The working poor and poverty reduction in stable macroeconomic environments.

日本没有官方的“贫困线”，因此没有关于贫困的官方统计数据(Sekine，2008)。然而，在实践中，经济学家们使用OECD指数(收入中位数的50%)，或“最低生活水平”方法来评估日本的贫困状况。OECD指数评估日本的相对贫困率为15.3%，在OECD国家中排名第五。然而，日本政府表示，根据绝对测量，不存在严重的贫困问题(Yuasa，2006)。然而，有一些令人担忧的迹象表明，低收入人口迅速增长，同时社会保障和公共援助等现有安全网在个人范围内存在很大差距(Kawai和Kanno，2006)。关于日本和其他发达经济体之间关于贫困劳动者的比较分析的学术文献很少。村上描述了日本的穷人，并将他们与加拿大(村上，2011)和英国(村上，2009)进行了比较。研究表明，贫困劳动者最有可能出现在年轻人中，他们的教育水平较低，就业不稳定，为小公司工作。就主要差异而言，在西方国家，男性更有可能属于贫困劳动人口，与英国和加拿大相比，日本25-44岁年龄段的贫困劳动人口数量明显较多。这项研究的目的是改进村上关于绝对贫困的统计分析的几个缺陷。(1)村上(2011)修订了贫困劳动者的定义(岩井和村上，2007)，并纳入了美国劳工统计局(BLS)提供的定义。然而，美国用贫困线来定义贫困，这在日本是不存在的。这种对定义的不一致使用导致了各国之间无与伦比的结果。(2)村上春树的数据基于1997年和2002年的劳工调查。他对数据强加了一个非常简单的统计模型，比较分析是基于简单的百分比计算。因此，他的结果的稳健性和估计的意义没有被调查。(3)数据集没有提供必要的结构来研究贫困劳动者的特征在群体之间和群体内的动态影响。我们对问题(1)的方法是回到贫困劳动者的原始定义(Klein，Bruce和Rones，1989)，并建立一个更严格的度量函数，允许区域内生活费用的变化，从而更好地衡量最低生活水平。此外，我们将总结欧洲所有主要发达经济体对贫困劳动者的所有定义，以便建立工作时数和贫困“线”定义之间的一般关系。这是第一次试图建立这种关系的研究。瑞士模型(Sekiguchi和Sato，2014)是一个非常好的观点，它定义了工作贫困，而不是就业小时数。然而，该模型的弱点在于它对贫困线的具体定义。我们试图在下面的第(3)项下解决这个问题。我们对问题(2)的做法是考虑一个相当大的面板数据集，包括至少5波劳工调查，并对这些数据应用先进的统计模型。这将使我们能够衡量我们估计的统计意义，并建立我们模型的稳健性。(3)我们数据集结构的丰富性使我们能够应用复杂的机器学习方法，如无监督方法(Casella，2015)，以提取更一般的“贫困线”定义。正是这种新的方法将导致在确定绝对贫困和相对贫困的一般度量函数方面取得重大的新结果。

Japan has no official "poverty line", and subsequently no official statistics on poverty exists (Sekine, 2008). In practice, however, economists have applied the OECD index (50% of median income), or the "minimum standard of living" approach in assessing Japan's poverty. The OECD index assesses Japan's relative poverty rate to be 15.3%, the fifth highest ratio amongst the OECD countries. The Japanese government, however, states that according to absolute measurements, there are no severe poverty issues (Yuasa, 2006). However, there are worrying sings of a fast-growing low-income population accompanied by the fact that current safety nets such as social security and public assistance have large gaps in their personal scope (Kawai and Kanno, 2006). The academic literature on comparative analysis of the working poor between Japan and other advanced economies is sparse. Murakami characterises the working poor in Japan and compares them with those in Canada (Murakami, 2011), and the UK (Murakami, 2009). The studies show that the working poor are most likely to be found among the young, those with a low level of education, unstable employment, and working for small companies. As regards to the main differences, male are more likely to belong to the working poor in western countries, and the number of working poor in the age group 25-44 is significantly lager in Japan compared to the UK and Canada. The aim of this research is to improve on several drawbacks of Murakami's statistical analysis of absolute poverty measurement. (1) Murakami (2011) revises the definition of the working poor (Iwai and Murakami, 2007) and incorporates the definition provided by the US Bureau of Labour Statistics (BLS). However, the US defines poverty by a poverty line, which does not exist in Japan. This inconsistent use of definitions leads to incomparable results across countries. (2) Murakami's data set is based on the labour survey 1997 and 2002. He imposes a very simple statistical model on the data, and comparative analysis is based on simple percentage calculations. Hence, the robustness of his results, and the significance of the estimates are not investigated. (3) The data set does not provide the necessary structure to study the dynamic effects of the characteristics of the working poor between and within groups. Our approach to problem (1) is to return to the original definition of the working poor (Klein, Bruce and Rones, 1989) and to establish a more rigorous metric function, which allows for intra regional cost of living variations and therewith a better measurement of the minimum living standards. Moreover, we will summarize all definitions of the working poor in all major advanced European economies in order to establish general relationships between number of hours of work, and definitions of poverty "line". This is the first study attempting to establish this relationship. The Swiss model (Sekiguchi and Sato, 2014) is a very good stating point, which defines working poor independent of the number of hours in employment. However, the model's weakness is its very country specific definition of the poverty line. We attempt to tackle that issue under (3) below. Our approach to problem (2) is to consider a much large panel data set including at least 5 waves of the labour survey and to impose advanced statistical models on the data. This will allows us to measure the statistical significance of our estimates and to establish the robustness of our model. (3) The richness of the structure of our data set allows us to apply sophisticated methods of machine learning such as unsupervised methods (Casella, 2015) to extract more general definitions of "poverty lines". It is precisely, this novel approach that will lead to significant new results in determining general metric functions in the measurement of absolute and relative poverty.