1、Policy Research Working Paper10930Disaggregated Impacts of Growth on Multidimensional PovertyDoes the Source of Growth Matter?Francis MulanguMokhtar BenlamineMichael KellerJean-Pascal NganouPoverty and Equity Global Practice September 2024 Public Disclosure AuthorizedPublic Disclosure AuthorizedPubl

2、ic Disclosure AuthorizedPublic Disclosure AuthorizedProduced by the Research Support TeamAbstractThe Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues.An objective of the series is to get the findings out qu

3、ickly,even if the presentations are less than fully polished.The papers carry the names of the authors and should be cited accordingly.The findings,interpretations,and conclusions expressed in this paper are entirely those of the authors.They do not necessarily represent the views of the Internation

4、al Bank for Reconstruction and Development/World Bank and its affiliated organizations,or those of the Executive Directors of the World Bank or the governments they represent.Policy Research Working Paper 10930This paper presents comprehensive findings on the rela-tionship between economic growth an

5、d poverty.Using a first-difference model applied to data from more than 80 countries spanning over 20 years,the paper investigates how changes in gross domestic product affect the Multidimen-sional Poverty Index and its subcomponents,considering variations in income level,region,and resource depende

6、ncy.The analysis confirms that economic growth generally reduces the Multidimensional Poverty Index,although the magnitude of the effect varies significantly.It is less pronounced in low-income countries,Sub-Saharan Africa,Latin America and the Caribbean,and resource-dependent countries.The paper di

7、saggregates gross domestic product growth by its dimensions,revealing that growth driven by total factor productivity,consumption,and sustain-able growth significantly decreases the Multidimensional Poverty Index.In contrast,factors such as human capital development,capital deepening,investment,gove

8、rnment spending,exports,and imports show ambiguous effects on the Multidimensional Poverty Index.These findings suggest that the effectiveness of these factors depends on country-level conditions.Given the clearer positive impact of total factor productivity,consumption,and sustainable growth on red

9、ucing multidimensional poverty,policy makers should prioritize strategies that promote these types of growth to fight poverty,especially in contexts where the effects of other growth contributors are uncertain or not well understood.This paper is a product of the Poverty and Equity Global Practice.I

10、t is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world.Policy Research Working Papers are also posted on the Web at http:/www.worldbank.org/prwp.The authors may be contacted at fmulanguworldbank

11、.org.1 Disaggregated Impacts of Growth on Multidimensional Poverty:Does the Source ofGrowth Matter?1 Francis Mulangu,Mokhtar Benlamine,Michael Keller,and Jean-Pascal Nganou Keywords:Multidimensional Poverty,Economic Growth,Sustainability,and Productivity JEL classification:I3,O1,Q54 2024 The World B

12、ank and International Monetary Fund.1 The authors would like to extend their heartfelt gratitude to Daniel Mahler,Aly Sanoh,Steve Pennings,and Clarence Tsimpo Nkengne for their invaluable feedback and insightful comments on earlier versions of this paper.Their contributions have significantly enhanc

13、ed the quality of this work.However,the views expressed in this paper are those of the authors and do not necessarily reflect those of the World Bank.2 1.Introduction The discourse on whether and how economic growth translates into poverty reduction is a topic of significant importance and enduring

14、debate in economic development.This debate extends far beyond academic circles,touching the core of policy making.A critical challenge is the ability to offer concrete guidance on prioritizing growth-oriented policies and identifying the specific nature of growth that effectively mitigates poverty.T

15、his ongoing discourse underscores the need for rigorous analysis and evidence to inform policy choices,particularly when considering the multiple facets of poverty beyond mere income metrics to encompass a broader range of deprivations affecting individual well-being.While a substantial body of empi

16、rical evidence exists pointing to a generally negative correlation between economic growth and poverty,it is noteworthy that the lions share of this literature has traditionally focused on income or monetary poverty.This narrow focus on money leaves a significant gap in our understanding of the dyna

17、mics between economic growth and multidimensional povertya form of poverty that considers various dimensions of deprivation,including education,health,and living standards(Alkire et al.,2021).The relationship between multidimensional poverty and economic growth remains relatively underexplored,with

18、most studies merely qualitatively tracing the trajectories of GDP growth and multidimensional poverty at the national level without engaging in formal statistical analysis(e.g.,Djossou et al.(2017),Tran et al.(2015).A handful of studies have ventured into examining the correlation between changes in

19、 GDP and shifts in composite indices of multidimensional poverty using cross-sectional data across different time points(Alkire et al.,2017;Burchi et al.,2019;Seth&Alkire,2021).However,rigorous assessments of the relationship between economic growth and multidimensional poverty are scarce.To our kno

20、wledge,only two attempts have been made in this direction,notably by Santos et al.(2019)and Balasubramanian et al.(2023).Both employed unbalanced panel data and a first-difference estimator(FDE)to uncover a negative correlation,albeit focusing on different indices of multidimensional poverty and stu

21、dy periods.Our paper endeavors to contribute meaningfully to this relatively uncharted but essential literature.We begin our investigation by constructing a panel dataset that integrates Multidimensional Poverty Index(MPI)data with indicators of economic growth.Following the methodological footsteps

22、 of Santos et al.(2019)and Balasubramanian et al.(2023),we employ a first-difference model to elucidate the fundamental relationship between MPI and economic growth.We begin the analysis by exploring the fundamental relationship between the MPI and economic growth,focusing on variations across diffe

23、rent income groups and regions.This initial analysis replicates existing findings from the literature,providing a solid foundation for our research.Building upon this,our study contributes significantly to the literature by examining the sub-components of MPInamely education,health,and living standa

24、rdsto identify any heterogeneity within these subcomponents.Furthermore,we investigate economic growth across various dimensions and components to provide deeper insights and distinguish our work from previous studies.In the second part of our study,we analyze the effects of economic growth by 3 exa

25、mining factors such as labor productivity,growth of factors of production,and expenditure approach components,and introduce a novel method for measuring sustainable growth.In the first part of this study,we identify a negative relationship between economic growth and the MPI,confirming prior researc

26、h findings.Our analysis further indicates that the effects magnitude varies notably across different income levels,regions,and degrees of resource dependency.Specifically,the reduction in MPI is less pronounced in low-income countries,as well as in the regions of Sub-Saharan Africa and Latin America

27、 and the Caribbean,and resource-dependent countries.Additionally,we find that while economic growth generally reduces the MPI components,these effects are often delayed.In the second part of our research,we focus on different dimensions of economic growth and their impact on multidimensional poverty

28、.Our findings show that not all growth dimensions contribute equally to MPI reduction.Notably,growth driven by total factor productivity(TFP),consumption,and sustainable practices consistently leads to significant declines in MPI,making these areas particularly effective in poverty reduction.Convers

29、ely,factors like human capital development,capital deepening,investment,government spending,exports,and imports show a statistically insignificant impact on MPI,suggesting that their effects may vary depending on country-specific conditions.The rest of the paper proceeds as follows:Section 2 discuss

30、es the relative importance of multidimensional poverty for informing policies.Sections 3 and 4 explain the data and estimation strategy used to analyze the impact of growth on multidimensional poverty,respectively.Section 5 presents the results,followed by robustness checks in section 6.Finally,sect

31、ion 7 concludes.4 2.Enhancing Policy Effectiveness through Multidimensional Poverty Measures Multidimensional poverty measures provide a broader perspective on human deprivation by encompassing various aspects such as health,education,and standard of living.Unlike monetary poverty measures that focu

32、s primarily on income,multidimensional indicators capture a wider range of deprivations,offering policy makers a more comprehensive understanding of the factors contributing to poverty.This approach allows for the design of integrated interventions that address the root causes of poverty across vari

33、ous dimensions.For example,policies aimed at reducing multidimensional poverty can simultaneously target issues such as access to health care,educational opportunities,and social inclusion,holistically tackling the interconnected aspects of poverty and improving overall well-being.Furthermore,multid

34、imensional poverty measures are particularly valuable for capturing the experiences of vulnerable and marginalized groups,who may face various forms of deprivation not fully represented by income-based measures alone.For instance,women,children,older people,and people with disabilities often face mu

35、ltidimensional forms of deprivation that income-based measures struggle to capture(Alkire and Foster,2011;UNDP,2019;World Bank,2011;UNICEF,2016).By incorporating dimensions that,one way or another,touch gender inequality,intergenerational poverty,and social exclusion,multidimensional poverty indices

36、 provide a more inclusive representation of poverty,ensuring that the most vulnerable populations are not overlooked in policy discussions and resource allocation.Research in development economics has highlighted that income-based poverty measures may not fully capture the complexity of human depriv

37、ation.Notably,Amartya Sens work underscored the importance of considering multiple dimensions of well-beingsuch as health,education,and social inclusionalongside income when assessing poverty(Sen,1999).This broader perspective has led to the development of the MPI,which measures poverty across these

38、 various dimensions(Alkire,2007).Studies using MPI suggest that traditional income-based poverty measures might underestimate the extent of poverty and overlook disparities in deprivation across different dimensions(Alkire et al.,2015).Moreover,empirical evidence indicates that economic growth does

39、not uniformly lead to reductions in monetary poverty,which underscores the value of a multidimensional approach to poverty measurement(Alkire et al.,2013).For instance,research by Cruz et al.(2015)demonstrated that economic growth can sometimes disproportionately benefit specific groups or regions,l

40、eaving others behind and leading to persistent monetary poverty.By considering multiple dimensions of deprivation,multidimensional poverty measures provide a more nuanced understanding of poverty dynamics.This allows policy makers to tailor interventions more effectively to meet the distinct needs o

41、f various population groups(Alkire et al.,2015).Studies have demonstrated that multidimensional poverty measures are particularly effective in identifying vulnerable and marginalized populations whom income-based measures may not fully represent.For example,Alkire&Santos(2013)found that women and ch

42、ildren are often disproportionately affected by multidimensional poverty,especially in access to education and health care.Similarly,Palmer et al.(2015)highlighted that people with disabilities face specific challenges that income-based measures may not capture,emphasizing the need to consider disab

43、ility as a dimension of poverty.These findings support multidimensional measures as a complementary tool to provide a more comprehensive and inclusive poverty assessment.5 The effect of economic growth on monetary poverty,especially in Sub-Saharan Africa,has been relatively modest(Wu et al.,2024).De

44、spite efforts to address initial disparities in poverty levels,income,and inequality,the elasticity of poverty reduction to growth has remained low in this region.The limited pace of poverty reduction since 2000 has often been attributed to economic growth rather than major changes in income distrib

45、ution.This situation has raised concerns about the effectiveness of growth-centric development strategies in significantly alleviating poverty(Aryeetey,2015;McKey and Thorbecke,2015).However,focusing solely on monetary poverty may overlook the broader benefits that economic growth can provide,partic

46、ularly in areas such as health care,education,and infrastructure.While income levels are a critical component of welfare,economic growth can also lead to improved access to essential services and opportunities that enhance overall well-being.These non-monetary benefits are particularly important in

47、regions starting from a low base,such as many Sub-Saharan African countries.Thus,a comprehensive approach that considers the various dimensions of poverty and the multiple pathways through which growth can enhance welfare is crucial.This paper contributes to the ongoing debate by highlighting the im

48、portance of integrating multidimensional poverty measures with traditional monetary metrics to provide a fuller picture of welfare and development.6 3.Data Multidimensional poverty index Our analysis uses the Multidimensional Poverty Index(MPI)and its indicators as dependent variables.The MPI,develo

49、ped by the Oxford Poverty and Human Development Initiative(OPHI),provides a nuanced understanding of global poverty beyond income-based metrics.It incorporates a range of indicators reflecting various deprivations that individuals face across different aspects of life.This multidimensional approach

50、aligns with the idea that poverty is not only about income,but also the inability to access essential services and enjoy a standard quality of life(Alkire et al.,2021).The MPI includes ten indicators grouped under three significant dimensions,as depicted in Table 1.Table 1:Dimensions,indicators,and

51、weights of the MPI Dimensions(Weight)Indicator(Weight)A person is deprived of living in a household where.Health(1/3)Nutrition(1/6)Any person under 70 years of age for whom nutritional information is available is undernourished.Child mortality(1/6)A child under 18 has died in the household in the fi

52、ve years preceding the survey.Education(1/3)Years of schooling(1/6)No eligible household member has completed six years of schooling.School attendance(1/6)Any school-aged child is not attending school up to the age at which he/she would complete class 8.Living Standards(1/3)Cooking fuel(1/18)A house

53、hold cooks using solid fuel,such as dung,crops,shrubs,wood,charcoal,or coal.Sanitation(1/18)The household has unimproved or no sanitation facility,or it is improved but shared with other households.Drinking water(1/18)The households source of drinking water is not safe,or safe drinking water is a 30

54、-minute or longer walk from home,roundtrip.Electricity(1/18)The household has no electricity.Housing(1/18)The household has inadequate housing materials in any of the three components:floor,roof,or walls.Assets(1/18)The household does not own more than one of these assets:radio,TV,telephone,computer

55、,animal cart,bicycle,motorbike,or refrigerator,and does not own a car or truck.Source:Alkire et al.(2021)To identify multidimensional poverty,each household is assessed against the ten indicators.For each indicator,if a household fails to meet the minimum standard,it is considered deprived.Each indi

56、cator is weighted equally within its dimension,and the dimensions have specific weights.A household is considered multidimensionally poor if the cumulative weighted score of its deprivations is 1/3 or higher.This threshold indicates that a household is deprived in at least one-third of the weighted

57、indicators.The MPI value is calculated by multiplying the proportion of people identified as multidimensionally poor(incidence)by the average proportion of deprivations experienced by this poor group(intensity)(Alkire et al.,2021).While the MPI captures both the incidence and intensity of poverty,it

58、 does not measure the depth of poverty,unlike the adjusted poverty gap.7 Other variables The independent variables encompass GDP per capita,labor productivity per worker,and several distinct methodologies for disaggregating GDP and labor productivity.GDP per capita data,measured in US dollars at con

59、stant 2015 prices,was sourced from the World Development Indicators(World Bank,2024).Productivity variables are drawn from the Cross-Country Database of Productivity developed by Dieppe(2021).The first productivity variable is labor productivity,defined as the amount of output produced per worker an

60、d calculated as GDP divided by the number of employed people.Labor productivity growth is further decomposed into three components:human capital growth,capital deepening growth,and total factor productivity(TFP)growth.Human capital growth measures how improvements in worker skills and capabilities t

61、hrough education,training,and health improvements increase productivity.Capital deepening measures changes in the capital-to-worker ratio,and TFP measures the efficiency with which factor inputs are combined.TFP often serves as a proxy for technological progress in growth accounting exercises(Dieppe

62、,2021).We further disentangled the components of economic growth from an expenditure perspective and calculated per capita values for consumption,(public and private)investment,government spending,exports,and imports measured in constant 2015 U.S.dollars.These measures were derived from the World De

63、velopment Indicators database.Moreover,we delineated economic growth into sustainable and unsustainable components,adopting the methodology outlined by Mahler(2021).The unsustainable part of national income was quantified as the aggregate of natural resource depletion,the economic costs attributable

64、 to CO2 and particulate emissions,and the consumption of fixed capital.Conversely,the sustainable portion was determined by subtracting these costs from the total national income,represented as Gross National Income(GNI).Data structure The data is organized into spells,as defined by Cox(2007).These

65、spells represent the intervals between two survey years,with the length of each spell determined by the time elapsed between the surveys.Within this framework,the dependent and most independent variables are computed as the average annual change in the logarithm of the variable for each spell,repres

66、enting the yearly(compound)proportional change.Santos et al.(2019)and Balasubramanian et al.(2023)argue that there is a time lag between changes in GDP per capita and variations in multidimensional poverty,suggesting that the impacts of economic growth on poverty do not manifest instantaneously but

67、rather unfold over time.Consequently,their multidimensional poverty estimate for a particular year is associated with the average GDP per capita over the five years preceding that year.For instance,the change in multidimensional poverty between 2010 and 2015 is linked to the average GDP per capita c

68、hange between 20052009 and 20102014.Our approach builds on this understanding by adjusting the data structure to account for the complexity of the MPI,which comprises ten subcomponents.This complexity leads us to propose that the timing of economic growths impact on the MPI likely varies across its

69、different components.For example,improvements in income from economic growth might immediately enhance nutrition levels as households gain the ability to purchase food.In contrast,effects on 8 educational attainment,such as years of schooling,may emerge more gradually as the benefits of increased in

70、come permeate societal structures.To explore these dynamics,we associate the change in multidimensional poverty with current and lagged GDP changes.For instance,the change in multidimensional poverty between 2012 and 2015 is linked to the change in GDP per capita during the same period,referred to a

71、s the current change in GDP.Additionally,we test for lagged impacts,where the change in GDP is defined as the change in the five years before the first survey or before the spell starts.In our example above,this would mean that the lagged effect is defined between 2007 and 2011.The lagged changes in

72、 GDP are consistently measured over the five years preceding the first survey.In contrast,the period for the current change in GDP varies depending on the interval between the surveys.Finally,we combine both current and lagged changes to assess the total(current+lagged)impact.By testing different sp

73、ecifications,we aim to discern whether the effects of economic growth on multidimensional poverty are immediate or if they occur with a delay.This will enhance our understanding of the intricate ways economic changes influence various aspects of poverty.Sample Our sample consists of 84 countries dis

74、tributed across six world regions following OPHIs region classification.Table 2 lists the six regions,details the number of countries and spells available for each region,and the first and last years for which we have MPI data.2 Sub-Saharan Africa has the most representation,with 36 countries and 60

75、 spells,whereas South Asia has the least representation,with five countries and seven spells.The number of observations per country ranges from two to four,and the length of each spell varies from one to twelve years,with an average spell length of five years.For further insights,Table A 2 in the ap

76、pendix provides descriptive statistics for both the dependent and independent variables used in our analysis.Table 2:Regions of the sample Regions Nr.of countries Nr.of spells First year Last year Arab States 9 10 2006 2020 East Asia and the Pacific 9 12 2010 2022 Europe and Central Asia 12 16 2005

77、2019 Latin America and the Caribbean 13 22 2001 2021 South Asia 5 7 2006 2021 Sub-Saharan Africa 36 60 2000 2021 Total 84 127 2000 2022 2 For a list of all countries by region refer to Table A 1 in the appendix.9 4.Estimation Strategy Following the approach of Balasubramanian et al.(2023)and Santos

78、et al.(2019),we adopted the first difference estimator(FDE)to investigate the impact of economic growth on multidimensional poverty.This method is particularly suited to address potential omitted variable biases by differencing out unobserved,time-invariant factors that might influence countries eco

79、nomic growth and poverty levels.Our methodology involves the construction of a panel dataset,focusing on the changes in the Multidimensional Poverty Index(MPI)and current and lagged changes in GDP per capita or productivity.The primary equations used in our analysis are as follow:,=+1,+,(1),=+1,+2,1

80、+,(2),=+1,1+,(3)In all three equations,stands for the change in the logarithm of the MPI or its components for country during period,capturing the annual proportional change in multidimensional poverty.The subscript denotes the period between two surveys,also known as the spell.The length of each sp

81、ell can vary depending on when the surveys were conducted.The subscript 1 refers to a lagged period,specifically the five years before the first survey in the spell.In general,log represents the change in the logarithm of GDP per capita or productivity.Specifically,in equation(1),covers the current

82、change in GDP per capita,where is the same period or spell as for the dependent variable.This measures the immediate impact of GDP per capita or productivity on poverty.In equation(2),we added,1,which measures the change in the logarithm of GDP per capita or productivity over a lagged five-year peri

83、od.This recognizes the temporal lag between economic performance and its potential impact on poverty.Thus,in equation(2),we estimate both the current and the lagged impact of GDP per capita on the MPI separately.Finally,in equation(3),we combine the current changes in GDP(,)and the lagged changes in

84、 GDP(,1)to calculate the total change in GDP(,1).Variable represents a constant term,1 and 2 are the coefficients of interest indicating the elasticity of poverty with respect to GDP or productivity,and,is the change in the error term.By employing this model,we directly estimate the growth elasticit

85、y of povertyhow changes in GDP per capita or productivity influence multidimensional poverty.This approach allows us to not only replicate the foundational analysis conducted by Balasubramanian et al.(2023)and Santos et al.(2019)but also extend our understanding by disaggregating the components of t

86、he MPI,economic growth,and productivity.We want to emphasize that our estimates do not establish causal relationships,as we do not exploit any exogenous variation in economic growth or productivity.Consequently,our analysis primarily demonstrates correlations rather than causation.This limitation me

87、ans that our findings are 10 susceptible to reverse causality and omitted variable bias,which should be considered when interpreting the results.Despite these limitations,we believe that the observed correlations likely reflect underlying causal relationships consistent with our economic understandi

88、ng.11 5.Results In this section,we discuss the results of our estimations,focusing on the coefficients of interest,1 and 2,from equations(1),(2),and(3).These coefficients measure the elasticity between poverty and economic growth.Elasticities reflect the responsiveness of the Multidimensional Povert

89、y Index(MPI)to changes in economic growth.Therefore,1 and 2 indicate the percentage change in MPI resulting from a one percent change in economic growth.The first part of our analysis concentrates on the overall impact of economic growth on the MPI and its subcomponents,assessing variations by regio

90、n,income level,and resource dependency.This section aims to validate findings from prior studies and contribute new insights regarding the disaggregated components of the MPI.In the second part,we employ advanced growth measures to enhance our understanding of the relationship between economic growt

91、h and the MPI.This analysis broadens the existing literature by exploring different facets of growth,including labor productivity,factors of production growth,the expenditure approach,and sustainability considerations.Economic growth and the MPI In the initial phase of our analysis,we focus on the M

92、PI and its subcomponents.We begin by estimating the impact of GDP changes on the MPI,as presented in Table 3.Utilizing a first-difference model,our analysis highlights how GDP changes affect variations in multidimensional poverty.This approach captures the dynamic relationship between economic growt

93、h and multidimensional poverty.12 Table 3:MPI and GDP (1)(2)(3)(4)(5)(6)VARIABLES MPI MPI MPI MPI MPI MPI Change in GDP(current)-0.685*-0.506*(0.201)(0.203)Change in GDP(lagged)-0.608*(0.258)Change in GDP(current+lagged)-1.144*-0.558*-1.136*(0.225)(0.317)(0.224)Change in GDP(following Santos and Bal

94、asubramanian)-0.828*(0.241)Initial level of MPI 0.001 (0.008)Change in GDP(current+lagged)x Initial level of MPI 0.197 (0.144)Change in Gini(current)-0.168 (0.756)Constant-0.052*-0.041*-0.038*-0.047*-0.037*-0.039*(0.006)(0.008)(0.008)(0.011)(0.014)(0.008)Observations 127 127 127 77 127 127 R-squared

95、 0.068 0.106 0.134 0.089 0.170 0.135 Note:Robust standard errors are presented in parentheses.*p 0.01,*p 0.05,*p 0.1.Model 1 investigates the elasticity of poverty with respect to economic growth,using the current change in GDP per capita as the independent variable.Our findings reveal a negative an

96、d significant elasticity of-0.685.This result implies that a 10%increase in GDP per capita is associated with a 6.85%reduction in the MPI.In Model 2,we explore whether the impact of economic growth on MPI is lagged,as suggested by Santos et al.(2019)and Balasubramanian et al.(2023).To this end,we in

97、corporate the lagged change in GDP over the five-year period preceding the first survey used to calculate the change in MPI.Our results indicate that the lagged change in GDP is also negative and significant,with a magnitude similar to the current change in GDP.Further analysis in Table 4 will demon

98、strate that specific subcomponents of the MPI account for the immediate impact while others are responsible for the lagged effect.Model 3 combines the current and lagged impacts of changes in GDP on the MPI to estimate the total effect.With a coefficient of-1.144,our results suggest that a 1%increas

99、e in GDP correlates with a 1.144%decrease in the MPI.Comparatively,this effect is notably smaller than the impact of GDP on monetary poverty as identified in recent research by Wu et al.(2024).They reported a growth elasticity of monetary poverty of approximately 2.5,using the World Banks$2.15-a-day

100、 poverty line.This discrepancy highlights that while economic growth does contribute to reductions in both monetary and multidimensional poverty,its effect is considerably more pronounced on 13 the former,underscoring the complexity of addressing the multidimensional aspects of poverty which include

101、 health,education,and living standards.Our finding in model 3 aligns directionally with previous findings by Santos et al.(2019)and Balasubramanian et al.(2023),reinforcing the negative relationship between economic growth and multidimensional poverty.However,it is noteworthy that our estimates are

102、approximately double those reported in the previous studies(Santos et al.(2019)estimated a coefficient of-0.56,and Balasubramanian et al.(2023)estimated a coefficient of-0.46).The disparity between our findings and those of the other studies may be attributed to differences in the calculation of eco

103、nomic growth rates,the use of different MPI indicators,or variations in the sample.In Model 4,we apply the same economic growth variable as defined by Balasubramanian et al.(2023),who followed Santos et al.(2019).Their approach calculates economic growth as the average annual differences in logarith

104、ms,assuming a lag between the impact of growth on poverty.Consequently,their multidimensional poverty estimate for a particular year is associated with the average GDP per capita over the five years preceding that year.For instance,the change in multidimensional poverty between 2010 and 2015 is link

105、ed to the average GDP per capita change between 20052009 and 20102014.Conversely,we associate the change in multidimensional poverty between 2010 and 2015 with the change in GDP per capita between 2005 and 2015.Model 4 yields a smaller coefficient compared to Model 3.With a coefficient of-0.828,the

106、impact in our sample,using the Balasubramanian/Santos growth rate,remains larger than the impacts found by Balasubramanian et al.(2023)and Santos et al.(2019).However,the reduction in magnitude indicates that solely including the lagged and smoothed change in GDP may overlook some of the effects.The

107、 remaining differences could be attributed to the aforementioned variations in the poverty index or sample differences.Model 5 examines whether the initial poverty level affects our analysis.Given that elasticities can depend on the initial level of the poverty measure(Alkire et al.,2023),it is poss

108、ible that countries at the higher or lower ends of the MPI distribution disproportionately influence the results compared to those in the middle.Including the initial poverty level as a control variable and interacting it with the change in GDP variable shows an insignificant coefficient for the ini

109、tial poverty level itself and for the interaction with change in GDP.This shows that the poverty-growth elasticity does not depend on the initial level of poverty in our sample.For the remainder of our analysis,we choose not to control for the initial poverty level to maintain simplicity.In Model 6,

110、we control for the Gini coefficient to ensure that the observed effects of growth are not confounded by changes in the distribution of income(Balasubramanian et al.,2023;Santos et al.,2019).The Gini coefficient is insignificant,and the GDP coefficients remain unchanged.This result is in line with fi

111、ndings from Balasubramanian et al.(2023)and Santos et al.(2019).Consequently,we exclude this variable in the remaining analysis to keep the model as straightforward as possible while testing our results for robustness.3 3 We also tested the effects of the lagged change in the Gini coefficient,the co

112、mbined current and lagged change in Gini,and the interaction between Gini and GDP change.In all cases,the coefficients related to Gini were found to be insignificant.14 Next,in Table 4,we analyze the impact of economic growth on the individual components of the MPI to identify potential heterogeneit

113、y and determine which components drive the current and lagged impacts.4 All significant coefficients follow the expected direction,indicating that economic growth reduces deprivation in each measured form.Table 4:MPI components and GDP (1)(2)(3)(4)(5)VARIABLES Nutrition Child mortality Years of scho

114、oling School attendance Cooking fuel Change in GDP(current)-0.390-0.153-0.365-0.762*-0.920 (0.325)(0.299)(0.382)(0.296)(0.775)Change in GDP(lagged)-0.644*-0.837*-0.191 0.222-1.022*(0.361)(0.483)(0.480)(0.416)(0.358)Constant-0.035*-0.055*-0.055*-0.051*-0.026 (0.010)(0.011)(0.015)(0.011)(0.026)Observa

115、tions 114 117 125 126 122 R-squared 0.052 0.044 0.008 0.041 0.078 (6)(7)(8)(9)(10)VARIABLES Sanitation Drinking water Electricity Housing Assets Change in GDP(current)-1.805*-0.502-0.580-0.948-1.142*(0.818)(0.516)(0.711)(0.724)(0.436)Change in GDP(lagged)-1.150*-1.765*-2.340*-1.291*-0.757*(0.477)(0.

116、609)(0.929)(0.478)(0.438)Constant-0.005-0.046*-0.043*-0.030-0.065*(0.026)(0.016)(0.021)(0.023)(0.015)Observations 126 124 118 122 124 R-squared 0.164 0.121 0.115 0.087 0.086 Note:Robust standard errors are presented in parentheses.*p 0.01,*p 0.05,*p 0.1.The health components,specifically nutrition a

117、nd child mortality,are presented in Models 1 and 2 of Table 3.Both components exhibit a negative lagged response to economic growth,indicating that higher economic growth reduces nutritional deprivation and lowers child mortality rates,albeit with a delay.Economic growth only weakly influences the e

118、ducation components of the MPI(Models 3 and 4).Years of schooling show no significant coefficient,while school attendance is only significant for the current change in GDP.This suggests that economic growth immediately impacts the decision to enroll or withdraw a child from school,whereas years of s

119、chooling is a more rigid measure that requires considerable time to change.Even the lagged coefficient for years of schooling is insignificant,underscoring its rigidity.This rigidity is expected,as the years of schooling indicator includes all household members aged ten and older,encompassing adults

120、.Consequently,economic growth is unlikely to significantly alter the educational attainment of older 4 For a table with the combined current and lagged impact refer to Table A 3 in the appendix.15 individuals,such as those aged 30 or 40,implying that this variable changes only gradually over time.La

121、gged changes in GDP significantly and negatively influence all living standards components of the MPI.It seems that higher economic growth reduces deprivation in terms of living standards with a delay.Current changes in GDP only affects sanitation and assets with a negative sign,i.e.higher economic

122、growth reduces sanitation and asset deprivations instantly.Overall,the components of the MPI are predominantly and negatively influenced by economic growth in a lagged manner,indicating that economic growth reduces most forms of deprivation,albeit with a delay.Only three componentsschool attendance,

123、sanitation,and assetsare influenced by current changes in GDP and are,therefore,likely responsible for the significant coefficients observed in Models 1 and 2 of Table 3.Next,we examine the relationship between economic growth and its impact on the MPI across different income groups,as shown in Tabl

124、e 5.To address the issue of small sample sizes,which could compromise the robustness of our findings if each income group was analyzed separately,we employ a methodological approach that interacts the GDP growth variable with income group dummies.This technique enables us to utilize the full sample

125、for our analysis without significantly increasing the degrees of freedom.To interpret the interaction approach,it is essential to consider both the coefficient of economic growth and the coefficient of the interaction term.The total effect for an income group is captured by the sum of these two coef

126、ficients.This further means that a negative coefficient for the interaction term indicates that economic growth is more effective at reducing the MPI,while a positive coefficient suggests that economic growth is less effective or may even exacerbate the MPI.The only significant interaction is found

127、in Model 1 for the low-income dummy(Table 5).The positive interaction term indicates a reduction in the overall poverty-reducing effect for low-income countries.In other words,the same economic growth reduces poverty less in low-income countries compared to lower-middle and upper-middle-income count

128、ries.Further analysis of the individual components of the MPI(Table A 4)reveals that low-income countries notably lag in leveraging economic growth to improve nutrition,sanitation,drinking water,and housing.Meanwhile,lower-middle-income countries fall behind in cooking fuel,sanitation,and drinking w

129、ater.Conversely,upper-middle-income countries perform better in transforming economic growth to enhance cooking fuel,sanitation,drinking water,and housing conditions.Next,we examine whether economic growth helps to reduce poverty differently in countries that depend heavily on natural resources.This

130、 investigation is driven by concerns that resource-dependent countries might suffer from the resource curse,which could lead to slower economic growth,increased corruption,deterioration of the tax base,and inefficient resource use(Keller,2020,2022;Leite&Weidmann,1999;Sachs&Warner,1995),which could c

131、ounteract poverty alleviation.For our analysis,we added a variable to Model 4 in Table 5 to identify countries considered resource-dependentdefined as those where natural resource rents exceed 10%of GDP on average over the last five years(Keller,2020).We then explored how this resource dependency in

132、teracts with economic growth in affecting the MPI.16 The findings present two main points.First,the resource dependency indicator by itself does not significantly affect MPI,indicating that resource-dependent countries,on average,do not differ in their level of poverty from other countries.Second,th

133、e interaction between economic growth and resource dependency is significant and positive.This suggests that economic growth has a less substantial impact on reducing poverty in resource-dependent countries compared to those that are not resource dependent.This result reveals an additional concern f

134、or policy makers:the already slow economic growth in resource-dependent countries further translates into less effective poverty reduction,amplifying the urgency of addressing the unique obstacles presented by natural resources.Table 5:MPI and GDP by income groups and resource dependency (1)(2)(3)(4

135、)VARIABLES MPI MPI MPI MPI Dummy low income lower-middle upper-middle Resource dependent Change in GDP(current+lagged)-1.261*-1.171*-1.003*-1.504*(0.251)(0.341)(0.210)(0.283)Dummy 0.026*-0.036*0.005-0.013 (0.013)(0.015)(0.026)(0.014)Dummy x change in GDP 0.727*0.172-0.479 1.003*(0.339)(0.410)(0.481)

136、(0.379)Constant-0.049*-0.024*-0.040*-0.031*(0.012)(0.012)(0.006)(0.012)Observations 126 126 126 125 R-squared 0.251 0.204 0.151 0.163 Note:Robust standard errors are presented in parentheses.*p 0.01,*p 0.05,*p 0.1.Wu et al.(2024)identified a different poverty-growth elasticity for Sub-Saharan Africa

137、 using income poverty,prompting us to examine the relationship between economic growth and its impact on the MPI across different world regions.To address the issue of small sample sizes,we employ the same approach as for the income groups(see above)by interacting individual region dummies with chan

138、ges in GDP.Table 6 presents the results by region for the MPI and highlights three regions:Latin America and the Caribbean,South Asia,and Sub-Saharan Africa.In the case of Sub-Saharan Africa,our findings align with Wu et al.(2024)in that,the growth-poverty relationship is smaller for non-monetary po

139、verty,similar to their findings for monetary poverty.However,we also observe that this reduced relationship holds for Latin America and the Caribbean.Conversely,the opposite is true for South Asia.This region appears to leverage economic growth more effectively in reducing poverty than the rest of t

140、he sample.When examining the subcomponents of the MPI,as detailed in Table A 5,we observe additional nuances.The Arab states,for example,exhibit poor performance in reducing child mortality.However,they excel in providing all living standard components of the MPI compared to the rest of the sample.T

141、he East Asia and Pacific region shows superior performance only regarding assets.The negative MPI result for Latin America and the Caribbean,shown in Table 6,is driven by poor performance in the subcomponents of years of schooling,housing,and assets.South Asia,on the other hand,outperforms the sampl

142、e in the subcomponents of nutrition,years of schooling,and 17 electricity.Lastly,Sub-Saharan Africa encounters significant difficulties with nutrition,cooking fuel,sanitation,drinking water,electricity,and housing.Table 6:MPI and GDP by region (1)(2)(3)(4)(5)(6)VARIABLES MPI MPI MPI MPI MPI MPI Regi

143、on Arab states East Asia and the Pacific Europe and Central Asia Latin America and the Caribbean South Asia Sub-Saharan Africa Change in GDP(current+lagged)-1.221*-0.843*-0.989*-1.290*-1.138*-1.245*(0.257)(0.212)(0.254)(0.237)(0.238)(0.319)Region dummy -0.018-0.005 0.062-0.037*0.047*0.012 (0.011)(0.

144、028)(0.155)(0.018)(0.022)(0.017)Region dummy x change in GDP-0.778-0.674-1.987 1.044*-1.043*1.011*(0.473)(0.564)(3.474)(0.596)(0.542)(0.374)Constant-0.033*-0.042*-0.040*-0.032*-0.038*-0.049*(0.010)(0.008)(0.006)(0.009)(0.008)(0.016)Observations 127 127 127 127 127 127 R-squared 0.153 0.164 0.157 0.1

145、52 0.136 0.230 Note:Robust standard errors are presented in parentheses.*p 0.01,*p 0.05,*p 0.1.To summarize the first part of the analyses,our results confirmed that economic growth consistently reduces the MPI,consistent with the findings from Balasubramanian et al.(2023)and Santos et al.(2019).The

146、 effect varies significantly across different income groups,regions,and levels of resource dependency.Specifically,the impact is less pronounced in low-income countries,Sub-Saharan Africa,Latin America,and the Caribbean,and in resource-dependent countries.Additionally,while economic growth generally

147、 leads to a reduction in MPI components,these effects often appear with a temporal lag.Advanced growth measures and the MPI In the second part of our analysis,we focus on how economic growth influences the MPI.In this section,we measure economic growth in various ways:1)as labor productivity,2)as fa

148、ctors of production growth,3)as expenditure approach components,and 4)from a sustainability perspective(Table 7).18 Table 7:MPI and growth (1)(2)(3)(4)VARIABLES MPI MPI MPI MPI Change in labour productivity-0.828*(0.332)Change in TFP -1.390*(0.434)Change in physical capital -0.638 (0.536)Change in h

149、uman capital -0.467 (0.712)Change in consumption -1.711*(0.421)Change in investment -0.081 (0.192)Change in government spending 0.154 (0.148)Change in export -0.120 (0.225)Change in import 0.299 (0.422)Sustainable change in GNI -0.911*(0.305)Unsustainable change in GNI -0.328*(0.156)Constant-0.049*-

150、0.054*-0.031*-0.038*(0.011)(0.015)(0.013)(0.012)Observations 91 65 89 80 R-squared 0.065 0.119 0.168 0.129 Note:Robust standard errors are presented in parentheses.*p 0.01,*p 0.05,*p 0.1.In model 1 of Table 7,we analyze the relationship between changes in labor productivity and the MPI.Labor product

151、ivity,characterized as the ratio of an economys GDP to its number of workers,is a crucial indicator of economic efficiency and capacity.We find that improvements in labor productivity are inversely related to the MPI.This finding echoes the negative correlation observed between economic growth and m

152、ultidimensional poverty,albeit with a slightly less pronounced effect.Furthermore,we observe a consistent trend when dissecting the impact of labor productivity improvements across the specific components of the MPInamely health,education,and living standards.The coefficients suggest that increases

153、in labor productivity contribute to positive outcomes in most dimensions of poverty,much like the effects attributed to economic growth(Table A 6 in the appendix).Delving deeper into the relationship between economic growth and multidimensional poverty,we dissect economic growth into factors of prod

154、uction such as total factor productivity(TFP),capital deepening,and human capital,each potentially influencing multidimensional poverty differently.TFP,which measures the efficiency of all inputs in the production process,can directly enhance economic output without additional inputs,potentially imp

155、roving employment opportunities and 19 wages and reducing poverty.Human capital,reflecting the education and skills of the workforce,theoretically reduces poverty by increasing employability and earning potential.Capital deepening,involving increased capital per worker,is expected to boost productiv

156、ity and wages but requires complementary skills and technologies to reduce poverty effectively.Our investigation,as presented in model 2 of Table 7,places a significant emphasis on the role of TFP.The data compellingly indicate that advancements in TFP are closely linked with reductions in the MPI.T

157、his finding highlights the crucial role of technological progress and efficiency improvements in diminishing multidimensional povertys various facets.Notably,the positive impact of TFP extends across all health,living standards,and the years of schooling dimensions of the MPI(Table A 7 in the append

158、ix).The insignificant results for human and physical capital growth are surprising,as these factors are typically expected to reduce poverty.One possible explanation,though speculative,is that the benefits of human capital and capital deepening may be concentrated among higher-income groups,not reac

159、hing the poorest segments of society.Additionally,if capital deepening is associated with automation and labor-saving technologies,it might reduce the demand for low-skilled labor,thereby failing to alleviate poverty among the least skilled workers.Overall,the insignificant coefficients for human ca

160、pital and capital deepening do not necessarily indicate a lack of relationship between these variables and poverty.It may be that in some countries the effect is positive,while in others it is negative,resulting in an overall average zero effect in our sample.This further suggests that underlying fa

161、ctors determine the direction of the effect.Further research could focus on identifying these factors,as it is beyond the scope of this study,to better understand the conditions under which human capital and capital deepening influence poverty.Since only TFP significantly and unambiguously reduces m

162、ultidimensional poverty,policies could focus on fostering technological development,innovation,and efficiency across industries.For example,policies advocated to improve TFP include tax incentives,subsidies,or direct funding to foster research and development,strengthening intellectual property righ

163、ts aiming to incentivize private sector investment in innovation,and education and training programs tailored to the evolving needs of industries to help workers adapt to and benefit from new technologies.Note that these policies are not exhaustive and,more importantly,we are not testing them in thi

164、s study;hence,their mention here should only be seen as a motivation for future research.Next,we evaluate the impact of economic growth on MPI from the expenditure perspective of GDP.According to the expenditure approach,GDP equals consumption,(public and private)investment,government spending,and e

165、xports minus imports.For each expenditure category,we calculated the per capita value and determined the average annual change in the logarithm of the variable.Generally,we would expect that growth in each expenditure component reduces poverty.Consumption growth,for instance,is directly associated w

166、ith improved living standards.Public and private investment in physical capital and infrastructure can improve productivity and access to services.Efficient government spending can directly impact poverty through social welfare programs and public services targeting the poor.Export growth might gene

167、rate employment opportunities,potentially reducing poverty.Similarly,imports can reduce costs and increase access to various goods,potentially enhancing welfare,provided that price stability is maintained.However,situations exist in which growth in one or more expenditure components can lead to an i

168、ncrease in poverty.For example,while consumption growth can improve living standards,it might 20 also lead to increased inequality or unsustainable debt levels,exacerbating poverty.Investment in physical capital and infrastructure,as well as export growth,for example due to natural resource projects

169、,although beneficial for productivity and employment,could displace communities or cause environmental degradation,negatively impacting poverty levels.Inefficient or corrupt government spending can divert resources away from those in need.Lastly,a surge in imports,while reducing costs,might harm dom

170、estic industries and lead to job losses,thereby increasing poverty.Model 3 of Table 7 shows the results for the impact of each expenditure component on multidimensional poverty.The only significant coefficient comes from consumption.Consumption growth reduces multidimensional poverty,while changes i

171、n investment,government spending,exports,and imports do not significantly impact the MPI.The subcomponents of the MPI,shown in Table A 8 in the appendix,reveal a similar result,namely that predominantly changes in consumption reduce poverty for almost all subcomponents,except for years of schooling.

172、The main results from Model 3 indicate that consumption growth unambiguously reduces poverty,while the remaining expenditure components do not show a significant overall impact.It is important to note that this does not imply that government spending,investment,or trade have no impact on poverty;rat

173、her,the total impact in our sample averages out to zero.This suggests that in some countries,for example effective government spending reduces poverty,whereas in others,wasteful spending or corruption exacerbates it.For example,countries rich in natural resources such as oil and gas often experience

174、 significant increases in investment,government spending,and export due to new resource projects.However,resource-rich countries with poor institutional quality frequently fall victim to the resource curse,characterized by higher corruption,waste,and mismanagement,which negatively impacts poverty(Bh

175、attacharyya&Hodler,2010;Keller,2020;Robinson&Torvik,2005).The example of resource-rich countries illustrates that underlying factors,such as institutional quality,can determine whether growth in government spending,investment,exports,and imports reduces or increases poverty.Identifying these factors

176、 is beyond the scope of this paper but could serve as a motivation for further research.For policy makers,our results suggest that,in case of doubt,it is safer to promote consumption-increasing policies to reduce poverty rather than policies aimed at increasing government spending,investment,exports

177、,or imports,especially when the underlying factors are unknown.Economic growth can be split into sustainable and unsustainable components,each with distinct implications for multidimensional poverty.Following Mahler(2021),the unsustainable part of national income can be defined as the sum of the val

178、ue of depletion of natural resources,the cost of damage to national income due to CO2 and particulate emissions,and the consumption of fixed capital.The sustainable part is the residual of total national income(GNI).Sustainable growth reflects income generated through means that do not compromise fu

179、ture economic potential.This growth is presumed to be inclusive and enduring,likely supporting long-term living standards and social welfare improvements.On the other hand,unsustainable growth includes elements like the depletion of natural resources and the consumption of fixed capital,which may in

180、itially stimulate economic activity and reduce poverty but potentially undermine future welfare and environmental sustainability.21 Model 4 of Table 7 presents the results with sustainable and unsustainable changes in GNI as independent variable.Both coefficients are negative and significant,indicat

181、ing that both forms of growth reduce multidimensional poverty.However,comparing the magnitudes reveals that sustainable growth outperforms unsustainable growth by a factor of approximately three.This result aligns with Mahlers(2021)findings,who used monetary variables in his analysis.The MPI subcomp

182、onents generally follow a similar pattern,showing a solid poverty-reducing effect through sustainable growth and a smaller or even insignificant effect through unsustainable growth(Table A 9 in the appendix).To summarize the second part of the analysis,our findings indicate that not all dimensions o

183、f economic growth equally contribute to the reduction of multidimensional poverty.Specifically,economic growth propelled by advancements in TFP,consumption growth,and sustainable growth consistently and significantly decreases the MPI in comparison to other contributors examined.Conversely,other con

184、tributors such as human capital development,capital deepening,investment,government spending,exports,and imports do not exhibit a statistically significant impact on the MPI.This lack of significance suggests an ambiguous effect;in some cases,these contributors may reduce poverty,while in others the

185、y may not,resulting in an overall average effect close to zero.We propose that the direction and magnitude of these effects are contingent upon country-specific underlying factors.Thus,further research is warranted to elucidate the conditions under which human capital,capital deepening,investment,go

186、vernment spending,exports,and imports can effectively reduce poverty.Given the more definitive impact of TFP,consumption,and sustainable growth on diminishing multidimensional poverty,it is advisable for policy makers to prioritize strategies that foster these specific forms of growth,particularly i

187、n contexts where the influence of other growth contributors is uncertain or poorly understood.22 6.Robustness Checks To ensure the integrity of our findings,we conducted robustness checks,mainly focusing on the potential influence of changes in inequality,applying alternative methodologies,and testi

188、ng for outliers.In the first robustness test,we controlled for changes in inequality,measured by the Gini coefficient,across all model specifications used in our analysis.This adjustment aimed to assess whether variations in income distribution might influence the relationship between economic growt

189、h and multidimensional poverty.The analysis demonstrates that the coefficients for changes in the Gini coefficient are statistically insignificant in nearly all cases.Importantly,including this inequality measure does not modify the significant coefficients for changes in GDP.This outcome indicates

190、that our main findingsthat economic growth contributes to reductions in multidimensional povertyare not affected by fluctuations in income inequality.Table A 10 in the appendix provides a comprehensive view of these robustness checks,detailing the effects of changes in GDP on the MPI and its compone

191、nts,with adjustments for changes in inequality.5 To further validate the robustness of our results,we explored alternative methodological approaches,starting with a simple Ordinary Least Squares(OLS)regression using levels instead of changes to estimate the correlation between the MPI and GDP per ca

192、pita.While our primary first difference model is also an OLS regression,it employs changes in variables rather than levels,effectively canceling out time-invariant effects,which is why we prefer it.In this robustness check,the OLS regression incorporates the variables levels.Consequently,although we

193、 do not expect the coefficients to exactly mirror those from the first difference model,we anticipate that the direction and significance of the coefficients will align.The findings from the level OLS regression are indeed in harmony with those from the first difference model;most coefficients are s

194、imilar in direction and significance,reinforcing the negative correlation between economic growth and multidimensional poverty observed in our primary analysis.Table A 11 in the appendix presents the results of this robustness check,detailing the relationship between GDP per capita and the MPI and i

195、ts components.6 Continuing our robustness checks,we evaluated the suitability of a non-linear model to account for potential biases related to the initial poverty level.Given that this relationship may not adhere to a linear model,as discussed in Alkire et al.(2023),a model capable of handling logis

196、tic distributions is recommended.To this end,we employed beta regression,a technique well-suited for continuous dependent variables constrained within a range of 0 to 1(Cribari-Neto&Zeileis,2010).While our preference remains with the first difference modelprimarily due to its ability to control for

197、country-specific fixed effects,which is challenging in beta regression due to our sample size limitationswe implemented beta regression to ensure our findings were not biased by the linear 5 Further results using the remaining independent variables of the analysis are not reported due to space reaso

198、ns and can be obtained upon request.6 Further results using the remaining independent variables of the analysis are not reported due to space reasons and can be obtained upon request.23 assumptions of our primary model.In this robustness check,we regressed on the levels of variables,similar to the O

199、LS approach.The results from the beta regression are consistent with those from the first difference model;most coefficients are negative,and the key coefficients remain significant.This consistency across different modeling approaches underscores the robustness of our findings,demonstrating that ou

200、r initial results are not an artifact of the linear modeling approach.Detailed outcomes of this analysis are presented in Table A 12 in the appendix,which showcases the relationships between GDP per capita and the MPI and its components.7 For the final robustness check,we conducted an analysis to ac

201、count for potential outliers that could disproportionately influence our findings.This involved re-running all specifications of our analysis while sequentially excluding one country at a time(Pennings,2021).The coefficients and confidence intervals obtained from this robustness check are presented

202、in Figure B 1 in Appendix B.The figure illustrates the relationships between GDP per capita and the MPI,along with its individual components.8 The results of this robustness check confirm the stability of our findings,as no single country was found to have a significant impact on the overall results

203、.7 Further results using the remaining independent variables of the analysis are not reported due to space reasons and can be obtained upon request.8 Further results using the remaining independent variables of the analysis are not reported due to space reasons and can be obtained upon request.24 7.

204、Conclusion This paper has explored the intricate dynamics between economic growth and multidimensional poverty.This concept goes beyond income-based measures to include other dimensions of deprivation,such as health,education,and living standards.Our analysis demonstrates a significant negative corr

205、elation between economic growth and the Multidimensional Poverty Index(MPI),suggesting that multidimensional poverty decreases as economies expand.However,this relationship is not uniform;it varies significantly across income groups,world regions,resource dependency,and growth dimension.In the first

206、 part of our analysis,we confirmed that economic growth has a consistent poverty-reducing effect on the MPI,aligning with findings from previous studies.However,the magnitude of this effect varies significantly across different income groups,regions,and levels of resource dependency.Specifically,the

207、 impact of economic growth on reducing MPI is less pronounced in low-income countries,regions such as Sub-Saharan Africa and Latin America and the Caribbean,and in countries identified as resource dependent.This underscores the importance of considering regional and income-specific contexts when des

208、igning and implementing economic growth policies aimed at poverty reduction.Furthermore,our analysis reveals that the components of the MPI are generally negatively influenced by economic growth,indicating a reduction in multidimensional poverty.However,these effects often manifest with a temporal l

209、ag,highlighting the complex and sometimes delayed nature of poverty reduction through economic growth.In the second part of our analysis,we explored the diverse dimensions of economic growth and their respective impacts on multidimensional poverty.Our findings reveal that these dimensions do not con

210、tribute equally to the reduction of the MPI.Specifically,growth driven by advancements in total factor productivity(TFP),consumption growth,and sustainable growth shows a consistent and significant decrease in MPI,standing out as particularly effective in reducing multidimensional poverty.In contras

211、t,other contributors to economic growthsuch as human capital development,capital deepening,investment,government spending,exports,and importsdid not demonstrate a statistically significant impact on MPI.This variability suggests that these factors may have an ambiguous effect on poverty reduction,so

212、metimes contributing positively and other times not,resulting in an overall neutral impact.We suggest that the direction and magnitude of these effects are influenced by specific country-level factors,such as institutional frameworks,economic structures,and social policies.Therefore,further research

213、 is needed to understand better the conditions under which human capital,capital deepening,investment,government spending,exports,and imports can effectively reduce poverty.Given the clearer positive impact of TFP,consumption,and sustainable growth on reducing multidimensional poverty,policy makers

214、should prioritize strategies that promote these types of growth.This is especially important in contexts where the effects of other growth contributors are uncertain or not well understood,ensuring that efforts to stimulate economic growth also lead to substantial poverty reduction.In conclusion,whi

215、le economic growth is a powerful tool for reducing multidimensional poverty,its effectiveness depends on the nature of the growth and the specific economic context.Future research should continue to explore these dynamics to inform more nuanced and effective poverty reduction strategies,ensuring tha

216、t economic policies are growth-oriented but also inclusive and sustainable.25 References Alkire,S.(2007).The missing dimensions of poverty data:Introduction to the special issue.Oxford Development Studies,35(4),347359.https:/doi.org/10.1080/13600810701701863 Alkire,S.,Foster,J.,Seth,S.,Santos,M.E.,R

217、oche,J.M.,&Balln,P.(2015).Multidimensional Poverty Measurement and Analysis.Oxford University.https:/doi.org/https:/doi.org/10.1093/acprof:oso/9780199689491.001.0001 Alkire,S.,Jindra,C.,Aguilar,G.R.,&Vaz,A.(2017).Multidimensional Poverty Reduction Among Countries in Sub-Saharan Africa.Forum for Soci

218、al Economics,46(2:Social Economics,Poverty,and Human Development).https:/doi.org/https:/doi.org/10.1080/07360932.2017.1310123 Alkire,S.,Kanagaratman,U.,&Suppa,N.(2021).The Global Multidimensional Poverty Index(MPI):2018 Revision(51;OPHI MPI Methodological Note).https:/ophi.org.uk/publications/MN-51#

219、:text=Citation%3A Alkire%2C S.%2C,OPHI)%2C University of Oxford.Alkire,S.,Nogales,R.,Quinn,N.N.,&Suppa,N.(2023).On track or not?Projecting the global Multidimensional Poverty Index.Journal of Development Economics,165(July).https:/doi.org/10.1016/j.jdeveco.2023.103150 Alkire,S.,Roche,J.M.,Seth,S.,&S

220、umner,A.(2013).WHERE DO THE WORLDS POOREST LIVE?A MULTIDIMENSIONAL APPROACH TO THE BOTTOM BILLON(Oxford Poverty and Human Development Initiative(OPHI),Issue May).https:/ophi.org.uk/publication/RP-39a Alkire,S.,&Santos,M.E.(2013).A Multidimensional Approach:Poverty Measurement&Beyond.Social Indicator

221、s Research,112(2),239257.https:/doi.org/10.1007/s11205-013-0257-3 Aryeetey Ernest The Informal Economy Economic Growth and Poverty in Sub Saharan Africa in Andrew McKay and Erik Thorbecke eds Economic Growth and Poverty Reduction in Sub-Saharan Africa:Current and Emerging Issues Oxford online edn Ox

222、ford Academic Nov Balasubramanian,P.,Burchi,F.,&Malerba,D.(2023).Does economic growth reduce multidimensional poverty?Evidence from low-and middle-income countries.World Development,161,106119.https:/doi.org/10.1016/j.worlddev.2022.106119 Bhattacharyya,S.,&Hodler,R.(2010).Natural resources,democracy

223、 and corruption.European Economic Review,54(4),608621.https:/doi.org/10.1016/J.EUROECOREV.2009.10.004 Burchi,F.,Malerba,D.,Rippin,N.,&Montenegro,C.E.(2019).Comparing global trends in multidimensional and income poverty and assessing horizontal inequalities(No.2/2019;Discussion Paper,Issue 2).https:/

224、doi.org/https:/doi.org/10.23661/dp2.2019 Cox,N.J.(2007).Speaking Stata:Identifying spells.The Stata Journal,7(2),249265.https:/doi.org/10.1177/1536867x0700700209 Cribari-Neto,F.,&Zeileis,A.(2010).Beta Regression in R.Journal of Statistical Software,34(2),124.https:/doi.org/https:/doi.org/10.18637/js

225、s.v034.i02 Cruz,M.,Foster,J.,Quillin,B.,&Schellekens,P.(2015).Ending Extreme Poverty and Sharing Prosperity:Progress and Policies(101740;Policy Research Note).https:/documents.worldbank.org/en/publication/documents-26 reports/documentdetail/801561468198533428/ending-extreme-poverty-and-sharing-prosp

226、erity-progress-and-policies Dieppe,A.(2021).Global Productivity:Trends,Drivers,and Policies(A.Dieppe(ed.);2021st ed.).World Bank.https:/doi.org/doi:10.1596/978-1-4648-1608-6 Djossou,G.N.,Kane,G.Q.,&Novignon,J.(2017).Is Growth Pro-Poor in Benin?Evidence Using a Multidimensional Measure of Poverty.Pov

227、erty&Public Policy,9(4),426443.https:/doi.org/https:/doi.org/10.1002/pop4.199 Keller,M.(2020).Wasted windfalls:Inefficiencies in health care spending in oil rich countries.Resources Policy,66(101618).https:/doi.org/10.1016/j.resourpol.2020.101618 Keller,M.(2022).Oil Revenues vs Domestic Taxation:Dee

228、per insights into the crowding-out effect.Resources Policy,76(June 2022),102560.https:/doi.org/10.1016/j.resourpol.2022.102560 Leite,C.,&Weidmann,J.(1999).Does Mother Nature Corrupt?Natural Resources,Corruption,and Economic Growth(WP/99/85;IMF Working Paper).https:/www.imf.org/external/pubs/ft/wp/19

229、99/wp9985.pdf Mahler,D.G.(2021).Is green growth more effective at reducing poverty than gray growth?World Bank Blogs.https:/blogs.worldbank.org/en/opendata/green-growth-more-effective-reducing-poverty-gray-growth McKay Andrew and Erik Thorbecke eds Economic Growth and Poverty Reduction in Sub-Sahara

230、n Africa:Current and Emerging Issues Oxford online edn Oxford Academic Nov Palmer,M.,Groce,N.,Mont,D.,Nguyen,O.H.,&Mitra,S.(2015).The economic lives of people with disabilities in Vietnam.PLoS ONE,10(7),116.https:/doi.org/10.1371/journal.pone.0133623 Pennings,S.(2021).Cross-Region Transfer Multiplie

231、rs in a Monetary Union:Evidence from Social Security and Stimulus Payments.American Economic Review,111(5),16891719.Robinson,J.A.,&Torvik,R.(2005).White elephants.Journal of Public Economics,89(23),197210.https:/doi.org/10.1016/j.jpubeco.2004.05.004 Sachs,J.,&Warner,A.(1995).Natural Resource Abundan

232、ce and Economic Growth.In National Bureau of Economic Research(No.5398;NBER Working Paper Series).https:/doi.org/10.3386/w5398 Santos,M.E.,Dabus,C.,&Delbianco,F.(2019).Growth and poverty revisited from a multidimensional perspective.The Journal of Development Studies,55(2),260277.https:/doi.org/http

233、s:/doi.org/10.1080/00220388.2017.1393520 Sen,A.(1999).Development as freedom.Oxfort University Press.Seth,S.,&Alkire,S.(2021).Multidimensional poverty and inclusive growth in india:An analysis using growth elasticities and semielasticities.In OPHI Working Paper(137;OPHI Working Paper).https:/doi.org

234、/10.1108/S1049-258520210000029007 Tran,V.Q.,Alkire,S.,&Klasen,S.(2015).Static and Dynamic Disparities between Monetary and Multidimensional Poverty Measurement:Evidence from Vietnam.Emerald Group Publishing Limited.https:/ 27 World Bank.(2024).GDP per capita(constant 2015 US$)Data set.World Developm

235、ent Indicators.https:/data.worldbank.org/indicator/NY.GDP.PCAP.KD Wu,H.,Atamanov,A.,Bundervoet,T.,&Paci,P.(2024).Is economic growth less welfare enhancing in Africa?Evidence from the last forty years.World Development,184,106759.doi:https:/doi.org/10.1016/j.worlddev.2024.106759 28 Appendix A 29 Tabl

236、e A 1:List of countries and survey years Region Country Survey years Arab States Algeria 2013,2019 Egypt,Arab Rep.2008,2014 Iraq 2011,2018 Jordan 2012,2018 Morocco 2011,2018 West Bank and Gaze 2010,2014,2020 Sudan 2010,2014 Tunisia 2012,2018 Yemen,Rep.2006,2013 East Asia and the Pacific Indonesia 20

237、12,2017 Cambodia 2010,2014,2022 Lao PDR 2012,2017 Mongolia 2010,2013,2018 Philippines 2013,2017 Thailand 2012,2016,2019 Timor-Leste 2010,2016 Viet Nam 2014,2021 Europe and Central Asia Albania 2009,2018 Armenia 2010,2016 Bosnia and Herzegovina 2006,2012 Kazakhstan 2011,2015 Kyrgyzstan 2006,2014,2018

238、 Moldova 2005,2012 North Macedonia 2006,2011,2019 Montenegro 2013,2018 Serbia 2010,2014,2019 Tajikistan 2012,2017 Turkmenistan 2006,2016,2019 Ukraine 2007,2012 Latin America and the Caribbean Belize 2011,2016 Bolivia 2003,2008,2016 Colombia 2010,2016 Dominican Republic 2007,2014,2019 Ecuador 2014,20

239、18 Guyana 2009,2014,2020 Honduras 2006,2012,2019 Haiti 2012,2017 Mexico 2012,2016,2020,2021 Nicaragua 2001,2012 Peru 2012,2018,2019,2021 Suriname 2006,2010,2018 Trinidad and Tobago 2006,2011 Afghanistan 2011,2016 South Asia Bangladesh 2014,2019 30 India 2006,2016,2021 Nepal 2011,2016,2019 Pakistan 2

240、013,2018 Sub-Saharan Africa Burundi 2010,2017 Benin 2014,2018 Burkina Faso 2006,2010 Central African Republic 2000,2010 Central African Republic 2010,2019,2019 Cte dIvoire 2012,2016 Cameroon 2011,2014,2018 Congo,Dem.Rep.2007,2014,2018 Congo,Rep.2005,2015 Ethiopia 2011,2016,2019 Gabon 2000,2012 Ghana

241、 2011,2014,2018 Guinea 2012,2016,2018 Gambia,The 2006,2013,2018,2020 Guinea-Bissau 2014,2019 Kenya 2009,2014 Liberia 2007,2013,2020 Lesotho 2009,2014,2018 Madagascar 2009,2018,2021 Mali 2006,2015,2018 Mozambique 2003,2011 Mauritania 2011,2015,2021 Malawi 2010,2016,2020 Namibia 2007,2013 Niger 2006,2

242、012 Nigeria 2013,2017,2018,2021 Rwanda 2010,2015,2020 Senegal 2005,2017,2019 Sierra Leone 2013,2017,2019 So Tom and Prncipe 2009,2014,2019 eSwatini 2010,2014 Chad 2010,2015,2019 Togo 2010,2014,2017 Tanzania 2010,2016 Uganda 2011,2016 Zambia 2007,2014,2018 Zimbabwe 2011,2015,2019 31 Table A 2:Summary

243、 statistics N Mean St.d.Min Max MPI 127-0.065 0.062-0.283 0.219 Spell length in years 127 5.283 2.138 1.000 12.000 Nutrition 114-0.059 0.077-0.338 0.466 Child mortality 117-0.075 0.089-0.524 0.314 Years of schooling 125-0.067 0.111-0.775 0.249 School attendance 126-0.057 0.090-0.331 0.268 Cooking fu

244、el 122-0.069 0.112-0.592 0.529 Sanitation 126-0.072 0.141-0.718 0.599 Drinking water 124-0.093 0.138-0.475 0.904 Electricity 118-0.098 0.184-0.669 1.371 Housing 122-0.080 0.134-0.955 0.469 Assets 124-0.100 0.113-0.582 0.159 Change in GDP(current+lagged)127 0.025 0.020-0.013 0.089 Change in labour pr

245、oductivity 91 0.022 0.020-0.019 0.090 Change in TFP 65 0.005 0.021-0.062 0.051 Change in physical capital 66 0.013 0.016-0.018 0.055 Change in human capital 66 0.005 0.010-0.060 0.021 Change in consumption 95 0.024 0.019-0.016 0.087 Change in investment 90 0.037 0.061-0.295 0.142 Change in governmen

246、t spending 92 0.029 0.039-0.119 0.199 Change in export 94 0.029 0.060-0.387 0.153 Change in import 94 0.037 0.051-0.303 0.147 Sustainable change in GNI 80 0.024 0.021-0.016 0.084 Unsustainable change in GNI 81 0.019 0.036-0.078 0.106 32 Table A 3:MPI components and GDP(current+lagged)(1)(2)(3)(4)(5)

247、(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-1.132*-0.955*-0.730-0.598-1.911*-2.815*-2.318*-2.992*-2.181*-1.658*(0.227)(0.339)(0.466)(0.367)(0.735)(0.819)(0.523

248、)(0.804)(0.739)(0.549)Constant-0.030*-0.054*-0.049*-0.047*-0.023-0.000-0.040*-0.036-0.027-0.066*(0.012)(0.011)(0.013)(0.011)(0.026)(0.027)(0.017)(0.025)(0.024)(0.014)Observations 114 117 125 126 122 126 124 118 122 124 R-squared 0.076 0.044 0.015 0.020 0.090 0.167 0.141 0.126 0.096 0.074 33 Table A

249、4:MPI components and GDP by income Low-income countries (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-1.240*-0.956*-0.716-0.671-2.242*-3.563*-2.54

250、0*-3.484*-2.982*-1.560*(0.318)(0.456)(0.603)(0.507)(1.116)(1.149)(0.698)(1.107)(1.134)(0.679)Low-income dummy 0.016 0.034*0.043*0.036*0.015-0.022 0.031 0.027-0.006 0.069*(0.020)(0.019)(0.021)(0.019)(0.043)(0.042)(0.027)(0.041)(0.040)(0.023)Change in GDP x inc.dummy 0.733*0.235-0.027 0.393 1.550 2.61

251、3*1.833*1.965 2.648*0.136 (0.415)(0.581)(0.764)(0.593)(1.170)(1.298)(0.780)(1.337)(1.169)(1.014)Constant-0.038*-0.068*-0.065*-0.061*-0.030 0.006-0.056*-0.049-0.025-0.093*(0.019)(0.016)(0.019)(0.016)(0.043)(0.042)(0.025)(0.041)(0.039)(0.020)Observations 114 116 124 125 121 125 123 117 121 123 R-squar

252、ed 0.116 0.087 0.048 0.091 0.137 0.221 0.226 0.188 0.168 0.158 Lower-middle income countries (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-0.974*-

253、1.054*-1.357*-0.677-2.677*-3.561*-2.873*-3.383*-2.707*-2.168*(0.289)(0.471)(0.615)(0.479)(0.998)(1.118)(0.707)(1.174)(1.118)(0.723)Lower-middle income dummy-0.034-0.047*-0.018-0.030-0.099*-0.092*-0.053*-0.076-0.073*-0.048 (0.020)(0.021)(0.024)(0.020)(0.045)(0.044)(0.028)(0.047)(0.041)(0.032)Change i

254、n GDP x inc.dummy-0.189 0.351 1.329 0.110 2.522*2.232*1.967*1.222 1.579 1.362 (0.465)(0.652)(0.827)(0.677)(1.247)(1.285)(0.858)(1.413)(1.228)(1.112)34 Constant-0.018-0.035*-0.044*-0.035*0.008 0.030-0.026-0.010-0.003-0.050*(0.017)(0.015)(0.018)(0.014)(0.035)(0.037)(0.023)(0.034)(0.033)(0.016)Observat

255、ions 114 116 124 125 121 125 123 117 121 123 R-squared 0.128 0.091 0.036 0.053 0.141 0.214 0.155 0.156 0.126 0.095 Upper-middle income countries (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity

256、Housing Assets Change in GDP(current+lagged)-1.066*-0.901*-0.470-0.633*-0.706*-1.403*-1.101*-2.239*-1.041*-1.379*(0.268)(0.325)(0.357)(0.336)(0.412)(0.404)(0.342)(0.554)(0.312)(0.554)Upper-middle income dummy 0.016 0.014-0.044-0.013 0.072 0.107-0.009 0.041 0.080-0.041 (0.042)(0.041)(0.040)(0.031)(0.

257、078)(0.077)(0.050)(0.082)(0.074)(0.030)Change in GDP x inc.dummy-0.189-0.316-1.104-0.126-3.546*-4.089*-2.822*-2.684-4.208*-1.052 (0.560)(0.856)(0.950)(0.928)(1.960)(1.902)(1.184)(1.976)(2.192)(0.932)Constant-0.035*-0.057*-0.041*-0.044*-0.049*-0.038*-0.049*-0.051*-0.051*-0.058*(0.007)(0.009)(0.009)(0

258、.010)(0.014)(0.013)(0.010)(0.017)(0.012)(0.015)Observations 114 116 124 125 121 125 123 117 121 123 R-squared 0.080 0.049 0.097 0.035 0.162 0.258 0.266 0.160 0.179 0.145 35 Table A 5:MPI components and GDP by region Arab states (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years

259、 of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-1.208*-1.142*-0.656-0.649-1.498*-2.160*-2.256*-3.312*-1.516*-1.846*(0.258)(0.368)(0.499)(0.414)(0.370)(0.552)(0.571)(0.863)(0.430)(0.580)Arab states dummy-0.022-0.042*0.016

260、-0.012 0.082 0.150*0.013-0.057 0.140*-0.046 (0.021)(0.019)(0.021)(0.014)(0.090)(0.088)(0.028)(0.039)(0.071)(0.030)Change in GDP x reg.dummy-0.341 1.533*-1.852-0.414-13.483*-14.232*-4.079*-9.806*-11.854*-4.066*(0.727)(0.679)(1.192)(0.640)(5.509)(5.908)(1.503)(4.056)(3.950)(2.112)Constant-0.026*-0.046

261、*-0.051*-0.044*-0.030*-0.018-0.039*-0.009-0.046*-0.052*(0.014)(0.013)(0.015)(0.013)(0.013)(0.014)(0.019)(0.027)(0.014)(0.016)Observations 114 117 125 126 122 126 124 118 122 124 R-squared 0.085 0.056 0.019 0.024 0.272 0.343 0.162 0.293 0.223 0.144 East Asia and the Pacific (1)(2)(3)(4)(5)(6)(7)(8)(9

262、)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-0.980*-0.922*-0.366-0.742*-1.784*-2.654*-2.355*-2.236*-2.249*-1.058*(0.267)(0.461)(0.586)(0.366)(0.956)(1.029)(0.674)(0.851)(0

263、.864)(0.568)East Asia and Pacific dummy-0.019-0.064*0.026-0.078-0.017 0.029-0.074 0.050-0.063 0.156*(0.040)(0.037)(0.053)(0.102)(0.057)(0.068)(0.083)(0.098)(0.061)(0.056)Change in GDP x reg.dummy-0.115 0.944-1.258 1.577-0.020-0.857 1.261-2.441 1.443-3.879*36 (0.622)(0.793)(1.270)(1.720)(1.338)(1.663

264、)(1.789)(2.190)(1.341)(1.147)Constant-0.032*-0.053*-0.054*-0.043*-0.024-0.003-0.038*-0.048*-0.026-0.077*(0.012)(0.011)(0.012)(0.011)(0.029)(0.030)(0.018)(0.026)(0.025)(0.013)Observations 114 117 125 126 122 126 124 118 122 124 R-squared 0.081 0.051 0.025 0.031 0.092 0.168 0.145 0.146 0.098 0.110 Eur

265、ope and Central Asia (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-0.951*-0.672*-0.620-0.280-1.712*-2.474*-1.911*-2.902*-1.580*-1.338*(0.201)(0.25

266、9)(0.438)(0.365)(0.738)(0.852)(0.515)(0.851)(0.737)(0.592)Europe and C.Asia dummy 0.186 0.183-0.122 0.008 0.012 0.053-0.198*-0.132-0.117-0.245 (0.244)(0.197)(0.258)(0.115)(0.296)(0.145)(0.119)(0.158)(0.258)(0.236)Change in GDP x reg.dummy-4.531-5.404 2.038-1.671-1.610-2.736 2.007 2.196-0.224 3.545 (

267、5.493)(4.880)(6.244)(2.483)(7.461)(3.921)(3.124)(4.092)(5.778)(6.043)Constant-0.037*-0.060*-0.045*-0.047*-0.022-0.003-0.033*-0.032-0.025-0.058*(0.005)(0.008)(0.012)(0.010)(0.026)(0.028)(0.017)(0.026)(0.024)(0.014)Observations 114 117 125 126 122 126 124 118 122 124 R-squared 0.148 0.107 0.039 0.074

268、0.106 0.187 0.263 0.141 0.193 0.195 Latin America and the Caribbean (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets 37 Change in GDP(current+lagged)-1.204*-1.030*-1.178*-0.698*-2

269、.071*-3.076*-2.178*-2.782*-2.477*-2.015*(0.268)(0.371)(0.422)(0.367)(0.848)(0.916)(0.479)(0.719)(0.845)(0.570)Latin America and Caribbean d.-0.030*-0.014-0.097*-0.022-0.042-0.058 0.023 0.084-0.049-0.061*(0.017)(0.023)(0.036)(0.042)(0.036)(0.054)(0.079)(0.117)(0.041)(0.028)Change in GDP x reg.dummy-0

270、.157 0.746 3.293*0.768 1.067 1.993-1.226-0.990 2.098*2.848*(0.480)(0.935)(1.318)(1.508)(0.976)(1.704)(2.387)(3.493)(1.246)(0.910)Constant-0.024*-0.053*-0.034*-0.043*-0.016 0.008-0.043*-0.051*-0.019-0.057*(0.014)(0.012)(0.012)(0.010)(0.031)(0.031)(0.014)(0.020)(0.028)(0.016)Observations 114 117 125 1

271、26 122 126 124 118 122 124 R-squared 0.094 0.046 0.053 0.024 0.096 0.177 0.145 0.146 0.107 0.098 South Asia (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+

272、lagged)-1.070*-0.988*-0.725-0.595-1.978*-2.858*-2.347*-2.886*-2.340*-1.765*(0.233)(0.367)(0.512)(0.395)(0.770)(0.856)(0.551)(0.855)(0.785)(0.586)South Asia dummy 0.049*0.005 0.084*0.051 0.006-0.046 0.100*0.244*0.022 0.019 (0.024)(0.028)(0.034)(0.037)(0.034)(0.049)(0.059)(0.090)(0.031)(0.022)Change i

273、n GDP x reg.dummy-1.613*0.173-1.814*-1.103 0.446 1.354-1.819-5.955*0.739 0.511 (0.665)(0.867)(0.904)(0.864)(0.938)(1.414)(1.606)(2.575)(0.925)(0.748)Constant-0.031*-0.054*-0.050*-0.047*-0.023 0.000-0.041*-0.039-0.026-0.066*(0.012)(0.011)(0.013)(0.011)(0.027)(0.028)(0.017)(0.026)(0.025)(0.014)Observa

274、tions 114 117 125 126 122 126 124 118 122 124 R-squared 0.080 0.045 0.017 0.022 0.093 0.168 0.144 0.135 0.105 0.080 38 Sub-Saharan Africa (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing

275、 Assets Change in GDP(current+lagged)-1.389*-0.905*-0.521-0.469-2.418*-3.852*-2.641*-3.543*-3.012*-1.441*(0.424)(0.507)(0.695)(0.558)(1.358)(1.369)(0.838)(1.398)(1.385)(0.761)Sub-Saharan Africa dummy-0.000 0.012 0.040 0.028-0.005-0.049 0.019 0.015-0.023 0.064*(0.027)(0.024)(0.028)(0.023)(0.057)(0.05

276、5)(0.035)(0.058)(0.054)(0.028)Change in GDP x reg.dummy 1.102*0.168 0.356 0.348 2.308*3.605*2.522*3.140*2.804*1.166 (0.479)(0.646)(0.744)(0.644)(1.374)(1.385)(0.870)(1.429)(1.407)(0.834)Constant-0.034-0.062*-0.075*-0.065*-0.027 0.018-0.061*-0.056-0.020-0.109*(0.026)(0.022)(0.027)(0.020)(0.056)(0.055

277、)(0.034)(0.057)(0.053)(0.026)Observations 114 117 125 126 122 126 124 118 122 124 R-squared 0.113 0.050 0.054 0.063 0.152 0.237 0.262 0.225 0.155 0.214 39 Table A 6:MPI components and labour productivity (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School att

278、endance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in labour productivity-1.062*-0.772*-0.574 0.194-1.486*-1.507*-0.874-1.786*-1.184-1.249*(0.396)(0.395)(0.562)(0.373)(0.737)(0.828)(0.581)(0.863)(0.788)(0.692)Constant-0.036*-0.064*-0.054*-0.073*-0.036-0.031-0.085*-0.066

279、*-0.058*-0.087*(0.016)(0.014)(0.014)(0.012)(0.025)(0.025)(0.017)(0.019)(0.023)(0.016)Observations 83 84 90 90 87 90 91 85 89 89 R-squared 0.068 0.027 0.009 0.002 0.053 0.070 0.026 0.099 0.024 0.043 40 Table A 7:MPI components and labour productivity (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition

280、 Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in TFP-1.497*-1.259*-1.430*-0.407-2.634*-2.216*-1.286*-2.484*-2.120*-1.806*(0.597)(0.738)(0.801)(0.527)(1.094)(1.076)(0.749)(1.136)(1.137)(0.939)Change in physical capital-0

281、.899-0.444 0.034 0.511-0.745-1.439-0.795-1.349-1.355-0.672 (0.621)(0.768)(0.827)(0.633)(1.072)(1.236)(0.942)(1.252)(1.378)(1.026)Change in human capital-1.193-0.941 1.676 1.044-2.903-0.489 2.397*0.355 1.451 0.114 (0.962)(0.786)(3.580)(0.716)(4.542)(1.099)(1.291)(3.023)(1.514)(2.985)Constant-0.036-0.

282、068*-0.074*-0.083*-0.031-0.036-0.113*-0.086*-0.072*-0.110*(0.023)(0.021)(0.035)(0.015)(0.054)(0.036)(0.022)(0.034)(0.035)(0.030)Observations 58 59 64 64 61 64 65 59 63 63 R-squared 0.086 0.043 0.056 0.044 0.098 0.103 0.120 0.157 0.086 0.065 41 Table A 8:MPI components and GDP expenditure components

283、(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in consumption-1.864*-2.444*-1.498-1.591*-3.500*-1.958*-3.079*-3.840*-2.813*-2.547*(0.798)(0.960)(1.736)(0.906)(1.250)(0.55

284、0)(1.031)(1.064)(0.830)(1.104)Change in investment 0.187-0.001-0.501-0.310-0.242-0.479-0.942*-1.372*-0.126-0.353 (0.376)(0.276)(0.403)(0.307)(0.360)(0.378)(0.388)(0.633)(0.299)(0.406)Change in gov.spending 0.299 0.508 0.094-0.083 0.421*-0.004 0.239-0.180 0.310 0.503*(0.247)(0.308)(0.311)(0.279)(0.20

285、6)(0.189)(0.253)(0.275)(0.222)(0.277)Change in export 0.022 0.107-0.128-0.214-0.113-0.773-0.973*-1.808*-0.297-0.252 (0.463)(0.459)(0.434)(0.288)(0.425)(0.529)(0.332)(0.754)(0.433)(0.459)Change in import-0.255-0.020 0.847 0.755 0.688 1.382 2.248*3.679*0.602 0.787 (0.873)(0.699)(0.952)(0.564)(0.879)(0

286、.999)(0.744)(1.478)(0.761)(0.960)Constant-0.019-0.034*-0.046*-0.031*-0.013-0.039*-0.045*-0.043-0.033*-0.064*(0.023)(0.017)(0.021)(0.016)(0.018)(0.018)(0.022)(0.031)(0.014)(0.019)Observations 82 81 89 88 86 88 87 84 85 89 R-squared 0.100 0.132 0.049 0.097 0.210 0.102 0.187 0.205 0.173 0.110 42 Table

287、A 9:MPI components and sustainable growth (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)VARIABLES Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Sustainable change in GNI-0.727*-0.740*-1.603-0.593-1.802*-1.199*-1.860*-3.021*-1.048-1.

288、542 (0.289)(0.351)(1.212)(0.606)(0.934)(0.543)(0.901)(1.095)(0.673)(0.939)Unsustainable change in GNI-0.354*-0.293 0.164 0.053 0.078-0.718*-0.345-0.304-0.557-0.387 (0.161)(0.386)(0.531)(0.286)(0.421)(0.262)(0.392)(0.443)(0.400)(0.415)Constant-0.035*-0.049*-0.043*-0.044*-0.027-0.036*-0.037-0.025-0.05

289、0*-0.062*(0.016)(0.016)(0.019)(0.015)(0.019)(0.017)(0.023)(0.033)(0.014)(0.019)Observations 75 72 80 79 78 79 80 76 77 80 R-squared 0.050 0.035 0.080 0.024 0.117 0.160 0.113 0.163 0.082 0.094 43 Table A 10:Robustness check:MPI,GDP,and Gini (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)VARIABLES MPI Nutrition C

290、hild mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets Change in GDP(current+lagged)-0.830*-1.232*-0.985*-0.715-0.575-1.957*-2.810*-2.206*-2.910*-2.224*-1.659*(0.232)(0.274)(0.346)(0.483)(0.370)(0.744)(0.827)(0.511)(0.793)(0.745)(0.555)C

291、hange in Gini 0.474 1.587 0.791-0.382-0.516 1.157-0.131-2.505-1.851 0.711 0.030 (1.014)(1.639)(1.393)(1.416)(1.087)(1.287)(1.092)(1.528)(1.837)(1.086)(1.461)Constant-0.036*-0.023-0.051*-0.051*-0.049*-0.018-0.001-0.051*-0.044*-0.024-0.066*(0.010)(0.017)(0.014)(0.013)(0.012)(0.027)(0.028)(0.016)(0.025

292、)(0.025)(0.014)Observations 127 114 117 125 126 122 126 124 118 122 124 R-squared 0.138 0.099 0.047 0.016 0.023 0.096 0.167 0.166 0.135 0.098 0.074 44 Table A 11:Robustness check:Alternative methodologies:OLS (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)VARIABLES(log)MPI(log)Nutrition(log)Child mortality(log)

293、Years of schooling(log)School attendance(log)Cooking fuel(log)Sanitation(log)Drinking water(log)Electricity(log)Housing(log)Assets (log)GDP-1.427*-1.298*-1.177*-1.500*-1.563*-1.778*-1.670*-1.786*-2.064*-1.636*-1.657*(0.082)(0.090)(0.108)(0.106)(0.078)(0.118)(0.114)(0.110)(0.131)(0.110)(0.108)Constan

294、t 8.092*7.253*5.255*8.468*8.751*11.065*10.060*10.535*12.647*9.803*9.477*(0.570)(0.618)(0.771)(0.765)(0.562)(0.809)(0.804)(0.761)(0.921)(0.758)(0.755)Observations 211 188 197 209 209 203 210 208 198 206 208 R-squared 0.554 0.540 0.442 0.439 0.585 0.506 0.460 0.524 0.524 0.461 0.493 45 Table A 12:Robu

295、stness check:Alternative methodologies:Beta regression (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)VARIABLES MPI Nutrition Child mortality Years of schooling School attendance Cooking fuel Sanitation Drinking water Electricity Housing Assets (log)GDP-1.180*-1.094*-0.773*-1.024*-0.992*-1.472*-1.230*-1.222*-1.

296、328*-1.287*-1.120*(0.058)(0.053)(0.058)(0.087)(0.057)(0.085)(0.095)(0.065)(0.110)(0.082)(0.079)Constant 6.982*6.371*3.085*5.924*5.474*9.942*7.951*7.376*8.605*8.316*6.524*(0.415)(0.373)(0.443)(0.642)(0.426)(0.626)(0.703)(0.470)(0.800)(0.598)(0.578)Observations 211 188 197 209 209 203 210 208 198 206 208 46 47 Appendix B 48 Figure B 1:Outlier test,excluding one country in each regression.49