Determinants and Pathways for Inclusive Growth in China: Investigation Based on Artificial Intelligence (AI) Algorithm

Abstract

Can artificial intelligence (AI) algorithms help policymakers in their decisions on inclusive growth? In this study, we introduce artificial intelligence algorithms to calibrate China's inclusive growth determinants. We uncover various factors that significantly influence inclusive growth using machine learning forecasts. Furthermore, our results using best practice methods outperform findings from traditional regression-based strategies on other dimensions, which miss non-linear interactions in their estimations. However, we observe that when the actual value of the inclusive growth index is too large, the accuracy of the machine learning model is diminished. Meanwhile, the results of heterogeneity analysis reveal that the determinants of inclusive growth in cities with different region and different marketization level are distinct. In addition, we adopt the scenario simulation and prediction approach to reveal the best policy measure to promote inclusive growth in China. Our findings indicate that machine learning holds promise for understanding how inclusive growth can be achieved and can assist real-world economies in enhancing inclusive growth.

Appendices

Appendix A: The Indicator System of Inclusive Growth

Dimension	Sub-dimension	Basic indicator	Direction	Unit
Economic growth	Economic output	GDP per capita	+	Yuan
	Economic efficiency	Labor productivity	+	%
		Capital productivity	+	%
	Economic resilience	Economic resilience index	+	–
Equality of opportunity	Opportunity of employment	Employment in the secondary and tertiary industries	+	%
		Registered urban unemployment rate	-	%
	Opportunity of employment	Investment intensity of education funds	+	%
		Teacher/student ratio in primary schools	+	–
		The ratio of teachers to students in secondary schools	+	–
	Opportunity of education	Health technicians per thousand population	+	People
		Number of beds in health institutions per 1,000 population	+	Piece
	Opportunity of medical	Endowment insurance participation rate	+	%
		Health insurance participation rate	+	%
		Unemployment insurance participation rate	+	%
Achievement sharing	The income gap between urban–rural	Per capita net income of rural residents	+	Yuan
		Per capita disposable income of urban residents	+	Yuan
		Urban–rural income ratio	-	–
	Regional income distribution	Regional income gap	-	–
	Economic output sharing	Per capita income to GDP ratio	+	%

1. “ +” represents a positive indicator, while “−“ indicates a negative indicator.

Appendix B: Definition of Acronyms and Variables

Acronyms	Definition
${\text{S}}$	The training set of machine learning models, which is mainly used to train models
${\text{N}}$	The number of features of a training set in a machine learning model
${\text{Q}}$	The total number of samples in the original data set
${\text{M}}$	The number of candidate features in a new feature set when constructing a decision tree
${\text{B}}$	The number of training subsets extracted from the original sample by placing back sampling
${\text{Y}}$	The average of prediction results of each decision tree in the random forest model
DFI	Digital financial index
IS	Industrial structure
FD	Traditional financial development
GI	Government intervention
URB	Urbanization level
FDI	Foreign direct investment
TRS	Transportation level
MAE	The mean absolute error
MAPE	The mean absolute percentage error
MSE	The mean square error
RMSE	The root mean square error
R2	The goodness of fit