The Influence of Investors' Jobs on Portfolios: Is there a

The Influence of Investors’ Jobs on Portfolios:

Is there a Job Industry Bias?

Ralf Gerhardt∗

This version: February 27, 2008(working paper - comments welcome)

Abstract

The major component of most investors’ wealth is the discounted value of(non-tradeable) future labor income and therefore an important backgroundrisk which should be adequately hedged by financial portfolio risk.We test this hypothesis empirically by using a unique dataset of 30,000 privateinvestors of whom we know their profession as well as their detailed portfoliocomposition. Instead of lower ratios we find that investors hold significantlyhigher ratios of equity in their job industry - a job industry bias. Investors tendnot only to ignore but to aggravate background risk in their overall wealth byholding biased financial portfolios.Rational and financially sophisticated investors better hedge their labor in-come risk. In contrast, a short event study shows that unsophisticated in-vestors can not expect to reduce their job industry bias by taking financialadvice.

JEL classification: D12, D10, G29

Keywords: Background Risk, Labor Income, Household Portfolios, Asset Allocation, Be-

havioral Finance

∗E-Finance Lab & Johann Wolfgang Goethe-University, Chair of Banking and Finance,Mertonstr. 17, 60054 Frankfurt am Main. E-Mail: [email protected],phone +49.162.2658081, fax +49.69.798.28585.

The Influence of Investors’ Jobs on Portfolios 2

1 Introduction

The largest component and thus the largest background risk in most households’ wealth is

the discounted value of future labor income i.e. non-tradeable human capital. Theoretical

models and empirical evidence suggest that dividends on human capital (labor income)

and dividends on stocks are correlated or at least cointegrated so it is of great impor-

tance for investors to adequately hedge this background risk when choosing the financial

portfolio to reduce risk exposure of overall wealth. Rational investors following financial

theory should hold a lower portion of their financial portfolio in their job industry than

peers working in other industries.

Using a dataset of almost 30,000 individual investors’ disaggregated portfolios with a

large German direct bank we are able to check empirically whether private investors act

according to financial theory in respect of hedging labor income risk. We construct five

industries (Financial Services, IT, Health Care, Technology and Aviation) and assign per-

sons’ professions as well as securities resulting in a final sample of 4,395 private investors

working in one of these industries. Then we compute the five industry ratios in each

investor’s equity portfolio. Instead of lower holdings - as one would expect from financial

theory - investors exhibit much higher holdings of equity in their respective job industry

across all five industries. This bias which we shall call job industry bias is robust against

several variations and indicates that most investors not only ignore background risk in

their financial portfolios but even aggravate it.

Further analysis shows that financial sophistication and rationality significantly reduce

the bias. This supports the assumption that investors have no rational motif to hold high

ratios of their job industry in financial portfolios but simply act irrationally and deviate

from financial theory - a common behavior that Campbell (2006) calls investment mis-

takes. Thus a job industry bias is very much in line with various findings in the behavioral

finance literature like home bias, overtrading, the disposition effect and many more.

Having found that financial sophistication reduces a job industry bias and thus sophisti-

cated investors better hedge their background risk, it is interesting to see if unsophisti-

cated investors can improve their portfolio by taking financial advice. Regression analyses

indeed show that advised customers exhibit a lower industry bias than non-advised cus-

tomers. However a small event study among the sub-sample of investors that switched

from acting self-directed to advised enables us to isolate the effect of advisors on portfo-

lios. It turns out that advised investors better hedge their background risk on their own

and not because of the influence of an advisor: The causality seems to be that better

hedged investors also take financial advice and not that financial advice leads to better

hedged background risks. So unsophisticated investors can not expect to improve their


portfolios in regard to a better hedge of background risk by taking financial advice.

To our knowledge this is the first empirical study - at least for German customers - that

concentrates on job industries to analyze whether private investors consider their back-

ground risk from labor income when choosing the financial portfolio. We contribute to the

literature on background risk and portfolio choice in providing another dimension where

investors significantly deviate from financial theory and make an investment mistake: Not

only ignorance but aggravation of background risks by their financial portfolio choice.

The rest of the paper is organized as follows: Section 2 reviews related work. Section 3

describes the empirical evidence for a job industry bias as well as its quantification and

determinants. Section 4 analyzes the impact of financial advice on hedging labor income,

section 5 concludes.

2 Theory and literature review

2.1 Human capital and portfolio choice

For most households, (non-tradeable) human capital is the largest component of wealth

as Campbell (2006) states. So it is not surprising that Cocco et al. (2005) showed that

ignoring the effects of background risk in portfolio composition generates large utility

costs for investors.

Literature on background risk like Heaton and Lucas (2000b) mostly identifies three ma-

jor components of background risk: (1) Most studies (see e.g. Heaton and Lucas (1997) or

Campbell and Viceira (2002)) focus on labor income as the primary source of background

risk. (2) The effects of housing on portfolio choice is still debated in the literature since

it can be seen as non-tradeable asset and therefore as background risk (see e.g. Cocco

(2005)) but also as hedge against renting cost risks as e.g. Sinai and Souleles (2005) show.

Since it is not yet clear which effect housing has on background risk we exclude housing

aspects here. (3) The last major component of background risk is entrepreneurial income.

Heaton and Lucas (2000a) found empirically that it leads to lower stock holdings as these

investors realize their background risk and adjust their financial portfolios to hedge their

entrepreneurial income. To further focus on the effects of industry-specific labor income,

we exclude entrepreneurs from the sample so that labor income remains as the main com-

ponent of background risk in our analysis.

Depending on the type of labor income it can be seen as rather riskless (e.g. for civil

servants) or risky (e.g. artists) asset and there is still debate in the literature on the risk


properties (Campbell (2006)). Depending on the riskiness of an investor’s labor income

it can be seen as (risky) stock or (low risk) bond investment paying dividends on human

wealth. Campbell and Viceira (2002) show that investors should adjust their financial

portfolio to compensate for implicitly holding human wealth and conclude that human

capital can crowd out certain assets. This effect is stronger for young investors as their

total wealth consists to larger parts of human capital and therefore should have stronger

impact on financial portfolio choice.

Labor income risk and stock market risk are likely to be correlated since both are influ-

enced by the same business cycle conditions and - in this specific analysis - furthermore

subject to the same industry business conditions. Unfortunately, empirical studies on

these correlations are very rare. Analyzing U.S. data from 1965 to 1994, Davis and

Willen (2000) find both small negative as well as positive correlations between wages and

security returns in specific industries.

According to Campbell et al. (1999), these correlations increase if labor income is used as

a 1-year lagged variable so that the authors assume a stable correlation between market

returns and human capital returns in their model. In a very recent article, Benzoni et al.

(2007) looked at various research evidence on the correlation of stock market and human

capital returns. They prove empirically that the two markets are cointegrated and show

mathematically that the assumption of cointegration allows for much higher correlation

between the two markets.

Hence we can assume that labor income and stock market returns are correlated and

follow that this correlation should be even stronger if focusing on a specific industry.

2.2 Related research on job industry weights in portfolios

Related research concentrating on job industries as well as portfolio industries is very rare

which is why we try to fill this gap by providing empirical evidence of private investor

behavior in this field.

In a recent study on Swedish households, Massa and Simonov (2006) find that investors

tilt their portfolios (relative to the Swedish stock market index) to stocks that are familiar

to them. They define familiarity as stocks with geographical and professional proximity

to the investors and stocks that have been held for a long time. So a partial result of their

analysis is that professional proximity leads to overweighting in the respective stock indus-

tries. The study is mainly focused on entrepreneurial income and geographical proximity

and the authors find only weak evidence of a professional proximity effect. Nevertheless

their findings indicate that investors do not hedge their background risk in financial port-

folios despite theory tells us that this would decrease overall portfolio risk exposure.


3 Empirical analysis

3.1 Data description

We use a unique data set of nearly 30,000 customers of a large German direct bank. The

portfolios of individual customers are available at security level dated February 2007 which

allows for a very detailed and up-to-date analysis. Table 10 shows some demographics

indicating that investors in the sample are a bit younger, more male and have higher risk

tolerance than other retail banking customers in previous German studies (see Bluethgen

et al. (2007) for example). However our sample consists of direct bank customers so that

these deviations are plausible. Furthermore all investor characteristics in our sample are

very much consistent with investors in other studies on German direct banking customers

like Glaser (2003). So our sample is not exactly representative for the typical German

retail banking customer but very much for the typical German online banking customer.

Table 10 also includes some portfolio characteristics: Dummy variables indicate if a per-

son is classified as heavy trader by the bank’s data warehouse (indicating high trade

frequency) or Eurex trader which shows that the investor is allowed to trade futures,

options and other derivatives. Deposit and cash values are computed as means over the

last six months to account for short-term fluctuation. Average return p.a. and Sharpe

Ratio are computed over the last 20 months. Number of securities is a simple proxy

for naıve diversification following Benartzi and Thaler (2001) and Mitton and Vorkink

(2007). Other measures for diversification are the Herfindahl-Hirschmann-Index (HHI)

used e.g. by Blume and Friend (1975) or Dorn and Huberman (2005), the ratio of funds

in the equity portfolio following Guiso and Jappelli (2005b), the ratio of international

equity - which is non-German equity here - in the equity portfolio following Bluethgen et

al. (2007) and the share of index funds in the portfolio.

Many securities were classified into certain industries in the original data sample already

by the bank’s internal system. Additionally, we used stock industries from vwd, a large

German market data supplier and added mutual fund industries from Feri Rating & Re-

search, a large German rating company and data supplier. As most bonds, money market

funds, and other securities cannot be assigned to a specific industry, this study focuses

on the equity portfolio of investors. The average equity share in our sample is above 85%

so that no major wealth components regarding industry allocation are ignored.

After adding security industries from the mentioned data sources, 6,325 out of 7,982 secu-

rities could be matched with an industry. The coverage is not bad regarding the fact that

many mutual funds have no clear industry focus but invest in regions or tax-optimized

assets.


Table 1:Industry assignment

Investor professions and security sectors are assigned to five industries. Security sectors aresometimes overlapping because they are extracted from three data sources and some Germandescriptions have no explicit translation. Italic security sectors are excluded from the analysislater when checking for robustness.

Industry Professions Security Sectors

Fin. Services banker banks, online brokers, insurers, financial service providers,financial sector, investment companies

IT computer computers/software, it-services, it, internet, internet servicescientist providers, phone/communication, it hardware,

information services

Health Care doctor, pharma/cosmetics/gene technology, health services,pharmacist health care, bio technology, chemistry/pharma

Technology engineer, automotive, electrics/electronics, energy technology, vehicletechnician, and machine manufacturing, aircraft construction, opticalelectrician, instruments/measurement, solar energy, environmentmotorcar engineering, technology, shipbuilding, semi-conductors,mechanic nanotechnology, defense technology, bio-energy

Aviation pilot airlines

The professions and security sectors are grouped into five industries: Financial Services,

IT, Health Care, Technology and Aviation. The respective assignments are shown in table

1.

We can not tell if the investors in our sample hold all their wealth in the securities ac-

counts in our sample or if they have considerable funds in other bank accounts. But the

size of securities accounts and cash holdings is quite close to the numbers in other studies

on German customers who could rule out deposits with other banks.

3.2 Hedging labor income risk

As shown in section 2.1, labor income and market returns are positively correlated so

that investors acting rationally and thus according to financial theory should hedge their

industry exposure when choosing the financial portfolio. The resulting portfolios should

have significantly lower ratios of equity from the respective job industry than the portfo-

lios of peers working in other industries. This leads to the first hypothesis:


Hypothesis 1: Investors hold a lower portion of equity from the industry they work in

compared to investors working in other industries.

3.2.1 Job Industry Bias

To test this hypothesis empirically we accumulate the equity portfolios of all investors

working in the same industry which leads to five aggregated industry investors and im-

plicitly comprises a value-weighting of the individuals’ portfolios. Then we look at the

industry ratios in each of the five aggregated investors and build a table that displays

aggregated investors in columns and their respective ratios of industry equity holdings in

rows.

Table 2:Analysis with five aggregated investors (value weighted)

Accumulating the equity portfolios of all investors that work in the same industry leadsto five aggregated industry investors and implicitly comprises a value-weighting of theindividuals’ portfolios. It should be read row by row to see that investors working in a spe-cific industry invest more in ’their own’ industry than their peers working in the other industries.

Job IndustryAsset Industry Financial Serv. IT Health Care Technology Pilots Obs.

Financial Services 15.44% 5.14% 8.51% 7.73% 8.39% 329IT 9.67% 20.96% 8.83% 13.34% 11.75% 355Health Care 1.37% 1.74% 3.32% 1.89% 2.02% 1,206Technology 10.85% 11.03% 8.11% 16.88% 10.11% 2,461Aviation 0.33% 0.66% 0.44% 0.92% 7.31% 44

Total 37.65% 39.53% 29.20% 40.75% 39.59% 4,395

Table 2 shows that the ratios on the main diagonal (bold) are much higher than the other

values in the same row. The first row for example tells us that investors working in the

Financial Services industry hold 15.44% of their equity in Financial Services assets while

investors working in other industries hold only between 5.14% and 8.51% in this industry.

Rational investors should hold less in their job industry compared to their peers so one

would expect a ratio below 8%.

Instead table 2 shows exactly the opposite: All five industry investors hold much higher

ratios in their job industry, most of them about twice as high as their peers working in

other industries.


For investors working in the technology industry, overweighting the job industry in the

portfolio - which we call a job industry bias - is not twice as high but still clear and large.

This might be due to the fact that this industry comprises many different professions so

that the effect is blurred since a more exact matching of professions and securities was

not possible. Still, even despite this blurring effect in technology, a strong job industry

bias is obvious.

Thus an industry bias could show up stronger the more selective our industry assignment

gets. This assumption is supported when we look at a very selective industry - aviation.

It consists only of pilots and airline stocks and shows the most extreme job industry bias:

Pilots allocate ten times more equity to airline companies than their peers. The numbers

in aviation have to be interpreted with care as they are based on only 44 observations.

Nevertheless, the bias is enormously strong and further analysis will show that the effect

remains strong even if we account for the small number of observations.

To tighten the above strong results, we conduct two robustness checks: First we exclude

some security classes from industry composition to make sure that the results are ro-

bust against variations in the industry definitions. Table 7 shows only marginal changes

in the results so we can assume that the job industry bias is robust for industry definition.

Table 3:Analysis with individual investors (unweighted)

Ratios in this table are the equally-weighted means of all investors’ portfolios in the respectiveindustry. Thus small investors with usually extremer ratios compared to big diversifiedportfolios enter with the same weight.



Total 36.51% 36.40% 29.36% 38.04% 33.72% 4,395

As a second check, we calculate the matrix in table 2 without value-weighting the port-

folios. The results shown in table 3 are again only marginally weaker than the value-

weighted numbers. A possible explanation is that investors with small portfolios and only

one stock (thus 0% or 100% industry ratios) for example enter the calculation with the


same weight as a EUR 100,000 investor holding a well-diversified portfolio. But still, the

effect is clearly visible across all five industries indicating robustness in the results.

So we found clear and strong evidence that investors do not hold lower ratios of equity in

their job industry but much higher portions. Investors not only fail to hedge their labor

income, they even increase their overall wealth exposure to industry risk when selecting

their financial portfolios.

A possible explanation for these high holdings in the job industry are employer stocks.

Benartzi (2001) analyzes retirement savings accounts and finds that the majority holds

large fractions (10% to 40%) of company stock. The information of investors’ employer is

not included in our dataset so we can not quantify the exact effect of employer stockhold-

ing plans. On the other hand, pension plans and employer stocks are much less popular in

Germany: Skiera (2007) found in a representative study among 20,000 German households

that only 1.5% hold employer stock. Furthermore, our analysis is not based on pension

plan portfolios but trading portfolios of a large direct bank so we can safely assume that

the influence of employer stock is marginal.

Nevertheless, this explanation does not affect the recommendation from financial theory

to minimize correlations between the assets of an investor’s total wealth - regardless of

the fact that the stocks might have been bought using a stock pension plan.

Another rationale for a job industry bias could be private information on companies in

the investors’ job industry. But this private information can be positive as well as nega-

tive information on a company so that the resulting decision is to sell/short-sell or buy

a stock from the job industry. Thus private information can be ruled out as motif for

overweighting the job industry in the financial portfolio.

So there is strong and robust evidence that investors exhibit a job industry bias; they

do not hold lower portions of their job industry but much higher portions. H1 is clearly

rejected by our portfolio data.

There are two possible reasons why the portfolio data does not confirm H1. First, investors

do act rationally and the theoretical recommendation to hedge background risk with the

financial portfolio is simply wrong. This seems unlikely given all the consistent research

that has been conducted in the area. A second possible explanation is that investors act

irrationally and not according to financial theory and therefore deviate significantly from

textbook recommendations. Section 3.3 will further analyze these two possibilities.


3.2.2 Quantification

Besides the fact that there is strong overweighting of the own job industry in investors’

portfolios, it would be interesting to have a clear overweight measure over all 4,395 in-

vestors in the five industries. To compute such a measure, a benchmark for every single

investor is necessary. A basis for this benchmark could be the share of the respective

industry in the market portfolio. Holding the same industry ratio as in the market port-

folio does of course not include a hedge against labor income (background) risk. However,

as there are no figures of correlations between the respective job and the asset industry

available, the task to compute a correct individual optimal ratio is almost impossible.

Nevertheless the respective industry ratio of the market portfolio represents a maximum

ratio that investors should hold in a respective industry. It serves as a good relative

measure of overweighting the own job industry or job industry bias: Investors can be

compared across job industries according to their distance to the benchmark where a

value of zero represents a job-neutral financial portfolio i.e. investors do not hedge their

labor income risk.

But even the exact industry shares of the market portfolio are impossible to assess so that

we approximate this industry-specific figure by the average of all other customers in the

sample. To give an example, we exclude the 329 persons working in the financial services

industry from our overall sample of 29,184 investors and see that the remaining 28,855

have on average 6.54% of their equity portfolio in the financial services industry. Then we

take the individual ratio of equity in financial services over all equity for each of the 329

investors working in the financial services industry, subtract the benchmark of 6.54% and

get 329 values for overweight. This calculation is done for all five industries and 4,395

investors.

As a robustness check, we also compute the benchmark of all other investors from the

other four industries. Table 4 shows that the two benchmarks are quite close to each other

and the investors in the respective industries are on average high above the benchmark -

consistent with our previous findings.

The variable overweight is computed for all 4,395 customers and should be significantly

below zero if investors act according to financial theory and hedge their job industry risk.

Once more, empirical evidence is in sharp contrast to theoretical recommendations. A t-

test on the portfolio data shows that overweight is significantly positive with a mean of 4.1

percentage points above the respective industry benchmark and a p-value of 0.0%. This

underlines previous evidence for a clear and strong job industry bias in investors portfolios.


Table 4:Benchmarks and industry weights

The first column shows investors’ value-weighted portfolio ratios in their own job industry asdisplayed on the main diagonal in table 2. The second column shows the unweighted ratios asdisplayed on the main diagonal in table 3. Columns two and three are used to calculate theoverweight for each investor.

Asset Industry Own Industry Own Industry Benchmark of Benchmark of(aggregated) (indiv. Investors) all other investors 4 other industries

Fin. Services 15.44% 13.11% 6.54% 6.18%IT 20.96% 16.29% 11.72% 11.37%Health Care 3.32% 3.07% 1.86% 1.58%Technology 16.88% 15.58% 10.56% 9.75%Aviation 7.31% 7.65% 0.66% 0.66%

3.2.3 Determinants

Having built a measure for job industry bias, regressing demographics and portfolio char-

acteristics can provide further information on the question what drives a job industry bias

i.e. not only ignorance but aggravation of the background risk in overall wealth.

We include a gender dummy for male investors because a possible explanation for over-

weighting the job industry is that investors are overconfident in their ability to pick

winning stocks in their job industry and gender is ”a natural proxy for overconfidence”

according to Barber and Odean (2001). Furthermore, behavioral finance tells us that

women generally act differently from men in their financial decisions.

To see if investors’ age has an effect on industry bias is of interest because younger and

older investors differ in their implicit holdings of human capital and thus also vary in their

need to hedge labor income risk. We also include investors’ self-reported risk tolerance on

a scale of 1 (investor wants to avoid any risk) to 6 (investor accepts extremely high risk)

to see if risk-averse investors tend to better hedge their labor income risk.

To see if investors’ portfolio characteristics like deposit value (depvalue) and the share of

equity have influence on an industry bias, we also include these two variables. The share

of international equity in the equity portfolio (intdiv) serves as a measure for diversifica-

tion to see if investors who care about diversification in their financial portfolio also better

hedge their background risk and thus gain better diversification in their overall wealth.

The dummies for heavy trader and Eurex trader are not included because these persons

are only a small fraction of the sample and exhibit high correlation with the variable risk

tolerance.


So our simple model is specified:

Overweight = α + β1male + β2age + β3risk + β4depvalue + β5equityshare + β6intdiv + ε

The results of the regression are displayed in table 5 on page 14. Because two more vari-

ables will be added later, factor loadings and p-values are noted in the first of the three

columns.

Table 5 shows that gender has no significant effect on background risk hedging; men and

women are equally affected by a job industry bias. So overconfidence in stockpicking

abilities in the job industry does not seem to be the main reason for high holdings of job

industry stock. Interestingly, older investors show more industry bias. With increasing

age, the negative effect of overweighting the job industry diminishes because the human

capital shrinks and the need to hedge labor income risk also decreases. Nevertheless it

seems unlikely that older investors consider this fact. The best explanation is probably

that older persons worked for a longer time in their industry and invest in companies they

are familiar with.

The significant negative influence of risk tolerance on overweighting is surprising as one

would suspect that investors with low risk tolerance should try harder to hedge their labor

income risk than investors willing to bear higher risks. On the other hand this is just

one more piece of evidence indicating that most investors do not understand the need to

hedge their labor income.

The measure of geographical diversification is highly significant and reduces a job industry

bias. It is plausible that investors who care and know about the importance of diversi-

fication in their financial portfolios are also more likely to perceive the need to diversify

their background risk.

The positive and highly significant loading of β5 demonstrates that portfolio structure is

an important determinant of job industry bias. Investors holding high ratios of stocks

and stock mutual funds in their portfolios tend to overweight their own job industry.

Summing up we find in this first regression that age and equity share in investors portfolios’

drive the probability to overweight the job industry while diversification motives and risk

tolerance reduce it. In the following sections, two more explaining variables will be added

to get a clearer picture of the effects.


3.3 Rationality, education and sophistication

This section analyzes the question whether investors’ rationality and sophistication affect

their job industry bias. The result will shed more light on the reason why H1 could not

be confirmed: Is financial theory mistaken or do investors act irrational and thus not

according to financial theory?

We take a look at a sub-sample of 962 investors that hold a Dr. (Ph.D.) or Professor’s

degree indicating the highest level of education. These persons are likely to better under-

stand complex relations between diversification, risk, and hedging. Furthermore, working

in academia for years should also increase the rationality of these persons so that we take

high education as a proxy for rationality.

So if H1 could not be confirmed because investors do not act rationally and thus according

to financial theory, the subsample of 962 Dr.’s and Prof.’s should exhibit a significantly

lower job industry bias than the other 3,433 investors. This leads to a second hypothesis:

Hypothesis 2: Investors holding a Dr. or Prof. degree act more according to financial

theory and thus have less job industry bias than the other investors in the sample.

A t-test shows that the individual overweights for the group of Dr.’s and Prof.’s (avg.

2.19%) are with a p-value of 0.1% lower than the overweights of investors who are not

Dr.’s or Prof.’s (avg. 4.62%). To check this result for robustness against the assumed

normal distribution, we conduct a median difference test and a Wilcoxon Mann-Whitney

test. Both non-parametric tests confirm with a p-value of 0.0% that Dr.’s and Prof.’s

have lower job industry bias in their portfolios than their peers.

Additionally, we add a dummy variable ’DrProf’ to our regression model to see if the

effect remains if we control for the other variables. The estimation is displayed in column

2 of table 5 and shows that the dummy has negative loading and is highly significant with

a p-value of 0.1%. Hence H2 is supported by comparison of two subsamples as well as a

regression model.

As most of the Dr.’s and Prof.’s work in the Health Care industry, there are not enough

observations in the other industries to set up a matrix like in tables 2 or 3 to demonstrate

the effect. Nevertheless the above tests indicate that more rational investors have a sig-

nificantly lower job industry bias.

So financial theory behind H1 is not challenged and the rejection of H1 is due to irrational

behavior of investors which is very much in line with household finance and behavioral

finance literature.


Table 5:Determinants of a job industry bias

Risk tolerance is reported by investors themselves when opening their account on a scalefrom 1 to 6. Share of international equity in the investor’s equity portfolio is a measure fordiversification. The dummy variable DrProf comprises only persons with a Dr. or Prof. degree.All values were computed using heteroskedasticity-robust standard errors. VIF values below1.08 and the correlation coefficients in table 8 demonstrate that the values are not affected bymulti-collinearity.

(1) (2) (3)Overweight Overweight Overweight

Gender (male) 0.0105 0.00803 0.00702(0.260) (0.390) (0.450)

Age 0.000585** 0.000673** 0.000723**(0.048) (0.024) (0.015)

Risk tolerance -0.00613** -0.00613** -0.00653***(0.014) (0.014) (0.009)

Deposit value in TEUR 0.0000311 0.0000386 0.0000475*(0.230) (0.130) (0.069)

Share of equity in portfolio 0.0905*** 0.0903*** 0.0827***(0.000) (0.000) (0.000)

Share of int. equity in equity portfolio -0.175*** -0.174*** -0.172***(0.000) (0.000) (0.000)

DrProf (dummy) -0.0202*** -0.0195***(0.001) (0.002)

Advised (dummy) -0.0282***(0.000)

Constant 0.0372 0.0389 0.0478**(0.110) (0.094) (0.042)

Observations 4389 4389 4389

R2 0.09 0.09 0.10

Robust p values in parentheses*** p<0.01, ** p<0.05, * p<0.1


4 Financial Advice and background risk

Empirical studies (see e.g. Dorn and Huberman (2002) or Guiso and Jappelli (2005a))

found that education and wealth are good proxies for financial sophistication. Looking

at the portfolio values in the sub-sample of Dr.’s and Prof.’s compared to the other 3,433

investors shows that Dr.’s an Prof.’s are with an average of 98,455 EUR much wealthier

than the other investors who have an average of 69,283 EUR in their portfolios. So it

is reasonable that the group of Dr.’s and Prof.’s seems to be not only more rational but

also more financially sophisticated than their fellow investors and thus exhibit a lower job

industry bias, i.e. better hedge the job industry risk in their portfolios.

This leads to the interesting question for alternatives for unsophisticated investors. Does

professional financial advice better integrate background risk in investors’ portfolios and

reduce a job industry bias? Financial advisors should consider their clients’ background

risk and thus hedge the labor income risk in the financial portfolios:

Hypothesis 3: Financial advisors consider their clients’ background risks which leads to

lower ratios of equity from the respective job industry in advised portfolios.

4.1 Sub-samples

To test this hypothesis, we again split our 4,395 customers in two sub-samples of 604

advised and 3,791 non-advised persons. Then we conduct a t-test and find that the vari-

able overweight is with a p-value of 0.0% lower for advised customers (avg. -0.2%) than

for non-advised customers (avg. 5.8%). Again we also conduct a median-difference and

Wilcoxon Mann-Whitney test to relax the assumption of normally distributed differences

between the groups and confirm the results of the t-test with p-values of 0.0%.

Furthermore we add a dummy for advised in our regression model and see in column 3

of table 5 that the dummy variable is with a p-value of 0.0% highly significant and has a

negative effect on overweighting the job industry. This supports the finding that financial

advice significantly reduces a job industry bias.

So the results indicate that financial advisors adjust their clients’ portfolios to hedge their

respective labor income risk and H3 can be confirmed.

However the lower industry bias of advised clients is not necessarily due to the advisors’

capabilities: It is also possible that sophisticated investors realize the need to hedge their

background risk themselves, thus do not exhibit a job industry bias but decide to take

financial advice. This could be due to a lack of time to deal with their financial affairs

in detail, for example. So the problem is that we can not rule out that the 604 advised

investors better hedge labor income risk regardless of their decision to take financial ad-

vice.


4.2 Event study

Econometric methods like the usage of instrumental variables or (propensity score) match-

ing first seem appropriate to shed light on this issue. However, concerning instrumental

variables, it is very difficult to identify and measure an instrumental variable in the

dataset. The matching method is common in medical or economic literature and is mainly

used to avoid a self-selection bias in managerial literature.

Regarding the very case of financial advice the question is not if more sophisticated in-

vestors self-select themselves to take financial advice but rather the chronology of events:

Neither method can distinguish between investors holding well-diversified portfolios who

decided to take financial advice and investors holding poorly diversified portfolios that

took financial advice and because of that now also exhibit well-diversified portfolios. It

can not be ruled out that the advised customers held well-diversified portfolios already

before taking advice. Thus these methods are not appropriate to reliably determine the

effect of financial advice.

Fortunately, our sample allows us to deal with this special endogeneity or causality prob-

lem in a very elegant way. The data comprises not only the disaggregated deposit holdings

in February 2007 but also monthly holdings from 01/2000 to 09/2007. Furthermore we

have the change dates of investors that decided to stop acting self-directed and take finan-

cial advice in the time between 02/2006 and 07/2007. So we are able to conduct a small

but selective event study. Even a strongly reduced sample size should suffice to support

or doubt the finding that financial advisors significantly reduce investors’ job industry

bias.

The selectiveness of this small subsample allows for a clear isolation of the advisor’s effect:

It enables us to analyze the portfolios of the same persons before and after the decision

to take financial advice.1

Out of the 4,395 investors working in one of the selected industries, 58 changed once

from self-directed to advised. Investors that changed their advice-status more than once

between 02/2006 and 07/2007 have been excluded form the study.

To see if the portfolios exhibit a lower job industry bias after taking advice, two points in

time ( ’pre’ and ’post’) are selected. Then the value for overweight before taking advice

(pre) is subtracted from the value after taking advice (post) and should be significantly

negative if financial advisors substantially reduce a job industry bias in their clients’ port-

folios. To check the results for robustness against changes in the time window, pre- and

1Our dataset comprises more than 700 persons that switched from self-directed to advised and issubject of another paper focusing on various effects of financial advice. Here we only consider investorsworking in one of our five selected industries and the effect on overweight.


post date are varied. The month of the decision to take advice is denoted 0 (and thus

individual for every investor) while pre- and post months are denoted with negative or

positive month numbers, respectively.

Table 6:Event Study

This table displays the difference in industry overweights of 58 investors before (pre) and after(post) each of them decided to take financial advice. If advisors reduce their clients’ job industrybias, the value ’overweight post - pre’ should be significantly negative. The difference is computedfor several points in time to see if results are robust to window changes.

Months Overweight Significance InvestorsPre Post Pre Post Post - Pre p-value (Obs.)

-1 1 2.55% 5.06% 2.51% 9.33% 58-1 2 2.55% 5.13% 2.57% 9.08% 58-1 3 2.67% 1.81% -0.86% 89.34% 55-1 4 3.17% 2.56% -0.61% 90.31% 53

-2 1 3.70% 5.02% 1.32% 27.97% 57-2 2 3.70% 5.07% 1.37% 27.49% 57-2 3 3.87% 2.04% -1.84% 89.33% 54-2 4 4.43% 2.81% -1.62% 86.74% 52

-3 1 3.81% 5.02% 1.21% 29.94% 57-3 2 3.81% 5.07% 1.26% 29.48% 57-3 3 4.09% 2.04% -2.06% 90.66% 54-3 4 4.66% 2.81% -1.85% 89.09% 52

Table 6 does not show the expected significant negative overweight differences. In fact,

the industry bias seems to increase slightly after taking financial advice and then de-

crease after the third month. Furthermore, almost all values are not significant - only

two out of 15 values show significance at the 10% level. Because of these insignificances

we can not say that the job industry bias rises after taking advice. However table 6

shows that there is no significant reduction in home bias after taking financial advice and

this should be the case if advisors correctly consider the background risk of their clients.

The positive levels of overweight also rule out that we see no effect because these 58 in-

vestors do not have a job industry bias. It can also be ruled out that certain customers

have been marked as adviced in the dataset by the bank despite they have no or rare

contact with a financial advisor: These persons actively decided to take financial advice,

signed a respective contract and speak regularly to their personal advisor (no callcenters).

So we first found in section 4.1 that advised investors exhibit lower job industry bias. A

second analysis to isolate the effect of financial advice showed that advisors do not reduce

their clients job industry bias. The evidence indicates that the group of advised investors


better hedge their background risk on their own and - independently - decided to take fi-

nancial advice that does not mitigate their job industry bias. This could be because these

sophisticated investors do not have enough time to take care of their financial affairs and

seek support from a financial advisor for example.

In the end, H3 can not be confirmed - the lower bias in advised investors’ portfolios can

not be credited to the advisors but to the investors themselves. Thus financial advisors

do not adequately consider their clients’ background risks in financial portfolios.

5 Conclusion

This study dealt with the question whether the industry an investor works in influences

his or her financial portfolio allocations. Rational agents should act according to financial

theory and hedge their labor income (background) risk in their financial portfolios. Hu-

man capital is the major component in most households’ wealth and dividends on human

capital - labor income - are correlated to stock market returns as the literature shows.

Thus rational investors should hold a lower portion of their financial portfolio in their job

industry than peers working in other industries.

Based on profession and portfolio data of almost 30,000 persons, we constructed five in-

dustries and allocated professions of 4,395 investors and securities, mainly stocks and

stock mutual funds with a clear industry focus. Analyzing the 4,395 portfolios we found

that investors allocate not less assets in their job industry but much more. This job in-

dustry bias in investors’ portfolios is in sharp contrast to financial theory and provides

one more case where investors do not act rationally and bear unnecessary high risk.

We further showed that financial sophistication, proxied by education and wealth, signif-

icantly reduces the bias. Financially sophisticated investors better hedge the background

risk in their financial portfolios which is in line with prior research finding that sophisti-

cated investors achieve better performance.

Furthermore we studied if unsophisticated investors can improve their poor situation and

reduce their job industry bias by taking financial advice. A regression indeed showed that

advised investors have lower job industry bias than non-advised investors. However we

could isolate the effect of advisors in a little event study and found that advised investors

better hedge their background risk in their portfolios anyway and not because the advisor

told them. This leads to the unpleasant conclusion that unsophisticated investors can not

adequately hedge their background risk and can not improve their situation by taking

professional financial advice.


References

Barber, B. M. and T. Odean, “Boys Will Be Boys: Gender, Overconfidence, And

Common Stock Investment”, Quarterly Journal of Economics, 2001, 116 (1), 261–292.

Benartzi, S., “Excessive Extrapolation and the Allocation of 401 (k) Accounts to Com-

pany Stock”, The Journal of Finance, 2001, 56 (5), 1747–1764.

Benartzi, S. and R.H. Thaler, “Naive Diversification Strategies in Defined Contribu-

tion Saving Plans”, American Economic Review, 2001, 91 (1), 79–98.

Benzoni, L., P. Collin-Dufresne, and R.S. Goldstein, “”Portfolio Choice over the

Life-Cycle when the Stock and Labor Markets Are Cointegrated””, Journal of Finance,

2007, 62 (5), 2123–2167.

Bluethgen, R., A. Gintschel, A. Hackethal, and A. Mueller, “Financial Advice

and Individual Investors’ Portfolios”, August 2007. Working Paper.

Blume, M. E. and I. Friend, “The Asset Structure of Individual Portfolios and Some

Implications for Utility Functions”, Journal of Finance, 1975, 30 (2), 585–603.

Campbell, J. Y., “Household Finance”, Journal of Finance, 2006, 61 (4), 1553–1604.

Campbell, J. Y. and L. M. Viceira, Strategic Asset Allocation.: Portfolio Choice for

Long-Term Investors, Oxford University Press, 2002.

Campbell, J.Y., J.F. Cocco, F.J. Gomes, and P.J. Maenhout, “Investing Re-

tirement Wealth: A Life-Cycle Model”, March 1999. Harvard Institute of Economic

Research Paper No. 1896.

Cocco, J. F., “Portfolio Choice in the Presence of Housing”, Review of Financial Studies,

2005, 18 (2), 535–567.

Cocco, J. F., F. J. Gomes, and P. J. Maenhout, “Consumption and Portfolio Choice

over the Life Cycle”, Review of Financial Studies, 2005, 18 (2), 491–533.

Davis, S.J. and P. Willen, “Using financial assets to hedge labor income risks: Esti-

mating the benefits”, March 2000. Working Paper.

Dorn, D. and G. Huberman, “Who trades?”, November 2002. Working Paper.

Dorn, D. and G. Huberman, “Talk and Action: What Individual Investors Say and

What They Do”, Review of Finance, 2005, 9 (4), 437–481.

Glaser, M., “Online Broker Investors: Demographic Information, Investment Strategy,

Portfolio Positions, and Trading Activity”, October 2003. Working Paper.


Guiso, L. and T. Jappelli, “Awareness and Stock Market Participation”, Review of

Finance, 2005, 9 (4), 537–567.

Guiso, L. and T. Jappelli, “Household Portfolio Choice and Diversification Strategies”,

July 2005. Working Paper.

Heaton, J. and D. Lucas, “Market Frictions, Savings Behavior, and Portfolio Choice.”,

Macroeconomic Dynamics, 1997, 1 (1), 76–101.

Heaton, J. and D. Lucas, “Portfolio Choice and Asset Prices: The Importance of

Entrepreneurial Risk”, Journal of Finance, 2000, 55 (3), 1163–1198.

Heaton, J. and D. Lucas, “Portfolio Choice in the Presence of Background Risk”, The

Economic Journal, 2000, 110 (460), 1–26.

Massa, M. and A. Simonov, “Hedging, Familiarity and Portfolio Choice”, Review of

Financial Studies, 2006, 19 (2), 633.

Mitton, T. and K. Vorkink, “Equilibrium Underdiversification and the Preference for

Skewness”, Review of Financial Studies, 2007, 20 (4), 1255.

Sinai, T. and N. Souleles, “Owner-Occupied Housing as a Hedge Against Rent Risk”,

Quarterly Journal of Economics, 2005, 120 (2), 763–789.

Skiera, B., “Status Quo im Retail Banking in Deutschland”, Research Report. University

of Frankfurt, 2007.


Appendix

Table 7:Robustness check for five aggregated investors

This is a robustness check to table 2 and differs in the asset industry composition:Phone/Communication, IT Hardware and IT Services are excluded from IT and Bio-Energy isexcluded from Technology.



Total 33.13% 31.47% 24.89% 33.80% 35.54% 4,395

Table 8:Correlations of socio-demographic variables

This table displays the pairwise correlation coefficients of exogenous variables used in theregression analysis on page 14. All correlation coefficients are below 20%, all VIFs are below1.08 so the variables are free of multi-collinearity.

Gender Age Risk Deposit Equity Intern. DrProf Advisedtoler. value share diversif.

Gender 1Age 0.053 1Risk tolerance 0.071 -0.016 1Deposit value 0.000 0.184 0.075 1Equity share 0.042 -0.005 0.007 -0.099 1Int. diversification 0.001 -0.035 0.076 0.011 -0.080 1DrPorf -0.082 0.120 0.002 0.116 -0.023 0.040 1Advised -0.041 0.071 -0.047 0.121 -0.181 0.092 0.054 1


Table 9:Professions of the investors in the dataset

This table shows the professions of the respective number of all investors in the dataset. Thebold professions were used to construct five industries, the other professions were deleted fromthe dataset. From investors working in one of the five industry, portfolios with less than EUR100 in February 2007 were excluded so that the final sample consists of 4,395 investors: 329in Financial Services, 355 in IT, 1,206 in Health Care, 2,461 in Technology and 44 working inAviation.

Profession Investors Profession Investors

White-collar worker 9,845 Field staff 69Retiree 2,198 Bookkeeper 69Engineer 2,120 Unemployed 68Self-employed 2,064 Educator 62Student 1,571 Salesman 56Official 1,557 Designer 53Doctor 1,200 Assistant 52Manager 850 Military / civilian service 52Merchant 821 Farmer 50Office administrator 499 Pilot 46Housewife 472 Lecturer 44Public servant 465 Artist 44Teacher 447 Restaurateur 40Computer scientist 377 Motorcar mechanic 39Blue-collar worker 368 Fitter 33Banker 366 Machine operator 30Business economist 349 Cabinetmaker 30Advisor 348 Photographer 27Solicitor 279 Orderly 25Technician 277 Locksmith 24Business consultant 187 Musician 23Electrician 175 Lab assistant 21Tax advisor 172 Printer 18Architect 158 Cook 18Independent 136 Gardener 16Industrial clerk 125 Hairdresser 15Craftsman (master) 119 Warehouseman 14Jurist 108 Sharper 13Craftsman 101 Painter 13Physicist 99 Sportsman 10Pharmacist 82 Bricklayer 9Nurse 77 Customer advisor 8Apprentice 72 Masseur 8

Croupier 1

Total 29,184


Table 10:Demographics

Risk tolerance is reported by the investors when opening an account on a scale from 1 (low) to 6(high). The digital variable Dr./Prof. comprises only persons holding a Dr. or Prof. title. Shareof int. equity in the investor’s equity portfolio is a measure for diversification following Bluethgenet al. (2007). Deposit and cash values are means over the last six months. Average return p.a.and Sharpe Ratio are computed over the last 20 months. The Herfindahl-Hirschmann-Index(HHI) is a measure for diversification where lower numbers indicate higher diversification.

All 5 industries Financial Services IT(n = 4,395) (n = 329) (n = 355)

mean median mean median mean median

Gender (male) 89.83% 100% 75.08% 100% 88.45% 100%Age 45.98 45 41.12 40 41.28 40Risk tolerance 4.42 5 5.25 6 4.51 5Dr./Prof. (dummy) 21.89% 0% 0.91% 0% 5.35% 0%Advised (dummy) 13.74% 0% 10.03% 0% 11.27% 0%Heavy trader (dummy) 13.11% 0% 19.76% 0% 12.11% 0%Eurex trader (dummy) 3.69% 0% 17.02% 0% 3.10% 0%Deposit value (EUR) 105,837 68,613 102,283 61,351 88,132 65,089Cash (EUR) 37,116 18,444 34,631 16,314 31,571 18,013Equity share 87.59% 100% 84.79% 99.16% 87.39% 100%Average return p.a. 12.31% 12.52% 9.77% 11.01% 13.19% 12.74%Sharpe Ratio 0.88 0.99 0.77 0.91 0.90 0.98No of Securities 12.42 10 11.52 9 11.10 9HHI 9.25% 1.89% 10.42% 2.24% 10.52% 2.18%Fund ratio in equity 39.39% 27.79% 36.43% 21.21% 42.48% 32.34%Share of int. equity 49.88% 51.28% 47.11% 46.07% 49.57% 51.99%Share of index funds 6.86% 100% 9.74% 0% 7.32% 0%

Pharma Technology Aviation(n = 1,206) (n = 2,279) (n = 44)

Gender (male) 80.60% 100% 96.38% 100% 97.73% 100%Age 47.70 46 46.53 45 42.32 40Risk tolerance 4.41 4 4.30 4 4.52 5Dr./Prof. (dummy) 68.49% 100% 4.59% 0% 2.27% 0%Advised (dummy) 16.92% 0% 13.08% 0% 11.36% 0%Heavy trader (dummy) 12.19% 0% 12.76% 0% 15.91% 0%Eurex trader (dummy) 3.23% 0% 2.23% 0% 2.27% 0%Deposit value (EUR) 126,642 75,935 99,102 66,735 81,734 62,821Cash (EUR) 45,077 18,552 34,476 18,647 29,890 17,024Equity share 87.33% 100% 88.07% 100% 90.55% 100%Average return p.a. 12.28% 12.27% 12.47% 12.99% 16.25% 13.15%Sharpe Ratio 0.92 1.04 0.87 0.98 0.98 0.81No of Secs 13.29 10 12.31 10 12.64 10HHI 8.44% 1.67% 9.41% 1.89% 3.54% 1.90%Fund ratio in equity 46.78% 43.75% 35.73% 21.11% 39.35% 38.01%Share of int. equity 53.16% 55.43% 48.62% 49.05% 53.84% 59.37%Share of index funds 6.79% 0% 6.41% 0% 8.90% 0%

Documents

The Influence of Investors' Jobs on Portfolios: Is there a