1. Economic Integration and Mature Portfolios Dimitris Christelis CSEF, University of Salerno Dimitris Georgarakos Goethe University Frankfurt and CFS Michael Haliassos Goethe University Frankfurt, CFS, MEA

2. 2 Main Approaches in Integration Literature Integration has been assessed by looking at: Considerable international flows across markets. Prices: In a fully integrated market for goods, the law of one price should hold. In a fully integrated asset market, the price of risk should be the same. Consumption behavior: international risk sharing. In a fully integrated world, households would insure against output risks idiosyncratic to their countries by holding securities in other countries subject to different shocks.

3. 3 Motivation An economic agent of given preferences and characteristics optimizes subject to a set of processes (e.g. for labor income and asset returns) policies (e.g. for taxation or retirement financing) and constraints (e.g. credit market imperfections, informational limitations) Resulting policy rules (e.g. for asset demands) interact with the supply side and produce observed asset participation patterns asset holdings among participants.

4. 4 Motivation (ctd) With greater integration, greater similarity in market conditions facing households of given characteristics greater similarity in supply-side conditions harmonization of policies and institutions cultivation of common ‘culture’/preferences greater access to foreign markets

5. 5 Motivation (ctd) Thus, controlling for differences in population characteristics, greater integration should be reflected in greater similarity of the relationship between household characteristics and asset behavior as regards: Participation in asset markets Asset holdings among market participants Newly available data can be used to assess extent of similarity in this relationship Across countries Across regions Across assets

6. 6 ‘Mature’ Portfolios In principle, all portfolios interesting ‘Mature’ Portfolios: portfolios of those 50+ More time to accumulate Greater experience Influenced by retirement prospects Also: topical to study Financing retirement in view of demographic transition Prospects of ‘asset meltdown’

7. 7 This paper: Data Three comparable data sets, sharing common design (about 37000 HH) HRS: US 2004 ELSA: EN 2004 SHARE: First wave took place in 2004 in Sweden, Denmark, Germany, the Netherlands, France, Switzerland, Austria, Italy, Spain, and Greece; and in 2005 in Belgium.

8. 8 This paper: Issues We decompose differences in: Participation rates in assets: configuration of characteristics of populations influence of characteristics on participation probability Asset holdings among market participants: configuration of characteristics of asset holder pools influence of characteristics on asset levels Comparisons: US versus each European country examined Within US (Midwest as benchmark) Within Europe (Germany as benchmark)

9. 9 Assets Considered Financial: Stocks held directly or indirectly Real assets: Private businesses Primary residence

10. 10 Participation Rates Observed Differences among Older HHs

11. 11 Counterfactual decomposition Decomposition is of the following form: Actual difference Covariate Effects Coefficient Effects

12. 12 Decomposing Covariate from Coefficient Effects on Participation Construct counterfactual: average predicted probability of participation for the population in country i if they faced coefficients of the base country. First run probit in base country Regressors: 2nd order age polynomial, household size, education (high school dropout; high school degree; College degree), recall ability, self-reported bad health (includes responses ‘fair’ and ‘poor’ in HRS), number of ADL, work status (retired/working/unemployed- other inactive), marital status (couple/widow/never married), subjective probability to leave a positive bequest, whether has received an inheritance, whether HH provides help to relatives/neighbors, whether is involved in voluntary activities, income quartile, wealth quartile. Apply this set of coefficients to the sample of country i and compute average predicted probability. Draw (with replacement) the full sample size from both countries and repeat 100 times to compute bootstrap standard errors.

13. 13 Results on Europe-US comparisons Stockholding: Neither the market conditions faced by households in most European countries nor their characteristics are as conducive to participation in stockholding (direct or indirect) as those of the US. Private business: Market conditions are largely responsible for lower participation in private business in Europe than in the US. Most European older populations are estimated to have characteristics as conducive to business ownership as those of the US population. Homeownership: Coefficient effects are positive in most cases suggesting favorable US market conditions Exceptions: southern countries, EN (and small for BE) where households would have lower probabilities if faced with US market conditions. All covariate effects are positive, in favor of the US.

14. 14 Stockholding Participation

15. 15 Participation in Private Business

16. 16 Participation in Homeownership

17. 17 Comparison of Roles of Specific Factors We compute marginal effects of a ‘unit change’ in a given factor and compare them across assets and countries. We estimate marginal effects for each household in the country and then average across households Both estimates and simulated standard errors are reported

18. 18 Marginal Effects of Higher Education Degree

19. 19 Marginal Effects of Self-declared Bad Health

20. 20 Participation: Within the US Market conditions in the Midwest are more conducive to participation in any of these asset classes. Exception: the South is estimated to have even more favorable conditions for homeownership than the Midwest. Though statistically significant, estimated differences are rather small.

21. 21 Participation: Within Europe Differences in participation rates arise mainly from differences in market conditions rather than in population characteristics. With very few exceptions, coefficient effects are statistically significant and often quite large small for business ownership

22. 22 Results on Within- Integration

23. 23 Counterfactual decomposition Levels of Asset Holdings Decomposition is of the following form: Actual difference Covariate Effects Coefficient Effects

24. 24 Implementation Adapt a technique proposed by Machado and Mata (2005), following Albrecht et al. (2003) First run 19 quantile regressions at every 5 th percentile for each asset, on base country wners Regressors: 2nd order age polynomial, household size, education (high school dropout; high school degree; College degree), recall ability, self-reported bad health (includes responses ‘fair’ and ‘poor’ in HRS), number of ADL, work status (retired/working/unemployed-other inactive), marital status (couple/widow/never married), subjective probability to leave a positive bequest, whether has received an inheritance, whether HH provides help to relatives/neighbors, whether is involved in voluntary activities, income quartile, wealth quartile. For each percentile, draw (with replacement) full sample size from comparison country and estimate holdings if in base country: Thresholds for income and wealth quartiles are defined for the base- country over all older households. Comparison country households are then placed in quartiles according to those thresholds. Rank estimated holdings; compare with actual holdings

25. 25 Stockholding Levels Strong coefficient effects: European stockholders would achieve considerably higher levels of stock holdings if they were confronted with US market conditions. Covariate effects are small and mostly insignificant across percentiles.

26. 26 Counterfactuals Stockholding Levels, Direct and Indirect (similar pictures for FR, EN)

27. 27 Counterfactuals Stockholding Levels, Direct and Indirect (similar pictures for ES, GR)

28. 28 Counterfactuals Stockholding Levels, Direct and Indirect

29. 29 Private Business Holdings Most of the differences can be accounted for by differences in market conditions. If European private business holders were faced with US markets conditions, they would be holding lower amounts in private businesses. England represents the only case where business holders (in particular small ones) would hold higher amounts. Covariate effects are insignificant.

30. 30 Counterfactuals Private Business Wealth (similar pictures for SE, DE)

31. 31 Counterfactuals Private Business Wealth (similar pictures for ES, GR)

32. 32 Counterfactuals Private Business Wealth

33. 33 Primary Residence European homeowners in virtually all countries would have smaller holdings if they faced US market conditions. Coefficient effects are particularly strong and well exceed the overall differences in home values observed in England and in southern countries. By contrast, US homeowners have characteristics more conducive to large home equity values than European homeowners.

34. 34 Counterfactuals Housing Wealth (similar pictures for BE, AT)

35. 35 Counterfactuals Housing Wealth (similar pictures for ES, GR, EN)

36. 36 Amounts: Within US Coefficient effects suggest that households in different regions face: Similar market conditions with respect to stockholding Greater incidence of statistically significant coefficient effects for private businesses Even greater for the primary residence

37. 37 Amounts: Within Europe Stockholding: Vast majority of countries exhibit strongly significant coefficient effects relative to Germany Private Business: Households with small or medium holdings tend to face comparable market conditions across European countries Statistically and economically significant differences are observed for those with the largest holdings Homes: Coefficient effects are statistically significant, but their estimated size and sign exhibit considerable variation across European countries when compared to Germany A number of statistically significant covariate effects, implying that differences in home values arise partly from differences in characteristics of homeowners across European countries

38. 38 Results on Within- Integration

39. 39 Concluding remarks Evidence of quite limited integration between US and major European countries, but also within Europe Considerable variation in observed: Asset participation Levels of asset holdings among holders Coefficient effects typically more important factor for international differences in asset holdings, compared to differences in household characteristics