부동산정책에 따른 투자심리와 주택가격 변화에 관한 연구 : 자연어처리를 활용한 온라인 커뮤니티 분석

부동산정책이 투자심리와 주택가격 변화에 미치는 영향을 분석했다.

Recently, housing prices in Korea have risen sharply. In response, President Moon Jae-in and his administration announced approximately 28 real estate policies one after another, starting with the June 19 measures in 2017 and ending with the October 26 measures in 2021. This study aims to examine the effect of stabilizing housing prices among the various effect expected by the government announcing the real estate policies. In particular, from the perspective of behavioral economics, in which housing prices can change depending on investor sentiment, we will examine the stabilizing effect of real estate policies on housing prices.

However, it is not easy to measure investor sentiment, and the dynamic relationship between investor sentiment and housing price or between investor sentiment and real estate policy is different depending on the study. Since the online community texts contain thoughts spontaneously expressed by many people, it will be possible to measure various dimensions of investment sentiment by analyzing them with Natural Language Processing methods. The purpose of this study is divided into three main parts. The first is to derive investor sentiment on various topics from online community texts and verify whether each investor sentiment explains housing prices well.

The second is to analyze the dynamic relationship between investor sentiment and real estate policy for each theme, and examine whether investor sentiment changes with real estate policy. Finally, we will see if real estate policy interventions that lower investment sentiment among online community users during periods of rising housing prices will have a stabilizing effect on housing prices. The main research results are as follows.

First, six interest indices("Future Price Expectations and Market Conditions", "Housing Desires", "Buying and Selling Recommendations", "Reconstruction and Remodeling", "Market Trends", "Politics") and three sentiment indices("Future Price Expectations and Market Conditions", "Buying and Selling Recommendations", "Politics") that show a two-way Granger casuality with housing prices were derived from real estate-related online communities as investment sentiment indices.

Among the nine investment sentiments, two interest indices (“Future Price Expectations and Market Conditions" and "Buying and Selling Recommendations") and one sentiment index (“"Buying and Selling Recommendations") were found to explain housing prices well. In particular, these three investor sentiment indices were found to explain house prices more than macroeconomic variables, even in VAR models that consider macroeconomic variables. We categorized the types of real estate policies into development regulation policies, tax system policies, financial regulation policies, and transaction regulation policies.

In the case of development regulation policies, housing prices barely reacted, while in the case of tax system policies and financial regulation policies, housing prices responded in a slightly downward direction. Even in the case of the transaction regulation policy, there were some positive and negative effects, indicating that it is difficult to expect changes in housing prices. In other words, we found that the effect of real estate policies on stabilizing housing prices is generally small.

As a result of the Granger causality test between real estate policy and investor sentiment, the tax system policy appears to be highly related to the interest index(“Future Price Expectations and Market Conditions" and "Buying and Selling Recommendations") and the sentiment index("Buying and Selling Recommendations"), which are highly related to housing prices. No association with psychological indices was found. When the impact of the tax system policy appeared, there was a reaction that 'Future Price Expectations and Market Conditions', an index of interest, dropped sharply.

In the VAR model, which considers the investor sentiment index, housing price, and real estate policies, it was found that the reaction of the investor sentiment index to the real estate policy shock changed the reaction of the housing price to the real estate policy shock. The impact of the tax system policy did not significantly reduce the ‘Buying and Selling Recommendations’', which is an interest index, and the ‘Buying and Selling Recommendations’, which is an sentiment index.

However, the shock of the tax system policy caused a relatively large decline in the ‘Market Conditions’, which is an interest index, and the ‘Future Price Expectations and Market Conditions’, which is an sentiment index. As a result, the impact of the tax system policy has resulted in a relatively large drop in housing prices. The academic implications of this study are: First, from the perspective of behavioral economics, we demonstrated the role of investor sentiment in the emergence of a bubble related to housing prices. Specifically, through text analysis, we demonstrated the relationship between online communities and house prices.

It was revealed that the housing price stabilization effect appears when investor sentiment declines due to the impact of real estate policies. Therefore, when housing prices rise sharply, strengthening the tax system can be expected to have the effect of stabilizing housing prices. The methodological implications are: First, a sentence-by-sentence text analysis methodology was applied. We also segmented the dimensions of investor sentiment using a deep learning-based language model. The policy implications are as follows. In order to achieve the goals of the policy, it is necessary to design and announce the content of the policy in consideration of investor sentiment.

Among the interest index and sentiment index constructed to measure investor sentiment in this study, it was shown that the interest index explains housing prices well, but frequent policy announcements and interviews attract people's attention. It is necessary to ponder