A New Approach for Estimating the Policy Positions of Parties and Districts Based on the Spatial Voting Theory‥‥Theoretical and Empirical Analysis

我指出，选举中政治行为者的经验主义的唐氏政策位置自然是由政党（候选人）获得的选票数量决定的。我开发了一个系统来执行这个计算。我将其应用于日本和英国（即英国的一部分）的全国选举。结果表明，唐氏主义意义上的选举主要议题分别是日本和英国的城乡矛盾和失业问题。以日本为例，如果采用唐斯式的框架，就会发现意识形态并不是选举的主要因素。然而，在我的方法中，有几个技术问题。首先，需要对高维空间中的函数进行全局优化，因此必须非常小心地得到正确的全局最优解。在这项研究中，我成功地在一个可接受的计算时间内得到了这个问题的看似正确的解决方案。其次，在我的方法中需要对每个区域的密度形状进行先验假设。作为第一个近似，我用正态密度。这里的一个自然问题是，结果是否对选择敏感。我检查了一下，如果我们改变t分布，而不是正态分布，结果会保持不变。第三，我们不使用有关各方政策的事先信息。因此，各方的顺序可能会随着时间的推移而不稳定地变化，这是由噪音或唐斯机制以外的因素引起的。我发现结果在日本和英国都很稳定。因此，我的方法现在实际上适用于全国选举。在研究过程中，我遇到了选举数据缺乏的问题。于是我安排了一个对公众开放的数据库。人们可以通过互联网访问这些数据。这是本研究的另一个贡献。

I point out that the empirical Downsian policy locations of political actors in an election are naturally determined from the number of votes gained by the parties (candidates). I developed a system for carrying out this calculation. I applied it to the national elections of Japan and England (i.e., a part of the United Kingdom.). The results show that the main issues of elections in the Downsian sense are urban=rural conflicts and the jobless problem, respectively, in Japan and in England. In the case of Japan, the result shows that the ideology is not the main factor of the election if we take the Downsian frame.In my approach, however, there are several technical problems.First, global optimization of a function in a high dimensional space is needed, so that one must be very careful in getting the correct global optimum. In this research I succeeded in getting seemingly correct solution to this problem within a acceptable computation time.Second, an a priori assumption on the shape of density on each district is needed in my approach. As a first approximation, I use the normal density. A natural question here is whether the result is sensitive to the choice or not. I checked that the results remains almost as it is if we change the t-distribution instead of the normal.Third, we use no prior information on the policy of parties. Thus it is possible that the order of parties changes unstable over time caused by noises or factors other than the Downsian mechanisms. I found that the results are very stable over time both in Japan and in England.Thus my approach is now practically applicable to national elections.In the process of research, I encountered a trouble of the lack of election data. Thus I arranged the database which is open to public. One can access the data through internet. It is another contribution of this research.