| It is emphasized in the 20 th National Congress of the Communist Party of China that "employment is the most fundamental matter of livelihood." This indicates that employment not only has the important economic significance,but also deeply affects social relations.The unemployment rate,as an important indicator of employment conditions,is always a hot research topic in academic society.In 2020,the rate of Anhui Province’s registered urban unemployment ranked 24 th out of 31 provinces and regions,lying at a medium level and possessing the representative.This article takes Anhui Province as an example to estimate its actual unemployment rate,and explores the relationship between the estimated unemployment rate and its influencing factors using a combination forecasting model,variable screening model,and additive partial linear model.The research content and the main conclusions are:(1)In terms of estimating the unemployment rate,in response to the provincial characteristics of Anhui Province,the hidden unemployment rate in rural areas was taken into consideration in the research.The relevant data from 1995 to 2020 were selected to estimate the total unemployment rate,which includes both the actual urban unemployment rate and the hidden rural unemployment rate.The actual urban unemployment rate was estimated using an adjusted coefficient method based on economic and population improvements,while the hidden rural unemployment rate was estimated using the calculation method of the rural labor force flow proposed by Wang Cheng in of the Chinese Academy of Social Sciences.The estimated results show that the total unemployment rate in Anhui Province has decreased by three-quarters from 28.33% in 1995 to 7.62% in 2020.In order to improve the reliability of empirical analysis,the estimated unemployment rate was compared with the unemployment rate data from the population census,and the comparison results showed that the estimated unemployment rate data agrees with the actual employment situation in Anhui Province better than the registered urban unemployment rate.(2)Based on the estimated real unemployment rate in Anhui Province,a time series model is established for a short-term prediction;Fifteen relevant indicators are selected for the linear relationship test,and seven variables with strong linear relationship with unemployment rate are selected,and the ordinary least squares regression model of unemployment rate is established through stepwise regression;The results of comparing the two models show that the time series model requires less data and the multiple linear regression model has higher prediction accuracy.Therefore,the generalized weighted average grey correlation combination prediction model is introduced to aggregate the advantages of the two benchmark models.(3)Three additive partial linear models with penalty terms,Lasso,adaptive Lasso,elastic network are reviewed in terms of formula,application,advantages and disadvantages.In view of the shortcomings of the benchmark model,three variable screening methods are used to screen multiple indicators,analyze and compare the mean square error of the three,and the adaptive Lasso is the smallest,or its screening effect is the best.The additive partial linear model is established with the screened indicators to clarify the impact of each effective indicator on the unemployment rate.It is showed that in the additive partial linear model,the impact of consumer price index,the growth rate of fiscal revenue and the average wage index of employees on the unemployment rate is significantly nonlinear,that is,the impact of the same factor on the unemployment rate is significantly different in different intervals;The additive partial linear model can be used effectively solve the problem of partial coefficient inconsistency in the OLS model,and its evaluation indicators are better than the OLS model,and the model reflects the actual situation in a better way. |
PDF Full Text Request