互联网搜索行为能帮助我们预测宏观经济吗? 阅读全文
下载全文 |
Title | Can Internet Search Behavior Help to Forecast the Macro Economy?
|
作者 | 刘涛雄 徐晓飞 范方达 |
Author | Liu Taoxiong, Xu Xiaofei and Fan Fangda |
作者单位 | 清华大学经济学研究所 |
Organization | Institute of Economics, Tsinghua University |
作者Email | liutx@tsinghua.edu.cn,xuxiaofei80@163.com |
中文关键词 | 宏观经济 互联网搜索行为 预测 “两步法” |
Key Words | Internet Search Behavior; Macro Economy; Forecast; “Two-Step Method” |
内容提要 | 在大数据被广泛应用的今天,能否以及如何利用大数据对宏观经济进行预测成为经济学研究的一个新领域,应用大数据对宏观经济总量进行预测迄今还鲜有研究。在宏观经济分析中,两种类型的数据可以被应用,即结构化数据和非结构化信息。政府统计指标属于结构化数据,而互联网搜索行为则属于非结构化信息。互联网搜索行为是在线大数据中较有代表性的信息,本文将探索互联网搜索行为能否帮助我们对宏观经济总量进行预测,分析利用结构化数据和非结构化信息的可行方法。本研究使用6种模型对宏观经济总量进行预测。通过不同模型的比较,最终选择最优预测模型。研究表明,互联网搜索行为可以帮助预测宏观经济,但必须依赖适当模型选择方法。搜索行为数据不是对现有统计数据的替代,而是补充。选择结构化数据与非结构化信息变量的正确方法是“两步法”。首先,仅使用政府统计信息选择初步最优预测模型;其次,将互联网搜索行为加入选择的模型中,最终确定最优模型。 |
Abstract | As big data are widely used today, whether and how to use big data in macroeconomic forecast has become a new field of economic research. In macro economy analysis, two types of data can be applied, namely the structured data and unstructured information. Government statistics are well structured while Internet search behavior information belongs to non-structured information. This research uses 6 types of models to forecast the macroeconomic aggregate. By comparing different models, the optimal forecast model is selected. We find that the Internet search behavior can help forecast the macro economy, but only conditionally on the appropriate way of model selection. We find that the correct way for variables selection with structured and unstructured information is the two-step method. Firstly only the government statistics are used and the conditional optimal model is selected. Secondly, the internet search data are added into these models and the optimal model is determined. |
文章编号 | WP912 |
登载时间 | 2015-08-14 |
|