우리나라 양파 가격 안정화 방안에 관한 연구 : 산지가격의 인과성 분석과 딥러닝을 활용한 가격예측을 중심으로

홍성민 강원대학교 대학원 2022 국내석사

농산물은 복잡한 유통구조와 날씨의 영향을 많이 받는 특성상 공급 불안정성이 내재하여 있으며 수요의 가격탄력성이 낮아 작은 공급변화에도 가격 변동성이 크게 나타난다. 이러한 원인 때문에 농산물은 2∼3년을 주기로 가격의 폭락과 폭등이 자주 발생하고 있다. 정부에서는 이러한 불안정한 농산물 가격 문제를 해결하고자 다양한 정책의 시행과 가격 예측을 하는 등의 노력을 하고 있다. 하지만 이러한 노력에도 불구하고 국내 농산물 가격 변동성은 미국, 일본 등의 국가와 비교할 때 매우 높은 편으로 정책의 효과성이 미흡한 편이다(김동환·류상모, 2016). 또한, 여전히 농산물 유통의 비효율성과 수급 불안으로 인해 가격의 급등락이 여러 번 발생하여 농가의 소득안정과 소비자의 가계 운영에 많은 어려움이 따르고 있다(감사원, 2017). 그중에서도 한국의 대표 조미채소 중 하나인 양파는 산지에서의 가격 형성기능이 미약하고, 수확기 출하 물량이 도매시장에 일시에 집중되면서 가격 등락 폭이 더욱 심화되고 있다(농림축산식품부, 2019). 따라서 양파를 비롯한 국내 농산물 가격안정화문제 해결을 위해 다양한 정책과 이와 관련된 연구가 필요한 시점이다. 본 연구의 목적은 국내 대표 조미채소 중 하나인 양파를 대상으로 가격안정화방안을 마련하는 것에 있다. 이를 위해 양파의 산지 가격을 권역별로 나누어 인과관계를 규명하여 가격선도를 이끄는 지역이 어디인지를 파악하고, 양파가격을 예측하여 농산물 가격 안정화 정책에 기반이 되는 연구를 하고자 한다. 먼저 양파가격이 산지 간의 어떤 인과성을 갖고 있으며, 가격선도가 일어나는 곳을 알아보기 위해 VAR 분석을 진행하였다. 분석 결과, 우리나라 최대 양파 주산지인 전라권(전남, 전북, 광주)이 산지 가격을 주도하는 것이 아닌 경상권(경남, 경북, 대구)에서 산지 가격을 주도하는 것으로 나타났다. 이런 원인으로는 다양한 원인이 있을 것이지만 양파의 유통구조 상의 특징에서 찾을 수 있다. 산지에서 출하된 양파는 대부분 가락시장이나 대구 공판장을 통해 가격이 형성되기 때문에 대구 공판장과 비교적 거리가 가까운 경상권의 가격이 타 권역에 영향을 미치는 것으로 추정되었다. 또한, 양파는 다른 농산물과 비교하여 저장성이 강해 생산량 대비 저장 비율이 높은 품목이다. 김성우 외(2016)에 의하면 국내 양파 저장업체의 현황은 경상권 445개 전라권 304개로 경상권의 저장업체 수가 더 많았다. 양파 저장업체 수가 해당 산지권역에 많이 분포되었다는 것은 산지가격에 반응하기 위한 충분한 시간적 여유가 있다는 것을 의미한다. 또 정현우 외(2017)에 의하면 전라도 지역의 농업생산기반시설이 경상도 지역보다 전반적으로 취약한 것으로 나타났다. 농업생산기반시설이 취약하다는 것은 양파품질 및 감모율에 영향을 미치기 때문이다. 다음으로 양파가격의 예측력을 향상하기 위해 구조모형과 머신러닝을 이용하여 양파의 도매시장 가격 예측력을 향상할 수 있는 모델을 구축하였다. 예측력 향상을 위한 예측모형 개발을 위해 양파 중기선행 관측모형을 구축하고 이를 월별로 재생산하여 단기 예측모형을 결합함으로써 일반 단기모형과 중기선행관측모형을 결합한 모형의 예측성을 분석했다. 머신러닝 분석에 양파 중기모형의 값을 반영하는 것은 머신러닝이 양파가격 예측을 위한 학습을 할 때 사전적 정보를 가지고 분석을 하는 것에 있어 중요한 의미가 있다. 분석결과 양파의 중기선행관측모형의 추정치 정보를 반영한 모형의 예측력이 더 높게 나타났다. 또한 머신러닝의 분석 성능에 대한 평가 지표로서 MAE, RMSE, MAPE를 활용하였다. 평가 결과 양파의 중기모형 정보를 반영한 LSTM모형과 반영하지 않은 LSTM모형의 가격 예측의 오차율은 각각 17.5%, 20.2%로 중기모형을 반영한 LSTM모형은 비교적 정확한 예측을 하는 것으로 나타났다. 반면, 중기모형을 반영하지 않은 LSTM모형은 비교적 합리적 예측을 하는 것으로 나타났다. 두 연구를 종합해 볼 때 향후 양파의 가격에 영향을 미치는 재고량에 관한 연구가 필요하다. 양파는 저장성이 강해 중·만생종은 수확기 저온저장 하여 9월 하순부터 출하되며, 매년 전체 생산량의 약 50∼60%가 저장되어 출하되고 있다. 본 연구에서는 양파의 재고량을 추정하여 예측 변수에 반영하였지만, 선행연구를 볼 때 재고량을 중심으로 한 연구는 많지 않다. 또 재고량의 경우 민간업체가 대부분 관리하여 정확한 물량 파악이 어려운 실정이다. 양파 가격안정화를 위해 다양한 노력이 필요하지만, 양파의 수급과 가격안정 체계를 효과적으로 이루기 위해서는 민간 저장업체의 적극적인 참여가 필요하다. 그리고 도매시장을 포함한 다양한 출하 장려금의 지원과 낙후된 저장창고 시설을 현대화하는 등의 서비스 제공을 검토할 필요가 있다. 정부와 지자체의 역할과 본 연구 결과를 기초로 양파 가격안정화 사업과 정책계획 수립에 있어 기초자료로 이용될 수 있을 것으로 기대한다. Unstable supply is inherent in Agricultural products due to complex distribution structure and its characteristics of being greatly affected by weather, and price fluctuation occurs often even with small changes in supply due to low price elasticity of demand. With these reasons, price slump and jump of agricultural products occur very often every two to three years. To solve the problem of this unstable price of agricultural products, the government is making effort such as implementing various policies and predicting prices. Despite these efforts, however, price fluctuation of domestic agricultural products is very high compared to that of USA and Japan and the effectiveness of policies are not sufficient (Kim Donghwan, Ryu Sangmo, 2016). In addition, the price slump and jump occured several times due to inefficiency in the distribution of agricultural products and unstable supply and demand, making many difficulties to stabilize rural household incomes and operate household finances of consumers (Gam Sawon, 2017). Onions, one of representative condiment vegetables in Korea, have weak function of price formation in the production area and the range of price fluctuation is getting deepen as the amount of shipment during the harvest time is concentrated in wholesale market at a time (Ministry of Agriculture, Food and Rural Affairs, 2019). Therefore, it is time when various policies and related studies are required to solve the issue of price stabilization of domestic agricultural products including onions. The purpose of this study is to prepare plans of price stabilization for onions, one of representative condiment vegetables in Korea. To do this, the study divided the price of onion production by region and identified the causal relationship to figure out the region that leads the price, and it predicted the price of onions to conduct study that can be the basis of price stabilization policy of agricultural products. First, VAR analysis was conducted to find out what kind of casual relationship onion price have between production areas and where price leadership occurs. The results of analysis found that Gyeongsangnam-do regions (Gyeongnam, Gyeongbuk, Daegu) is leading the price of production areas, not Jeollanam-do regions (Jeonnam, Jeonbuk, Gwangju), the largest onion producing areas in Korea. There may be various causes for this, but it can be found in the characteristics of distribution structure of onions. Since most of onions shipped from the production areas form their prices through Garak Market or Daegu joint market, it was estimated that the price in Gyeongsang regions where is relatively close to Daegu joint market affects other regions. In addition, onions has high stocking ratio compared to output with high storability unlike other agricultural products. According to Korea Rural Economic Institute (2016), the numbers of domestic storing enterprises in Gyeongsang regions were 445 which is more than Jeolla regions with 304. The fact that onion storing enterprises are distributed in these production areas mean that there is enough time to respond to the price of production area. In addition, Jeong Hyeonwoo et al., (2017) found that agricultural production infrastructure in Jeollado regions is vulnerable than that of Gyeongsangdo regions. This is because vulnerable agricultural production infrastructure affects the quality of onions and loss rate. Then, the model that can improve predictive power of the price in wholesales market of onions was built by using structural model and machine learning to improve the predictive power of onion price. To develop predictive model to improve predictive power, mid-term preceding observation model of onions was built and reproduced it monthly. And then it was combined with short-term predictive model to analyze the predictive power of model which was combined with general short-term and mid-term preceding observation model. Reflecting the value of mid-term model of onion in the machine learning has an important meaning in analyzing with prior information when machine learning learns to predict the price of onions. The results of analysis found that model reflecting estimated information of mid-term preceding observation model of onions had higher predictive power. In addition, MAE, RMSE and MAPE were used as an evaluation indicator for analysis performance of machine learning. As a result of evaluation, error rates of LSTM model that reflected mid-term model information of onions and LSTM model that didn't reflect it were 17.5% and 20.2% respectively, and this indicates that LSTM that reflected mid-term model showed relatively accurate predictions. On the other hand, LSTM model that didn't reflect mid-term model was found to have relatively reasonable prediction. When the both studies were combined, we can know that study on inventories affecting the price of onions is required in the future. Since onions have strong storability, medium and late maturing varieties are stored at low temperature during harvest time and shipped from the end of September. About 50 to 60% of total production is stored and shipped every year. This study estimated inventories of onions and reflected it to predictor variables, but there are not many preceding studies that focused on inventories. In addition, private enterprises usually manage the inventories, so it is hard to identify the accurate inventory. Various efforts are required to stabilize the price of onions, but active participation of private storing enterprises are also required to effectively achieve the supply and demand of onions and stable system of price. And it is necessary to review the provision of services such as supporting grants for various shipment including wholesale market and modernizing old storehouse facilities. Based on roles of government and local governments and the results of this study, it is expected that this can be used as a basic data for establishing stabilization project of onion price and policy planning.

산지가격의 변화에 따른 신용평가의 변화에 대한 연구

문길수 전북대학교 경영대학원 2012 국내석사

FTA, 규모화, 자본집약화 등에 따라 최근 몇 년 동안 사육두수 및 가격변화가 이루어지고 있는 시점에서 한우는 한우농가의 축산활동에 중심이 되는 재고자산으로 농업인의 재무상태 및 영농활동의 의지도 등을 파악할 수 있는 중요한 수단이 된다. 따라서 본 연구는 한우의 미래의 산지가격이 변화할 경우를 대비하여 농신보(영어) 신용평가모형의 현재를 진단하고자 한다. 과거의 산지가격 데이터를 중심으로 하여 산지가격이 변화할 경우 농신보 신용평가에 미치는 변화를 예상 및 평가하고, 이에 따르는 제도 개선 방향을 제시하고자 한다. 이를 위하여 김제시(GimJe) 지역의 한우농가 중 15곳의 한우농가를 대상으로 신용조사시점별로 총 7차례의 산지가격을 조사하여 농신보의 신용조사 기법으로 직접 면접 조사하였다. 시장상황에 따른 산지가격의 변화는 한우농가의 생애주기별로 그룹화된 개별자산의 비준가격의 변화를 가져오고 이를 통하여 신용조사시점별로 두당평균가격, 재고자산평가액, 매출액, 각종 재무비율, 농신보 신용보증 가능금액이 변화됨에 알 수 있었다. 이런 변화 속에서 각종 변화된 정보들과 농신보 보증가능금액과의 분산분석결과 농신보 신용보증 가능금액에 직접적인 영향을 미치고 있음을 알 수 있었다. 본 연구결과 미래의 산지가격의 변화와 연관되는 과대·과소계상 문제점과 다양한 신용평가구간의 생성으로 신용평점을 산출하는 보완점의 필요함을 알 수 있었다. 이를 위하여 산지가격의 변화의 특성과 재무비율, 신용평점 및 보증가능금액의 변화의 특성을 반영한 농신보만의 특화된 심사시스템 체제 마련이 필요하다고 볼 수 있겠다. 이를 위하여 다음과 같은 제안을 하고자 한다. 첫째, 신용조사시점의 적정한 산지가격을 적용하여 가격의 변화에 따른 과소 및 과대계상의 문제를 최소화하여야 한다. 둘째, 현 개인신용평가시스템의 정비가 필요하다. 셋째, 물가상승 및 하락시에 자산을 안정적으로 평가할 수 있는 모형이 필요하다. 넷째, 품질관리 시스템을 도입한 컨설팅 개념의 도입이 필요하다. 신용평가 모형의 제도적 개선을 통하여 한우농가의 미래 성장성의 가치증대효과를 가져올 수 있을 것으로 기대해 본다. According to FTAs, growth in scale and capital-intensification of livestock farmers, the number of livestock and price of livestock kept changing during past several years. As an inventory asset in the center of stockbreeding activity of livestock farmers, Korean native cattle is an important means to learn the financial status of livestock farmers and their will on stockbreeding. The purpose of this study is diagnosing the present of credit rating model in the [Credit Guarantee Fund for Farmers and Fishermen] in preparation for future change in producing area price of Korean native cattle. In this study, the impact of changes in producing area price of Korean native cattle on the credit rating of Credit Guarantee Fund for Farmers and Fishermen is forecasted and evaluated based on past producing area price data. As conclusion, system improvement suggestions are given based on study results. 15 livestock farmers in Gimje area had been visited for this study. 7 times of producing area prices has been collected for each credit-survey time. The credit survey method of Credit Guarantee Fund for Farmers and Fishermen was used during the interviews and data collection. It was possible to learn that the changes in producing area price dependent on market situation cause the change in calculated price of individual assets, which are grouped by lifecycle of livestock farmers. It was possible to know that the average price per cattle, evaluated amount of inventory asset, sales revenue, various financial ratios and available amount for credit guarantee by [Credit Guarantee Fund for Farmers and Fishermen] are different at each credit-survey time. Variance analysis was done on these changed information and available amount for credit guarantee by [Credit Guarantee Fund for Farmers and Fishermen]. By the result of variance analysis, it was possible to know that they have direct impact on the available amount for credit guarantee by Credit Guarantee Fund for Farmers and Fishermen. The results of this study suggested that there is an issue of over-estimating and under-estimating related to the change in future producing area price. In addition, it was possible to know that a system improvement is required which would create various credit rating segments and calculate the credit points based on those. In order to do this, it is required to establish a credit rating system exclusive to [Credit Guarantee Fund for Farmers and Fishermen], which would reflect characteristics of changes in producing area price, financial ratios, credit point and available amount for credit guarantee. As conclusion, following suggestions are given. First, the issue of under-estimating and over-estimating after producing area price change should be minimized by applying proper producing area price at the time of credit-survey. Second, current individual credit evaluation system should be improved. Third, a model, which can stably evaluate assets when prices rise and fall, is required. Fourth, introduction of consulting concept together with quality control system are required. It is expected that the result of this study would cause the systematic improvement of credit rating and Korean native cattle farmers would have more growth in the future.

畜産物價格의 變動要因 및 豫測模型에 關한 硏究 : 肥育豚, 韓牛, 肉鷄의 産地價格을 中心으로

김태완 慶尙大學校 2003 국내박사

굴 생산량과 산지가격 및 도매가격의 동적 관계에 관한 연구

변지환 경상대학교 대학원 2019 국내석사

The purpose of this study is to analyze whether there are stable and unstable trends in landing prices and wholesale prices according to the change of volume of oyster in producing area. And if there is a dynamic causal relationship, this study can derive implications by analyzing the influence of mutual influence using VECM model. The data used in this study is monthly volume and price data from January 2008 to December 2018. The main results were as follows. First, changes in production have had a positive impact on the changes in production and wholesale prices. It is reasonable to assume that the changes in the oyster farmers' production have led to a change in the oyster supply and demand status rather than a change in the spot and wholesale price. Second, changes in landing prices have a negative impact on changes in current production. This implies that oyster farmers can not control the amount of excrement and shipment by looking at the change in the price of producing area. The reason that as oyster demand and breeding status are different from each other, shipments are not easy to control.

돼지산지가격의 시계열분석 및 예측

윤두진 건국대학교 농축대학원 2001 국내석사

1. The Background and Purpose of Study The Korean hog raising industry is typically characterized by the long-term variation of production, distribution and price. Especially, the price variation must be the determinant in the amount of work done, sales volume and the like as well as the adjustment of output of producers, meat processing business and distribution business, etc. Until now the future prices of the hog price have been presented simple price forecasting(in fact, this work requiring a tremendous amount of time) which is trends and pattern analysis through past materials, several complicated analysis methods(e.g. moving average method, box-jenkins method and the like). But how accurately the future price of the hog price of the producing area, certainly influenced by seasonality and festive days, is presented by trend analysis method, seasonal trigonometric function method and box-jenkins method is very doubtful. Accordingly, both the research on the analysis techniques well reflective of the characteristics of the hog raising industry with seasonality and the presentation of the future price for hog through it become very important management data for all employees related to the hog raising industry. The hog price of the producing areas, the important factor of changes in livestock raising with hog raising as the pivot and in the meat market, is time-series data given with the passage of time as it is in all economic and business data, and these time-series data are made up of trend variation, cyclical movement and irregular variation. The time-series analysis of the hog price of the producing area with seasonality properly provides hog raisers with the time of intensive hog growing and management and the time of increase and decrease in its output by making a reliable presentation of the predicted value of hog prices based upon statistical theories. Therefore, the time-series analysis of the hog price of the producing area with seasonality is intended to provide basic materials for improving the profitability of the hog raising industry and making an attempt at the stable and normal growth and change of the stockbreeding related to hog raising and the meat market. And this study attempted to analyze time-series data using the stochastic seasonal model, the analysis model thought to be new and more accurate in prediction, and present the predicted value by the model. 2. The Research Method and Discussion The analysis of hog raisers has traditionally used such analysis methods as the moving average analysis of time-series by the trend of the curve and moving average smoothing and hanning, econometric analysis using the regression model, decomposition time series analysis and the like. Time series analysis methods used in combination with Box-Jenkins methods as well as X11, one of representative decomposition time series analysis methods, have recently been presented. It is not only easy to analyze data with seasonal factors in a traditional statistical analysis method. It is very controversial to make a statistical analysis of those data in the formula with the error term added to the deterministic average model as in the seasonal average model or seasonal trigonometric function model of decomposition time series analysis and also it is very questionable how accurately the overall trend of time series can be represented by the linear trend curve and seasonal trigonometric function model. For this reason, this study attempted to present the stochastic seasonal model as another alternative and conduct the time series analysis of hog raisers using the presented stochastic seasonal model. This study used one of statistical packages called MINTAB, the analysis instrument of time series, and set the model through the seasonal autocorrelation function(SACF) and the seasonal partial correlation function(SPACF) in accordance with the time series analysis procedures of Box-Jenkins models. And it made an estimation of the parameters of the model set in a final prediction error(FPE) method, tested the normality of residuals through the normal stochastic plot of residuals of data and tested independence in the stochastic component through the autocorrelation function(ACF), partial correlation function(PACF) and Box-Pierce statistic of residuals. And this study tested the significance of the estimated parameter for the estimate of final prediction error(FPE) through the T-value and then presented the hog price of the producing area after investigation into the goodness-of-fit of the set model through overfitting. Also, data were modified before analysis using the producer price index. This indicates that the principle of parameter redundancy was taken into full consideration by eliminating the trends caused by a rise in price of data. In Figure 4-13, the actual value and the 1-year predicted value were plotted by using the tested model. The ② line of the center on the right thin line of the graph(from the 125th data) is the predicted value and ① and ③ are the upper and lower limit values of the confidence interval. And the predicted value is shown in Figure 4-14. The predicted value can not be blindly believed but is of sufficient value only as reasonable data because it is dependent on past data, which is the limitation of time-series analysis, even though its analysis method and process may be statistical and reasonable. And predicted values presented in this study show the general stream of time series data and the data predicted though the general stream, so that they can be practically used as referential data in medium- and short-term policy or business management analysis. And since actual print-out data are data printed out without a factor of the rise in prices, it is difficult to see the multiplication of all the printed-out values by 100 as the real price including even the factor of accurate price rise. But seeing that government implements the policy of curbing the price rise rate to some extent annually in macroeconomic terms, the visual deviation of a rise in price for hog arising from the price rise is not much large. 3. Conclusion Predicted values presented through reliable data analysis enables hog raisers to determine the intensive hog management period and are conducive to improving the profitability of the hog raising industry and stabilizing the hog market through the adjustment of demand and supply. Predicted values of predicted hog price were shown to be 280,721won in May, 291,512won(highest value) in June, 277,329won in July, 271,213won in August, 260,684won in September, 249,632won in October(lowest value) and 249,656won in November. And it was predicted that the hog price would be 258,290won in December, and 257,834won , 257,910won, 250,985won and 25647won, respectively between January and April in 2002. The most highly predicted value of the hog price in June reflects that the lowered delivery rate is lowered in July and August due to the sultry climate and the efficiency of hog raising farms is lowered. Therefore, the income of the hog raising farms will considerably increase if they best maintain the condition of boars and foster the optimal environment in July and August. Importantly, the hog raising farms are very sensitive to the climate such as seasonality and the like, and so it is very urgent to develop the proper hog raising management technique that is not influenced by the climate.

豚肉의 産地生豚價格變化에 대하여 小賣精肉價格의 反應에 관한 硏究

박정기 成均館大學校 1991 국내석사

製品의 評價에 영향을 미치는 價格,原産地 그리고 本質的 特性

박정균 漢陽大學校 1988 국내석사

우리나라 主要 畜産物의 産地價格 및 飼育頭數豫測에 關한 硏究

양병태 成均館大學校 1993 국내석사

人蔘의 産地流通構造와 價格形成에 關한 硏究 : 錦山地域 事例分析을 中心으로

전현선 高麗大學校 1986 국내석사

부동산지수 활용 방안의 효과성에 관한 연구 : 주택가격지수 연동금리 모기지를 중심으로

김경인 건국대학교 대학원 2010 국내석사