital provide: Information (technology), market performance, and welfare in the South Indian fisheries sector . The quarterly journal of economics , 122(3), 879 - 924. Kasprzyk, D. (2005). Measurement error in household surveys: sources and measurement . Mathematica Policy Research. Lee, K. H., & Bellemare, M. F. (2013). Look who's talking: the impacts of the intrahousehold allocation of mobile phones on agric ultural prices. The Journal of Development Studies , 49(5), 624 - 640. Magesa, M. M., Michael, K., & Ko, J. (2014). Agricultural market information services in developing countries: A review. Advances in Computer Science: an International Journal, 3 (3), 38 - 47. Mather, D., Boughton, D., & Jayne, T. S. (2011). Smallholder heterogeneity and maize market participation in Southern and Eastern Africa: Implications for investment strategies to increase marketed food staple supply (No. 1096 - 2016 - 88373). Mi zzima (2018) . Mobile phone usage in Myanmar increases 110 percent 2018. (http://mizzima.com/business - domestic/mobile - phone - usage - myanmar - increases - 110 - percent) [Accessed on March 18, 2019] Nakasone, E., Torero, M., & Minten, B. (2014) . The power of information: The ICT revolution in agricultural development. Annu. Rev. Resour. Econ ., 6(1), 533 - 550. Poap ongsakorn, N., & Pantakua, K. (2014). Assessing the Thai paddy pledging policy: Its performance and social costs. World Food Policy , 1 (2), 2 - 29. Poole, N. (2017). Smallholder agriculture and market participation . Practical Action Publishing. Rahman, M. A., Lalon, S. B., & Surya, M. H. (2016). Information sources preferred by the farmers in receiving farm information. International Jou rnal of Agricultural Extension and Rural Development, 3(12), 258 - 262. Salasya, B. D. S., & Burger, C. P. J. (2010). Determinants of the place of sell and price of kale for Kiambu, Kenya. African Journal of Agricultural Research , 5(9), 805 - 812. Singleton Jr, R., Straits, B. C., Straits, M. M., & McAllister, R. J. (1988). Approaches to social research . Oxford Uni versity Press. Svensson, J., & David Yanagizawa, I. I. E. S. (2008). Stockholm University. Getting Prices Right: The Impact of the Market Informa tion Service in Uganda. 36 Ethiopia . World development , 68, 296 - 307. USDA (2018). Burma Union of grain and feed annual (2018 annual report). USDA Foreign Agricultural Se rvice. Wilson, C. M. (2012). Market frictions: A unified model of search costs and switching costs. European Economic Review , 56(6), 1070 - 1086. World Bank ( 2016 ) . Myanmar: Analysis of farm production economics (English). Washington, D.C. : World Bank Gr oup. ___ _______ (2018). World Development Indicators. (https://databank.worldbank.org/source/world - development - indicators) [Accessed on March 17, 2019]