Factor Decomposition of the Wealth Distribution in the Euro Area

Transcription

1 Factor Decomposition of the Wealth Distribution in the Euro Area Peter Lindner 1 (Economic Analysis Division, OeNB) Conference: The Future of Capitalism 25 th September Additional to the usual disclaimer, the opinions expressed in this paper and this presentation solely represent those of the author and do not necessarily reflect the official viewpoint of the Oesterreichische Nationalbank or of the Eurosystem. Lindner (OeNB) HFCS 25 th September / 38

2 Motivation Policy relevance Policies often target specific assets Decomposition allows to see the distributional impact Deepening the understanding of the foundation of distribution Investigate cross-country differences Contribution to the literature Factor decomposition of wealth rather than income Application to a large cross country database New decomposition method based in the recentered influence function Benoît Cœuré Lindner (OeNB) HFCS 25 th September / 38

3 Outline 1 Background / Literature 2 Decomposition Methods 3 Data 4 Results 5 Conclusions Lindner (OeNB) HFCS 25 th September / 38

4 Background / Literature Methodology Methodology mostly for income decomposition Long history of authors concerned with decomposition of the distribution: Shorrocks (1982), Lambert and Aronson (1993) natural decomposition Lerman and Yitzhaki (1985), Silber (1989) additive decomposition of Gini Coefficient Sastre and Trannoy (2002), Araar (2006), or Shorrocks (2013) decomposition based on Shapley value Fortin et al. (2011), Firpo et al. (2007) decomposition based on recentered influence function Lindner (OeNB) HFCS 25 th September / 38

5 Background / Literature Income Huge literature on various ways of decomposing income Subgroup decompositions: Okamoto (2011), Chernozhukov et al. (2009), Firpo et al. (2007), Yun (2006), Fields (2003), Juhn et al. (1993) for the U.S. Machado and Mata (2005) for Portugal; Celestin and Clovis (2012) for Cameroon; Fournier (2001) for Taiwan; Morduch and Sicular (2002) for China; Bourguignon et al. (2008) for a comparison of Brazil and the US; López-Feldman et al. (2006) for Mexico. Change over time: Gunatilaka and Chotikapanich (2009), Firpo et al. (2007), Mookherjee and Shorrocks (1982), Fiorio (2006), Yun (2006) Decomposition according to income sources: Paul (2004), Fräßdorf et al. (2011) Lerman and Yitzhaki (1985) Lindner (OeNB) HFCS 25 th September / 38

6 Background / Literature Wealth Few approaches in the literature tackling wealth Sierminska et al. (2012) and (2008): Differences of certain socio-economic characteristics in explaining the wealth portfolio of households in the United States, Germany, Italy, Luxembourg and Spain. Decomposition of the gender wealth gap in Germany Regression based approach yields wealth gap can be attributed mostly to differences in individuals characteristics. Azpitarte (2008): Includes in his study about the household wealth distribution in Spain a decomposition according to components finding that in Spain for the year 2002 around four fifth of the inequality is explained by real assets and only one fifth by financial assets. Brandoliniet al. (2004) Includes a decomposition with similar results for Italian data. Lindner (OeNB) HFCS 25 th September / 38

7 Decomposition Methods Natural Early approach based on Shorrocks (1982) showing the contribution is s k = cov(w k, w) σ 2 (w) Lindner (OeNB) HFCS 25 th September / 38

8 Decomposition Methods Additive Lerman and Yitzhaki (1985) use the additive separability of the covariance to show GC(w) = K k=1 cov[w k, F (w)] 2cov[w k, F (w k )] µ k cov[w k, F (w k )] µ k µ GC(w) = K s k gc k a k. k=1 (Relative) Partial derivative (chain rule) GC(w) ε k := ε k GC(w) = a ks k gc k GC(w) a k. Lindner (OeNB) HFCS 25 th September / 38

9 Decomposition Methods Shapley Shorrocks (2013) and Chantreuil and Trannoy (2013) provide analytical framework to get the contribution of factor k as s k = z Z k z (z 1)!(K z )! [GC(w(z)) GC(w(z k))] K! Lindner (OeNB) HFCS 25 th September / 38

10 Decomposition Methods RIF I Based on Fortin et al. (2011) and Firpo et al. (2007) using RIF regressions to decompose income inequality according to subgroups. Extend this approach for factor decomposition. Suppose small changes in source w k, i.e. think about the counterfactual contribution of w corresponding to a proportional increase in w k w(α) := K (1 + α k ) w k, k=1 marginal contribution of factor k to inequality (inequality measure ν(w(α))) can be defined as s k := ν(w(α)) α k If ν has a differentiable recentered influence function, we can write this as RIF (ν(w)) s k = E[ w k ]. w Lindner (OeNB) HFCS 25 th September / 38

11 Decomposition Methods RIF II For the Gini Coefficient RIF (GC(w)) = w µ GC + 1 w µ + 2 µ hence w s k = E[( GC µ 1 µ + 2 µ F (w)) w k] 0 F (x)dx Since Gini Coefficent is homogeneous of degree zero, i.e. ν(κ α) = ν(α), we have s k = 0 k Lindner (OeNB) HFCS 25 th September / 38

12 Data Euro Area HFCS Unit of collection: household (some information at the personal level) Information collected includes assets, liabilities, income, socio-demographic characteristics All euro area countries (except Ireland and Estonia in the 1 st wave) Field period of the first wave 2010/2011 (exception ES with some households in 2008) A priori harmonized questionnaire and survey techniques (e.g. CAPI interviewing) Harmonisation also of editing and imputation ECB coordinates and supervises quality and methodological consistency International State of the Art concerning data production Complex probability sampling design representative at the country level households surveyed in the euro area (response rate varies across countries) Missing data are multiple imputed (Bayesian approach, 5 implicates) Estimations use final household weights and takes the imputation structure into account Lindner (OeNB) HFCS 25 th September / 38

13 Data Classification of Real Assets Households main residence (HMR) Further real estate (FRE): All forms of real estate such as holiday homes, landed property, rented out property, but also forests or rented out office buildings; excluding HMR (or potentially owned as BUS). Business assets (BUS): Net value of business assets that are owned by the household and where at least one household member works. Other real assets (REST REAL): Other tangible assets; mainly there are all forms of vehicles, but in some cases also other valuables such as collections of pieces of art. Lindner (OeNB) HFCS 25 th September / 38

14 Data Classification of Financial Assets Safe financial assets (SAFE): Sight and savings accounts (including savings in building societies accounts) as well as life insurances and voluntary private pension plans. Risky financial investments (RISKY): Shares, bonds, and mutual funds. Other financial assets (REST FIN): All other forms of financial assets such as silent partnerships, managed accounts, or money owed to a household. Lindner (OeNB) HFCS 25 th September / 38

15 Results Shares I Figure: Share of Total Real Asset Components in terms of Gross Wealth Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

16 Results Shares II Figure: Share of Total Financial Asset Components in terms of Gross Wealth Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

17 Results Descriptive Statistics HMR FRE BUS REST REAL SAFE RISKY REST FIN Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

18 Results Descriptive Statistics HMR FRE BUS REST REAL SAFE RISKY REST FIN Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

19 Results Descriptive Statistics HMR FRE BUS REST REAL SAFE RISKY REST FIN Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

20 Results Descriptive Statistics HMR FRE BUS REST REAL SAFE RISKY REST FIN Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

21 Results Descriptive Statistics HMR FRE BUS REST REAL SAFE RISKY REST FIN Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median Part. Median (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) (in k e ) EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

22 Results Distribution of Gross Wealth Ratios Ineq. Measures Shares P90/ P75/ P90/ Mean/ Top Top Bottom P10 P25 P50 Median Gini GE2 5% 10% 50% EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

23 Results Distribution of Gross Wealth Ratios Ineq. Measures Shares P90/ P75/ P90/ Mean/ Top Top Bottom P10 P25 P50 Median Gini GE2 5% 10% 50% EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

24 Results Distribution of Gross Wealth Ratios Ineq. Measures Shares P90/ P75/ P90/ Mean/ Top Top Bottom P10 P25 P50 Median Gini GE2 5% 10% 50% EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38

25 Results Natural Table Lindner (OeNB) HFCS 25 th September / 38

26 Results Additive, Contribution Table Lindner (OeNB) HFCS 25 th September / 38

27 Results Shapley Table Lindner (OeNB) HFCS 25 th September / 38

28 Results Additive, Elasticities Table Lindner (OeNB) HFCS 25 th September / 38

29 Results RIF Table Lindner (OeNB) HFCS 25 th September / 38

30 Conclusions Real Assets are major contributing factor to the distribution of wealth in absolute terms In relative terms less widely held asset types (also with lower median) gain importance Elasticities vary across countries and across asset categories Policies that are not considered to target distribution do affect it Impact of monetary policy Impact of support for HMR Different decomposition methods provide different insights Future research: Sub-group decomposition, maybe according to countries/regions Get some standard errors of the estimates Lindner (OeNB) HFCS 25 th September / 38

31 Thank you for your attention! Lindner (OeNB) HFCS 25 th September / 38

32 Appendix Lindner (OeNB) HFCS 25 th September / 38

33 Appendix: Motivation Benoît Cœuré Monetary policy interventions have distributional effects across the population depending on the composition of the assets and liabilities of households. One could argue that monetary policy is in essence distributional in the intertemporal dimension and the euro area crisis has uncovered distributional effects of monetary policy in the spatial dimension. Such effects are however unintended, or in other words, they are a means to an end, that is price stability. Central banks should steer away from distributional politics. But there can be a thin line between intended and unintended consequences, which makes it even more necessary to understand the cross-sectional effects of monetary policy interventions. Opening remarks at the HFC-Conference of the ECB (Frankfurt, 17 th October 2013); published on [accessed 22 nd September 2014] Back Lindner (OeNB) HFCS 25 th September / 38

34 Appendix: Results Natural Source: Eurosystem HFCS HMR FRE BUS REST SAFE RISKY REST REAL FIN EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Lindner (OeNB) HFCS 25 th September / 38 Back

35 Appendix: Results Additive, Contribution Source: Eurosystem HFCS HMR FRE BUS REST SAFE RISKY REST REAL FIN s k s k s k s k s k s k s k EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Back Lindner (OeNB) HFCS 25 th September / 38

36 Appendix: Results Additive, Elasticities HMR FRE BUS REST SAFE RISKY REST REAL FIN ε k ε k ε k ε k ε k ε k ε k EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Back Lindner (OeNB) HFCS 25 th September / 38

37 Appendix: Results Shapley Source: Eurosystem HFCS HMR FRE BUS REST SAFE RISKY REST REAL FIN EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Lindner (OeNB) HFCS 25 th September / 38 Back

38 Appendix: Results RIF HMR FRE BUS REST SAFE RISKY REST REAL FIN EA AT BE CY DE ES FI FR GR IT LU MT NL PT SI SK Source: Eurosystem HFCS Lindner (OeNB) HFCS 25 th September / 38 Back