The Impact Of News Sentiment On Wine Prices

With the gradual improvement of the world financial market system,the diversification of investment has continuously expanded the scope of investment.Wine,which has the characteristics of risk and return ratio between equity investment and risk-free assets,has gradually entered the vision of investors and became an increasingly popular alternative investment.With the growing popularity of wine as an alternative investment,more and more scholars have studied the pricing of wine investment products from various aspects.However,most studies use traditional methods similar to stock pricing to price wine,and only some scholars have discussed the factors that affect the price of wine from the perspective of behavioral finance.At present,the more popular pricing factors for wine mainly include the weather conditions of the wine producing area,the wine review rating,and the wine awards.Considering that news sentiment can be used for the pricing of assets such as equity and gold,and few studies use news sentiment to price wine.So this article will study the relationship between news sentiment and wine prices.The main research question is whether news sentiment will affect the price of wine.Natural language processing is an important direction of computer science and a branch of artificial intelligence.It has been frequently used in the economic field in recent years.As a method of natural language processing,the sentiment dictionary matching algorithm,as one of the most mature methods in news sentiment judgment,is very suitable for the research of this paper.The basic principle of the sentiment dictionary matching algorithm is to establish sentiment dictionaries containing positive words,negative words,and degree adverbs respectively,and match the words in the news with the words in the sentiment dictionaries,and use a specific algorithm to measure the news’ s sentiment trend judge.The news data to be used in this article is from Google News.I collect news related to wine through a Python-based web crawler method.I use "wine" as the searching keyword to collect the titles,sources,and publishing dates of all news from January 1,2014 to May 31,2019.The data used in this article to reflect the price of red wine comes from the Liv-ex 100 Index launched by the London International Wine Exchange as the price of wine.The global macroeconomic data used in this article is the MSCI ACWI Index from Morgan Stanley Capital International.For wine price data and global macroeconomic data,I directly calculated the logarithmic returns on the monthly data of the Liv-ex 100 index and the MSCI ACWI index.I used time series data for analysis,the data range from 01/04/2014 to 30/04/2019,61 months in total.For news data,I first established three sentiment dictionaries: positive dictionary,negative dictionary,and degree adverb dictionary.After that,I preprocessed the news title corpus in three steps: Word Tokenization,Removing Stopwords,Stemming & Lemmatization.During Word Tokenization,I took the sentence apart into separate words,during Removing Stopwords,I removed words and punctuations which are meaningless to the whole sentence,during Stemming & Lemmatization,I transferred plural form nouns to singular nouns,verbs with tense to the original form,etc.,and converted uppercase to lowercase.After preprocessing the news corpus,I run the sentiment dictionary matching algorithm to calculate the sentiment score of each piece of news.The sentiment score of each news can be any positive,0 or negative number.The larger the value,the more positive the emotional trend of the sentence;the smaller the value,the more negative the emotional trend of the sentence;the value 0,which indicates that the news is neutral.Finally,I manually checked the matching results to discover errors in the matching process and make corrections to the sentiment dictionary.In the end,the final versions of the three sentiment dictionaries are a positive dictionary containing 2,167 positive words,a negative dictionary containing 6,294 negative words,a degree adverb dictionary containing 71 degree adverbs with their corresponding emotional degree multipliers.Finally,I calculated the arithmetic mean of the sentiment scores of all news corpora for each month.This article selects 60 months of data from 01/04/2014 to 31/03/2019 as news sentiment score data for subsequent regression analysis.In the regression analysis,I used the return of the wine price index for the current month as the explanatory variable of the model,the news sentiment score of the current month and the previous month,and the return of the wine price index for the previous month as the explanatory variables of the model.The rate was regressed as a control variable,and it was found that positive news sentiment will cause the price of wine to rise in the month,but because investors will overreact,the increase will be higher than it could be.In the next month,the price of wine will tend to reverse.And because of "momentum effect" reasons,the increase in the price of wine in the previous month will also make the price of wine this month tend to increase.The news sentiment also contains information about the global macroeconomy.After that,through the robustness test,the model used in this paper does not have the problem of heteroscedasticity or autoregression.In the design of the research method,I have used some methods such as modeling in differential form,adding the global economy index as a control variable and modifying the sentiment dictionaries to reduce the influence of endogeneity.This article studies the relationship between news sentiment and the price of wine,enriches the pricing method of wine,also studies the effectiveness of the red wine market,and also proposes some directions for further research.The research on the investment pricing of red wine and the influence of news sentiment on investment products has some enlightening effects.