Financial Literacy in the Age of Green Investment

Anders Anderson*

David T. Robinson†

First version: December 22, 2018

This version: November 16, 2021

Abstract

We survey a large sample of Swedish households and connect the responses to

administrative data to relate pro-environmental attitudes and values to actual in-

vestment decisions. Pro-environment households are not more likely to hold pro-

environment portfolios. This results from financial disengagement: they are less likely

to own stocks, check pension balances, or make green active retirement planning

choices. Green financial engagement is stronger in settings where financial literacy

is higher or where informational hurdles are lower. Informational barriers appear to

prevent financial market prices and returns from fully reflecting household environ-

mental preferences.

* Swedish House of Finance at the Stockholm School of Economics, Drottninggatan 98, SE-111 60 Stock-
holm, Sweden, Phone: +46-8-736 9110, E-mail: [Link]@[Link]. Corresponding author.
†
Fuqua School of Business, Duke University, Swedish House of Finance, and NBER, E-mail:
davidr@[Link]. We are especially grateful to Alex Edmans (the editor), to the Associate Editor, and
to anonymous referees for helpful comments. We are also grateful to Paul Smeets, Yang Yang, and to semi-
nar participants at the European Finance Association 2021 meetings, the Behavioural Economics workshop
at the Stockholm School of Economics, George Washington University, Linköping University, London Busi-
ness School, Northwestern and Peking University for comments. Yao Fu and Karl Sigfrid provided excel-
lent research assistance and we thank Vinnova, Mistra Financial Systems and the Jan Wallander and Tom
Hedelius Foundation for generous funding.

Electronic copy available at: [Link]

1 Introduction

In the fifty years since Milton Friedman’s famous adage, “The business of business is
profits” (Friedman, 1970), there has been an enormous growth in corporate social respon-
sibility and socially responsible investing. Not only do firms tout corporate social re-
sponsibility when articulating their corporate strategy to outside stakeholders, financial
market participants place growing importance on ESG considerations. Ultimately, these
trends raise questions about the nature of underlying household demand for socially re-
sponsible behavior from firms and other financial market participants.
To better understand the impact of ESG preferences on capital market outcomes and
firm behavior, a number of recent theoretical papers have taken up the connection be-
tween pro-social shareholder preferences and corporate behavior or asset prices. Peder-
sen, Fitzgibbons, and Pomorski (2020) derive a version of the CAPM in a setting in which
some investors are environmentally unaware, some are aware of the potential cash flow
implications of environmental issues but do not care about the environment, while some
derive utility from holding stocks with higher ESG scores. In their model these three types
of investors interact to determine asset prices. Pastor, Stambaugh, and Taylor (2020) de-
velop a framework in which some investors harbor social preferences and derive positive
utility from holding green stocks, thus affecting returns through their willingness to pay
higher prices for green stocks. These models share a common assumption, which is that
investors express their pro-social preferences through their portfolio choice decisions.
In this paper we empirically test this basic premise. Focusing specifically on environ-
mental considerations (the E in ESG), we ask a simple question: do green households
make green financial decisions? Exploring this question requires both an apparatus for
measuring ESG-related preferences as well as data on actual financial decisions. We de-
velop both.
Inspired by the Big-5 financial literacy test pioneered by Lusardi and Mitchell (2007),
we develop a five-question test of environmental literacy that measures knowledge of
basic facts related to climate change. We include this test in a larger survey that contains
standard measures of financial literacy as well as beliefs about the financial tradeoffs as-
sociated with green investment and environmental values. These questions are designed
to identify empirically the types of agents inhabiting the models of Pastor, Stambaugh,

Electronic copy available at: [Link]

and Taylor (2020) and Pedersen, Fitzgibbons, and Pomorski (2020); in particular, our
survey questions allow us to proxy for ESG-aware investors (those who think environ-
mental considerations will have cash-flow implications for certain types of firms) and
distinguish them from ESG-motivated investors (those who derive utility from holding
ESG-tilted portfolios). Then we administer the survey to a large cross-section of Swedish
households, where we can link the survey responses to administrative data that include
detailed socio-demographics, stock ownership, and retirement savings decisions in a
mandatory-participation, national pension plan.
Our central finding is that households with stronger pro-environmental values do not
hold greener portfolios. A key reason for this is that households with pro-environmental
views are disengaged from financial decisions. Individuals who place a high priority
on environmental considerations are around 10% less likely than others to hold stock
directly, controlling for demographics. The magnitude of this effect is comparable to that
associated with not attending post-secondary education. Not only are they less likely to
be direct stock owners, but even in a mandatory-participation retirement plan they show
limited signs of engagement: they are more than 20% more likely to never check their
pension balances, and they are more likely than others to leave their pension savings
in the default fund choice rather than make an active choice. At the same time, they
score high on measures of everyday pro-environmental behaviors—they recycle more
than their neighbors, and they are willing to pay more for green products. For most
individuals, however, this environmental engagement simply does not to cross over into
the realm of financial engagement. They report that environmental issues are interesting
but financial issues are boring.
These findings are striking in part because average levels of both environmental and
financial engagement in Sweden are high by international standards. The fact that envi-
ronmentally engaged households are financially disengaged presents a challenge to the
idea of “people’s capitalism”—the idea that ordinary citizens buy shares in firms and
hold these firms to account, implicit in many models of corporate social responsibility.
Understanding the reasons behind this disengagement are important for policymakers.
Because pro-environmental views are so often associated with the political left, which is
often critical of financial institutions, corporations, and capitalism in general (Kaustia and

Electronic copy available at: [Link]

Torstila (2012)), this could represent a form of virtue-signaling. Or it could simply reflect
the complexity of making financial decisions.
While there could easily be a virtue-signaling component to our findings, we find that
the complexity of many financial decisions presents a hurdle for many environmentally
engaged individuals. For example, the Swedish Pension Authority uses a classification
system that labels certain funds as ESG-compliant; this indicator was available in printed
and on-line reference material. At the same time, some funds actually contain words in
their name that signal their ESG status.1 Using this variation in the difficulty of determin-
ing whether a fund is environmentally friendly, we find that only more financially literate
individuals with pro-environmental preferences hold system-labeled ESG funds. Among
lower literacy individuals, environmentally engaged respondents are much more likely to
allocate their portfolios towards funds with green-sounding names. Even if this reflects a
lack of trust in the government’s labeling system, this finding suggests that informational
hurdles create investment challenges that prevent preferences from being expressed in
portfolio choices. An implication of this finding is that greater transparency and ESG-
related information could allow households to better express their pro-environmental
values through their investment decisions.
These results speak to several distinct literatures in household finance, behavioral fi-
nance and asset pricing. First, our results augment a large body of work in financial
literacy demonstrating its connection to financial engagement—we expand this connec-
tion to include environmental matters.2 Bauer and Smeets (2015), Riedl and Smeets (2017)
and Brière and Ramelli (2020) offer evidence for increased sustainable investments among
those with green-oriented values, but our results illustrate the importance of financial lit-
eracy and engagement for the expression of these beliefs. In that regard, our findings
also relate to those in Giglio et al (2020), who exclusively survey high net worth indi-
viduals, and find that beliefs about discount rates and expected returns are reflected in
the investment decisions people make. In contrast, our sample is based on a nationally
representative cross-section of households, which is why we find limited evidence that
environmental beliefs carry over to actual investment decisions.
1
Specifically, if the fund name contains the Swedish-language or English-language words for SRI, ESG,
water, sustainable, environmental, ethical, earth, clean or green.
2
See, for example, Ameriks, Caplin, and Leahy (2003), Beverly, Higert and Hogarth (2003), Anderson,
Baker, and Robinson (2017), Lusardi and Mitchell (2014) and Van Rooij, Lusardi, and Alessie (2011).

Electronic copy available at: [Link]

 The remainder of the paper proceeds as follows. First, in Section 2, we discuss our data
in greater detail and develop our main measures of financial and environmental literacy,
pro environmental beliefs and values as well as stockholdings and pension decisions.
Then we move directly to our main finding by examining the link between environmen-
tal values and pension holdings. This appears in Section 3. In Section 4 we examine direct
stock holdings. To dig deeper into the mechanisms behind the lack of green financial en-
gagement, in Section 5 we relate our measures to a number of non-financial decisions and
attitudes with clear implications for green behavior. Section 6 concludes with a discussion
of the implications of these findings for policy and future research.

2 Empirical Design

The centerpiece of our empirical strategy is a survey that we developed containing three
basic types of questions: questions about household values towards environmental is-
sues, questions about pro-environmental and investment behaviour, and questions about
financial and environmental literacy. In cooperation with Statistics Sweden, the govern-
ment statistical agency, we invited a random sample of 20,000 Swedish households to
take this survey online. The first survey was sent out on February 7, 2018, and after two
reminders sent out on February 22 and March 5, the survey was closed on April 5, 2018. A
total of 4,257 respondents completed the survey, for a response rate of just over 21%. After
deleting incomplete survey responses and matching the data to the characteristics leaves
us with at total of 3,993 observations.3 Table I provides a demographic breakdown of
the respondents along with mean values of the key variables of interest by demographic
category.

Table I here

The demographics include registry data obtained from Statistics Sweden for gender,
age, income, education and location. Location makes it possible for us to control for pop-
ulation density in each of the 290 municipalities in Sweden where each respondent lives.
Our urban/rural indicator variable assigns urban status to locations with population over
3
Details of the sampling procedure are presented in Appendix A.

Electronic copy available at: [Link]

100,000 and with a density over 100 people per square kilometer, which corresponds to
the twelve largest cities in Sweden. We match the vote count from the 2014 election for
the Swedish Green Party (Miljöpartiet), which has been part of the left wing coalition in
Sweden since 2014, at the voting-precinct level to obtain a measure of the degree to which
one’s neighbors are likely to hold pro-environmental attitudes. This allows us to gauge
the validity of some of our measures of pro-environmental attitudes and behaviours, and
also allows us to account for peer effects that may shape attitudes toward the environ-
ment.
Through Statistics Sweden we can also compare our survey data to the overall Swedish
working-age population.4 We have an over-representation of women and of older, wealth-
ier, higher-educated respondents in our sample: 57% of our respondents are 45 or older,
while only 41% of the Swedish working age population is in this age range. We received
15% of the responses from the lowest income bracket, while 25% of the relevant popula-
tion belong to this group. Slightly more than half of our survey respondents went to a
university programme, while just under 40% of the Swedish population has completed
high school.

2.1 Environmental values, beliefs and behaviours

One of our goals is to empirically distinguish respondents who harbor pro-environmental

views, and therefore may hold ESG-tilted portfolios for utility-driven reasons, from re-
spondents who may hold ESG-tilted portfolios for purely pecuniary reasons. This dis-
tinction corresponds to that made by Pedersen et al (2020) between ESG-aware and ESG-
motivated investors. Towards this end, a key variable in our analysis is our measure of
environmental value orientation. Schwartz (1982) defines a value as: “a desirable trans-
situational goal varying in importance, which serves as a guiding principle in the life of a
person or other social entity.” Values therefore reflect a belief concerning the desirability
of a certain end-state; they serve as a guiding principle for selecting or evaluating be-
haviour, people and events; they can be ordered in a system of priorities. To capture the
degree to which an individual’s personal values are oriented towards the environment,
4
Statistics Sweden provides sampling weights are based on gender, age, income, education and im-
migration background. All our results are similar with or without using weights, therefore in the tables
presented here we do not include them.

Electronic copy available at: [Link]

we ask the following question intended to capture their willingness to accept financial
tradeoffs for better environmental conditions:

• Clean planet. “A clean planet is more important to me than financial welfare.” (25%)

Possible answers are presented on a five point scale from “Strongly agree” to “Strongly
disagree”, “Don’t know” and “Prefer not to say”. Twenty-five percent of respondents
strongly agree to this statement, 39% agree and only about 8% disagree to some extent.5
We code a dummy equal to one for those who strongly agree to this statement and present
the fractions across demographic groups in Table I. For brevity we refer to this variable
as Clean planet in the tables of results. This response is more prevalent among women,
those with lower income, the highly educated, and city dwellers. The results on age are
u-shaped, with Clean planet being more common for those at the peak of their working
age, rather among the youngest in our sample.
Although the preceding question is intended to home in on individuals who have
preferences for ESG-portfolios based on non-pecuniary motives, these individuals may
also believe that investing in ESG-tilted portfolios offers a good risk/return tradeoff. To
better identify ESG-aware attitudes as a distinct from ESG-motivated attitudes, we also
ask respondents to what extent they agree that sustainable investments outperform, and
whether they are prepare to pay higher fees for financial products with that focus exclu-
sively on green investments:

• Higher return. “Environmentally sustainable investments generate higher returns in

the long run.” (14%)

• Higher cost. “I’m willing to pay higher fees for a fund that only makes sustainable
investments.” (9%)

About 14% strongly agree to the statement that green investments outperform, 30% agree
and less than 13% disagree to some extent. The willingness to pay higher fees for green
investments has less support. Only 9% strongly agree, 32% agree, while 17% disagree to
some extent.
5
Related measures of personal values include Hong and Kostovetsky (2012) who relate political affilia-
tion to SRI investments and Kahan et al (2012) who relate individualism vs. collectivism to beliefs about
climate change. Similarly, Bolton et al (2020) use proxy votes to infer preferences for sustainability vs. pure
financial considerations among institutional investors.

Electronic copy available at: [Link]

 These measures are correlated, as we confirm in the appendix. While they are in-
tended to capture specific motivations behind a pro-ESG stance, they can be taken to-
gether as an expression of a general preference for ESG-tilted investments, as in Pastor
et al (2019). To capture this broader preference succinctly, we average the raw responses
from the three variables (assigning -2 to strongly disagree and +2 to strongly agree so that
a middle score has a value of zero) and present that beside the individual measures in
Table I. We refer to this variable as Environmental preferences in the tables of results. Urban
dwellers, especially those who live in areas with high Green Party voter support, score
high on this measure. Those who studied environment-related fields, or biology, score
highest on this measure.

Table II here

Finally, we add three statements to gauge how our measures of environmental value
orientation relates to household everyday environmental behaviours:

• Green products. “I am willing to pay more for environmentally friendly products.”

(28%)

• Recycles more. “I recycle a great deal more than my neighbors.” (15%)

• Never checks pension. “I never review my retirement savings.” (26%)

These variables are summarized in Table II and are grouped together because they are
based on self-reflections of one’s behaviour and time use. The idea behind Green products
and Recycles more is to capture pro-environmental engagement in ordinary, day-to-day
consumer transactions as both a test of the validity of our measures of investor sentiment,
as well as a point of contrast against measures of financial engagement. The responses to
the question about green products reflect a strong willingness to pay for environmentally
friendly products in Sweden: 28% of survey respondents strongly agree to this statement,
41% agree, while only 10% disagree to some extent. Our recycling question is “Which
sentence best describes how much you recycle?” Potential answers ranges from “A great
deal less than my neighbors” to “A great deal more than my neighbors”. Only 7% of
respondents believe that they recycle less, but 43% believe that they recycle more than

Electronic copy available at: [Link]

their neighbors which is an illustration of overplacement and the well-known “better
than average” effect (see Svenson (1981)).
The question “How often do you review your retirement savings” is intended to
gauge the degree of a respondent’s engagement in financial matters that are important for
household financial planning decisions. While 16% state that they do so several times per
year, and 51% at least once a year, 26% report that they never do so. We use the response
“I never review my retirement savings” as a dummy variable indicating disengagement,
and refer to this as Never checks pension in the tables of results.
Our survey also includes questions measuring financial and environmental literacy.
For financial literacy, we use the “Big-5” test first developed by Lusardi and Mitchell
(2007). To measure environmental literacy in a comparable manner, we develop and ad-
minister a five-question test aimed at capturing knowledge of the environmental aspects
of various household consumer choices. These measures are discussed in detail in the
Appendix.

2.2 Stock Ownership and Pension Savings Data

To complete our dataset we connect the behavioural measures obtained through the sur-
vey described above to data on direct stock ownership and pension fund holdings. The
average level of stock market participation in Sweden is high by international standards.
For example, direct stock ownership as a fraction of total savings is considerably larger
in Sweden than the US. Sweden had the second largest share of direct stock ownership in
Europe in 2017.6 Sweden ranks third among European countries in terms of share of sav-
ings allocated to mutual funds. The Swedish premium pension plan, which is excluded
from the European statics and which we explain in more detail below, in effect also make
the whole working population mutual fund investors. Swedes are also important owners
of financial assets on an absolute scale, ranking ninth in Europe after Spain and Ireland,
but before Denmark, Austria and Finland.
6
Data from the European Fund and Asset Management Association, [Link].

Electronic copy available at: [Link]

2.2.1 Direct Stock Ownership

The individual stock data is obtained from Euroclear, Sweden’s central security reposi-
tory. We obtain stock ownership of listed individual Swedish equities for 2016, which is
matched anonymously to our survey data. This file contains the identity of individual
stocks, number of stocks held and market value at the end of each year. This is our mea-
sure of stock ownership, where we find that 946 respondents held equity in 471 firms.
Stock ownership in our sample is slightly higher than the average 23% for the whole
population.

Table III here

Panel A of table III presents the holdings of individual stocks. The stock ownership
sample covers 476 firms, and as can be seen in the size break-up, many of them are very
small. Even if the average firm has a market capitalization of SEK 14 billion (about USD
1.6 billion), 75% of the firms have a market cap well below one billion dollars, which
makes it very difficult to match the majority of firms on ESG-measures that normally are
available only for big firms through standard data vendors. Instead, we turn to Statistics
Sweden, who maintains a data base over aggregate Scope 1 carbon dioxide emissions
across industries. We match firms to this data set on two-digit standardized industry
codes (SNI) to assign each firm to 37 industries. This way, we obtain an overall measure
for the carbon intensity of the industry for each company’s stock. We also identify 16
energy stocks through the same industry classification of which eleven operates within
the oil and gas sector..
Stock ownership in our sample is slightly higher than the average 23% for the whole
population. The stock ownership sample is skewed as can be seen from the size break-up,
where the average portfolio holding is SEK 597,000, but the median just above SEK 33,000.
The average number of stocks in the portfolio is 4.4 and the median is 2, which indicates
that many portfolios are small. The median value of the holdings is SEK 6,420 (less than
one thousand USD). Some portfolios are large—the maximum number of stocks in the
portfolio in our data is 42.

Electronic copy available at: [Link]

2.2.2 Premium Pension Plan

Our second data source comes from the Swedish pension system. The Premium Pen-
sion Agency (PPA) shares data on investments and trades for the mandatory Swedish
premium pension plan launched in 2000, which is a national, tax-shielded plan similar
to corporate-sponsored 401(k) plans in the US. A total of 2.5% of worker’s wages are di-
rected into a designated personal account, and investors can choose up to five funds from
the participating pool of around 900 funds (892 at the end of 2017). Investors who do not
actively make a choice gets their contribution allocated to a well-diversified, target-date,
default fund. The objective of maintaining a high level of diversification means that the
fund has investments in all main equity markets and industries. The Swedish premium
pension system is described in detail in Palme, Sunden, and Söderlind (2007) and has
been analyzed before in for instance Cronqvist and Thaler (2004) and Dahlquist and Mar-
tinez (2015). A particular advantage of the system is its mandatory component with a
fixed contribution proportional to income, making it possible to abstract from potentially
important selection bias that would arise from generalizing decisions from people within
a given voluntary pension plan with variable contribution rates.
Panel B of Table III shows that there are 3,667 people in the plan among the 3,993
collected survey responses at the end of 2017. Around two-thirds of the savers had made
an active choice at some point participating in the pension plan, one-third never made a
choice and so hold only the default fund.
In 2004, the PPA introduced a designation to indicate whether a fund followed envi-
ronmental and social guidelines. This designation allows fund companies to label them-
selves as funds that invest with restrictions determined by ethical or environmental con-
siderations (so-called negative selection funds as in Hong and Kacperczyk (2009)). This des-
ignation takes the form of a green leaf that appears in the fund description, and appears
in all information and marketing about the funds, but there is no standard or minimum
requirements given by the PPA to which funds must adhere in order to earn this desig-
nation. From 2018 and onwards, after our sample period, all mutual fund companies on
the premium pension platform are required to follow the UN Principles for Responsible
Investment. At the end of 2017 about one-third of the funds (331 out of 892) carried this
label. At the same time, almost 74% of individuals who have made an active choice hold

10

Electronic copy available at: [Link]

some ESG-labeled funds in their portfolios, while 58% hold a majority of their portfolio
in ESG-labeled funds, and 34% of respondents hold an all-ESG portfolio in their pension
accounts.
When the flag was introduced, the PPA announced that the default fund would also
apply ethical guidelines and exclusion strategies for non-compliant firms, ostensibly mak-
ing it a green fund.7 Yet at the same time, the fund’s mandate requires it to be globally
diversified, and indeed, it has been widely criticized in the Swedish press for holding siz-
able positions in heavy-polluting companies, particularly in the fossil fuel industry. Be-
cause of this, it is unclear whether holding the default fund should be considered “making
a green choice.” As such, we provide distinct analyses based on whether or not holding
investments in the default fund is taken as evidence of making a green choice (along with
evidence later in the paper that speaks against the green-choice interpretation).
We also manually flag funds if the name of the fund suggests that the it has a so-
cially responsible investment profile. There are 46 funds containing keywords suggesting
that they follow guidelines related to sustainability. Labels are based on nine keywords:
“SRI”, “ESG”, “Sustainable”, “Environmental”, “Green”,“Ethical”, “Clean”, “Earth” and
“Water”, appearing either in Swedish or English. Of the 2,474 respondents who made an
active choice, we find that 1,827 had at least some exposure to ESG-labeled funds, while
167 respondents held ESG-named funds in their portfolio.
There is no reason, based on objective classification or ESG-reporting criteria, to be-
lieve that a green-named fund is greener than a green-labeled fund. (Indeed, Cooper, Gulen,
and Rao (2005) and Cooper, Dimitrov, and Rao (2001) illustrate how name changes can be
used to manipulate investor perceptions.) Yet because some funds have ESG-related in-
formation in the name itself, while other funds can only be determined as ESG-compliant
by consulting marketing materials, the two labeling standards for ESG-funds allow us
to test an information-related hypothesis concerning ESG-tilted fund ownership. Specif-
ically, if information about the fund’s actual investment behaviour provides a barrier to
financial decision-making, we would expect the connections between pro-environmental
preferences and ESG-tilted portfolios to be stronger for the ESG-named funds, where
their status is more transparently available to the investor. This is a weak test because
7
There were 65 blacklisted companies in 2018, meaning that they were classified as violating interna-
tional norms regarding the environment, human rights, labor rights or corruption.

11

Electronic copy available at: [Link]

the flag that indicates ESG-compliance is clearly available in publicly available market-
ing materials. Another mechanism that could drive a difference in uptake between the
types of funds is trust in the labeling system: perhaps investors find the ESG-compliant
label ambiguous, while a fund using ESG-related terms in its title itself sends a more
credible signal of its ESG-related intentions. Note however that under this mechanism,
the channel connecting investor behaviour to investment choices is still sophistication in
information processing.

3 Do Green Households Hold Green Retirement Portfolios?

For the central part of our analysis, we explore the connection between our measures of
environmental attitudes and portfolio holdings in the Swedish pension system. We start
here because the mandatory nature of the retirement system means that we can explore
issues related to financial engagement and the expression of environmental beliefs with
less concern that unobserved variation in idiosyncratic liquidity considerations or other
household situational factors confound our analysis. In other words, we start here be-
cause every working age individual contributes on an equal basis proportional to their
income, and every contributor has access to the same investment information, albeit con-
cerning a very large menu of approximately 900 mutual funds from which to choose. The
only question is whether they use this information to hold portfolios that express their
environmental beliefs.
To frame the question in the simplest possible terms, we begin by taking the full sam-
ple of all 3,667 survey respondent retirement accounts and ask whether green households
hold green portfolios. We consider two possibilities based on the categorization of the
default fund. In the first, we assume that the default fund is not a green choice because
of its index benchmarking across industries which may appear controversial for green
investors. We code a dummy to be equal to one for portfolios containing 100% ESG-
funds based on this assumption and label it Active ESG. Then we repeat the analysis but
code the default fund as ESG-compliant. This assumes that investors in the default fund
made a conscious decision not to move their money out of the default fund because they
deemed it to be sufficiently ESG-compliant. We will evaluate the reasonableness of that

12

Electronic copy available at: [Link]

assumption below.
Table IV displays the results. In Columns (1) through (3) the dependent variable is
coded so that the default fund is not a green choice. The regressions control for income,
age, gender, education, population density and green party voting outcomes. The Clean
planet variable loads negatively on making an active ESG choice, but is only significant at
the 10 % level. This holds when we separately categorize ESG-aware and ESG-motivated
invested by including measures of willingness to pay a higher cost for green funds or
believing green investments generate higher returns. Of course, as discussed in Section 2,
these variables are highly correlated, and therefore multicollinearity may be driving the
lack of a result. To account for this, we replace the three distinct measures with their sum
in order to proxy for a general ESG-preference, as in Pastor et al (2020). This is presented
in Column (3). The pooled measure of green preferences is also insignificant. The results
offer no evidence that pro-environmental investors choose greener portfolios.

Table IV here

What if, instead, we assume that leaving retirement savings in the default fund is
making an active choice to be green? Columns (4) through (8) of Table IV explore this
interpretation of the default option. Here the dependent variable is coded one if ESG-
labeled or default fund, zero otherwise. (Thus, an individual must make an active brown
choice in order to not be considered green.) Again, the clean planet variable does not
predict that respondents hold green investments, even if the default is considered to be
an active choice to hold an environmentally sensitive portfolio. Higher return, one of the
proxies for “ESG-aware,” is significant at the ten percent level. When we combine these
three into the more general measure of green preferences in Column (6), the resulting
variable is significant at the 1% level.
Interpreting this as making a green choice requires the assumption that pension savers
knowingly choose to leave their investment in the default fund because of their under-
standing of its pro-ESG properties. To asses the validity of this assumption, we split the
sample on financial literacy. Columns (7) and (8) report sub-samples based on low and
high financial literacy respondents, respectively, where high financial literacy is indicated
by answering the Big-3 financial literacy questions correctly. All the significance in Col-

13

Electronic copy available at: [Link]

umn (6) is coming from the high financial literacy sub-sample. This suggests that this
interpretation is only relevant among financially sophisticated individuals.
Thus, the results of Table IV provide weak evidence, at best, in favor of the idea that
green households hold green portfolios. There is some evidence that green households do
not actively seek out brown portfolios, but this relies on assuming that leaving savings in
the default fund—making no choice at all—is making the green choice. This stands some-
what in contrast to Riedl and Smeets (2017) and Brière and Ramelli (2020) who show that
evidence for increased sustainable investments among those with green-oriented values,
as well as results in Giglio et al (2020), who survey sample high net worth individuals
exclusively, and find that beliefs about discount rates and expected returns are reflected
in the investment decisions people make.
To understand why there is no relation between green beliefs and green investments,
we decompose the decision into (a) whether to leave their investment in the default fund,
and (b) whether to make green investments conditional on leaving the default fund. This
is presented in Table V. Columns (1) through (3) of Table V presents the results of a Probit
regression where the dependent variable is an indicator for making an active choice. Col-
umn (1) shows that active investors are those with higher financial literacy, but there is no
connection between environmental literacy and active pension choice. The Clean planet
variable is associated with about a 7% lower probability of making an active choice. In
Column (2), we introduce the environmental measures designed to identify ESG-aware
investors. Although they come in with a negative sign, indicating a lower probability of
active choice, they are not statistically significantly different than zero. Thus, Columns
(1) and (2) taken together show that ESG-motivated investors are less likely to make an
active retirement choice (of any kind, green or not) while ESG-aware individuals are not
different than ESG-naive investors. Column (3) presents the pooled, general measure of
green preferences, which shows that general environmental preferences are associated
with less active choice, significant at the 10% level.

Table V here

The first three columns do not distinguish between green or brown active choices,
only whether the respondent made an active or passive choice in the retirement system.
Beginning in Column (4) of Table V, we focus on the 2,474 investors who made an active

14

Electronic copy available at: [Link]

choice. The dependent variable is the weight that ESG-labeled funds have in the respon-
dent’s portfolio. There is no evidence that any measure of green orientation is related to
holding a green retirement portfolio.
There are a number of potential hypotheses behind the lack of financial engagement
by the environmentally engaged. Because many environmental activists are often criti-
cal of financial institutions, corporations, and capitalism in general (Kaustia and Torstila
(2012)), financial disengagement could be a way to show allegiance to environmental
causes, a form of virtue-signaling. Or it could simply reflect the fact that financial deci-
sions are often complex. To explore these hypotheses, in Column (7), we include only the
1,738 investors who get the Big-3 questions correct on the financial literacy test. If virtue-
signaling were solely responsible, we would not expect to find a difference here, but if
informational hurdles are important, we would expect a tighter link between values and
investments in this sample. Indeed, here we find evidence that green values are related
to green investment. This echoes the results presented in Column (8) of Table IV.
To explore the role of informational hurdles further, in Columns (8) through (10) we
redefine ESG-investment by focusing on ESG-named funds, rather than funds with an
ESG flag available in the Premium Pension marketing materials. The idea here is that
if ESG-signals are present in the fund’s name itself, individuals who lack the financial
sophistication to engage with the prospectus and marketing materials can nevertheless
select these funds based on name alone. The dependent variable now is the portfolio
weight in funds that contain some key word suggesting that it is environmentally or eth-
ically compliant. In Column (8) we see a highly statistically significant positive loading
on Clean Planet, indicating that ESG-motivated investors are more likely to hold these
funds. This result fades in Column (9), when we include the other measures of envi-
ronmental engagement, but this may be due to the high correlation between the three
measures. In Column (10), we account for this correlation by replacing the individual
ESG measures with the broader environmental preferences variable. Here again we see a
highly statistically significant positive relationship between holding an ESG-named.
In sum, this evidence indicates that green households do not, on average, actively
choose green retirement portfolios. One key challenge behind their lack of choice is the
information hurdles associated with making the choice; even if ESG-compliant funds are

15

Electronic copy available at: [Link]

labeled clearly, in a menu of almost 900 funds, the task of choosing a green portfolio in line
with one’s environmental preferences can be daunting. In line with this hypothesis, we
find that only the highly financially literate select ESG-labeled funds in keeping with their
environmental preferences when accessing the label involves consulting performance in-
formation provided by the Premium Pension system. When the informational hurdle is
lowered, we do see uptake of funds that advertise their sustainability focus through their
name among individuals people with a strong, pro-social value orientation.

4 Do Green Households Buy Green Stocks?

For additional evidence on the role of financial disengagement, we turn to direct stock
holdings. The very act of owning stock itself is a measure of financial engagement, be-
cause unlike the premium pension plan, which is a mandatory-participation plan, in or-
der to own stock in the first place a respondent has to take the active step of opening
an account at a brokerage and making a stock purchase. Presumably respondents who
belong to this group are much closer in disposition and sophistication to those surveyed
in Giglio et al. (2020). Do we see green preferences manifest in green portfolio holdings
in this group?
To explore this question, in the first three columns of Table VI we study whether green
households are likely to be direct stock owners. Here, the dependent variable is an in-
dicator equaling for who own shares, zero otherwise. The control variables show that
stock owners tend to be higher income, older, male, university-educated property own-
ers with an economics/business degree. They also tend to have higher financial literacy,
which is in line with the results in Van Rooij, Lusardi, and Alessie (2011). Column (1)
shows a negative loading on Clean planet, statistically significant at the 10% level, which
indicates that ESG-motivated people are less likely to own stocks. In Column (2), the
high correlation between the three variables measuring environmental attitudes renders
them statistically insignificant, but in Column (3), when we use the broader measure of
environmental preferences, we find a strong, negative relation between environmental
preferences and direct stock ownership. This is further evidence supporting the idea that
environmentally engaged households are financially disengaged.

16

Electronic copy available at: [Link]

 Table VI here

Nonetheless, some households with pro-environmental preferences do own stock di-

rectly. In the remainder of the table we focus on the 976 individual stockholders to ex-
plore whether environmental engaged households with high financial engagement ex-
press their green preferences through their stockholdings. In Columns (4) through (6) the
dependent variable is a dummy variable for whether the individual holds stocks in the
energy sector. This is a natural place to explore the tension between ESG-motivated and
ESG-aware investors. Because oil consumption is so heavily implicated in climate change,
environmentally motivated investors are likely to eschew energy stocks for virtue-signaling
motives, while ESG-aware investors may see profit opportunities as oil companies em-
brace alternative sources of energy production. Also, as we discussed in Section 2, these
stocks include eleven stocks in the oil and gas sector, but five belonging primarily to the
solar, wind and hydro energy sector. Less sophisticated investors may be unaware of the
potential for energy stocks to include companies that focus on renewable energy.
In Column (4) we find weak evidence that ESG-motivated stockholders are less likely
to hold energy stocks. There are no students of environmental science that hold energy
stocks, so the variable is dropped from the regression. The Clean planet variable has
a negative loading, but is statistically insignificant. When we introduce the proxies for
ESG-aware and include them alongside the ESG-motivated measure, we indeed see evi-
dence supporting the predictions of Pedersen et al (2020). In particular, we find a strong
positive loading on Higher returns, and a strong negative loading on Clean planet. The
interpretation here is that ESG-motivated individuals are loathe to hold energy stocks be-
cause they derive disutility from holding portfolios that can be construed as contributing
to pollution, whereas ESG-aware investors tilt their portfolios in favor of these stocks,
either because they expect positive cash flow shocks or because, along the lines of Hong
and Kostovetsky (2012) they anticipate earning excess returns for holding sin stocks. In
Column (6) the opposing signs cancel each other out and we find no effect on our general
measure of environmental preferences. While this could be viewed as a statistical artifact,
it is also in keeping with the interpretation that there is weak evidence supporting the
idea that green investors hold green portfolios because the “people’s capitalism” motive
for holding energy stocks cuts against the virtue-signaling motive for avoiding them.

17

Electronic copy available at: [Link]

 To further explore the role of informational impediments preventing the expression
of green preferences, in Columns (7) through (9), we replace the sector classification with
the measure of sector-level carbon dioxide emission described in Section 2. The compari-
son between energy sector holdings and holdings based on sector carbon emissions tests
the information hypothesis in a manner analogous to the comparison presented in the
previous section between ESG-labelled and ESG-named mutual funds. We present OLS
regressions in which the dependent variable is the portfolio carbon dioxide emissions.
Here, there is no connection between a stock holders’ green orientation and the carbon
footprint of their individual stock portfolio.
To summarize, direct stock ownership is presumably a channel through which people
with more sophistication and enthusiasm engage in financial markets. Environmentally
engaged respondents are significantly less likely to belong to this group, even controlling
for demographic factors like wealth and age, which affect stock market participation. Yet
by focusing on those individuals who do hold stock, we can shut down the extensive
margin of participation and examine portfolio holdings among a selected set of high-
participation individuals. When we do, we find that green preferences manifest in stock
holdings through both a virtue-signaling and a cash-flow channel. Consistent with the
evidence presented above showing that information hurdles present a challenge to green
investing, when we replace the sector classification with a more scientifically precise, but
more difficult to observe classification, we find no evidence that green investors tilt their
investment towards portfolios with lower carbon emissions.

5 More Evidence on Financial Disengagement

The preceding sections points to financial disengagement as a key reason why green in-
vestors do not hold green retirement portfolios. In this section we further explore the
mechanisms behind this disengagement. To explore this further, our final piece of anal-
ysis considers engagement across different domains. This helps guard against the pos-
sibility that our results stem from the survey questions having low power to detect en-
gagement because they measure preferences with error. It also helps illustrate differences
in attention across domains of household decision-making that are important for under-

18

Electronic copy available at: [Link]

standing our main findings.
Table VII presents results from three types of engagement. In Columns (1)-(3), the
dependent variable is a dummy equaling one if the respondent strongly agrees with the
statement “I am willing to pay more for environmentally friendly products.” Column (1)
shows a strong positive loading on Clean planet, our measure of ESG-motivated house-
holds. In Column (2) we include the ESG-motivated measure alongside the ESG-aware
measures, a willingness to pay higher fees and a belief that green investments outper-
form. All three carry highly statistically significant, positive loadings. The third column
replaces these three with the sum to obtain a broader measure of ESG preferences. Each
of these three specifications shows a clear link between our measures of environmental
attitudes and a measure of environmental engagement.

Table VII here

The next three columns focus on whether respondents think they recycle more than
their neighbors. This is coded as a dummy equaling one if respondents strongly agree
to that statement. Here we see a pattern in loadings consistent with the previous three
columns, where each of the environmental attitudes variables carries a large, highly sta-
tistically significant loading. The control variables include a dummy for living in an area
with a high Green Party vote count in the last election, which carries a negative sign,
indicating that respondents are less convinced that they recycle more than their neigh-
bors in areas where there neighbors harbor similarly pro-environmental views. This is
encouraging as a safety check that survey respondents are answering thoughtfully.
Columns (7) through (9) of Table VII turn to a measure of financial engagement, where
the dependent variable is an indicator variable equaling one for those who report that
they never check their pension balance. High financial literacy respondents are much less
likely to be those who never check their retirement balances.8 In contrast, respondents
with strong environmental values are more likely to respond that they never check their
balance. Interestingly, in Column (8) when we introduce the measures intended to proxy
for ESG-aware investors, we do not see the same pattern as in the environmental engage-
ment measures. A willingness to pay higher fees for green mutual funds and believing
8
The unreported control variables show that older and wealthier respondents are less likely to report
that they never check their retirement balances.

19

Electronic copy available at: [Link]

that green investments outperform is not associated with being less likely to check retire-
ment pension balances. This supports the ESG-awareness interpretation of the Higher
Cost and Higher Return variables, because ESG-aware investors derive no special util-
ity from being green, they simply see profit opportunity, therefore they are less likely to
eschew financial engagement on virtue-signaling grounds.
To explore this further, the remaining six columns focus on interest in environmen-
tal and financial matters. If virtue-signaling motives generate less financial engagement
among ESG-motivated individuals, and less common among ESG-aware investors, we
should see both types of investors interested in environmental matters, but only ESG-
motivated individuals interested in financial matters. In Columns (10) through (12) we
find a robust relationship between environmental interest and our measures of both ESG-
awareness and ESG-motivation. This holds whether we treat them separately or whether
we aggregate to the broader measure of ESG preferences. But in Columns (13) through
(15), we see that only a measure of ESG-awareness is significantly correlated with finan-
cial interest. In particular, in Column (14), when we include ESG-awareness measures
(Higher return and Higher cost) alongside ESG-motivation (Clean planet), we see that
there is a strong positive relationship between Higher returns and financial interest, but
a weak negative relationship in the other measures. In line with previous results, this
suggests that people need to show an interest in finance in order to form return expecta-
tions about green investments. The positive correlation in these measures is likely behind
the modest positive loading in Column (13), but when we combine these measures into
an aggregate measure of general environmental interest (column 15), we see no relation
between financial interest and green preferences.
Of course, respondents with strong environmental orientation may not actually recycle
more than their neighbors, they may not actually pay more for environmentally friendly
products: the literature on motivated reasoning (Carlson et al (2020) and FeldmanHall
et al (2012)) shows that people may misremember their past actions in order to main-
tain their moral self-image or act self-serving when faced with real choices. Whether
or not people actually engage in these behaviours is less important for our findings in
this section, because in contrast, their retirement planning responses point to disengage-
ment with financial markets, especially for ESG-motivated investors. Under a motivated

20

Electronic copy available at: [Link]

reasoning explanation, financial engagement is not part of the moral self-image that envi-
ronmentally engaged respondents wish to maintain (see also Brière and Ramelli (2020)).
The contrast between ESG-aware and ESG-motivated investors regarding the degree to
which they are interested in financial matters bears this out: it suggests that the virtue-
signaling motives behind a lack of financial interest are consistent with the idea that the
positive utility that attaches to holding ESG funds for ESG-motivated goes hand in hand
with a willful disengagement from financial matters.

6 Conclusion

This paper shows that households with green preferences by and large do not make green
financial decisions. There are two key factors that drive this result. The first is finan-
cial disengagement. Households that exhibit strong pro-environmental behaviors and
beliefs are financially disengaged and generally uninterested in financial matters. In a
mandatory-participation national retirement plan, they are less likely than others to check
their retirement balance, and they are less likely to make an active allocation decision, in-
stead relying on the default. Outside their retirement plan, they are less likely to own
individual stocks.
The second factor appears to be informational constraints associated with making
green investment decisions. Simply put, making green investment decisions involves
added layers of informational complexity that prevent less financially sophisticated indi-
viduals from expressing their green preferences in their portfolio holdings. For example,
among investors who do make an active choice in their retirement portfolios, environ-
mentally oriented investors are more likely to buy mutual funds with pro-environmental
names, but they are not more likely to buy funds that are labeled as ESG-compliant in
the more complex, statements and investment information that households receive every
year. Individual stockholders—i.e., individuals who have made an active stock market
participation decision—hold stocks in a manner consistent with environmental prefer-
ences when we classify stocks according to whether they are in the energy sector, but
there is no connection between their preferences and their stockholdings when we rank
according to a more complex measure, the carbon emissions ranking of the industry in

21

Electronic copy available at: [Link]

which the company operates.
Our findings have policy implications. Policy makers, regulators, and financial mar-
ket participants are currently in a debate over how much ESG-related information should
be made available to market participants. Some argue that more information is essen-
tial for sound decision-making, while others argue that additional information can easily
overwhelm decision-makers because the information is too complex. Indeed, in a public
statement on August 26, 2020, SEC Commissioner Allison Lee argued that “It’s time for
the SEC to lead a discussion—to bring all interested parties to the table and begin to work
through how to get investors the standardized, consistent, reliable, and comparable ESG
disclosures they need to protect their investments and allocate capital toward a sustain-
able economy.” Our findings suggest that is likely necessary, but not sufficient, for most
households’ environmental preferences to be expressed through their portfolio choice de-
cisions. Financial sophistication—the ability to digest and act upon that information—is
a necessary pre-condition for this information to steer capital towards more sustainable
uses.
The fact that greener households do not generally hold greener portfolios cuts in two
directions in terms of the debate around whether firms should pursue social and envin-
ronmental responsibility at the behest of their shareholders. On the one hand, one can
argue that firms have no special environmental obligation vis-a-vis their shareholders
because clientele effects in terms of environmental preferences are blunted by the lack
of financial engagement: if brown investors hold green stocks and green investors hold
brown stocks, why should companies go green on behalf of their shareholders? The op-
posing argument, however, is that all firms should pursue pro-ESG policies because ag-
gregate preferences towards green investment are far broader than would be suggested
by the relatively small group of investors who are both financially engaged and envi-
ronmentally engaged. One potential remedy here would be to concentrate the task of
gathering and processing ESG-related information into the hands of financial intermedi-
aries, such as fund managers, rather than leave this to individual shareholders. The EU
taxonomy to assist with incorporating ESG factors into financial decision-making is one
such effort on this front, but our findings suggest that it will be challenging to strike a bal-
ance between clarity and simplicity required to facilitate decision-making and an accurate

22

Electronic copy available at: [Link]

representation of the complexities of climate impact.
These results also have implications for ongoing theoretical research. Although when
we focus on high-engagement subsamples we find clear evidence consistent with the pre-
dictions of recent theoretical models, the fact that financial engagement is so low among
most environmentally engaged households suggests that the economic mechanisms at
work in these models may have limited empirical traction. Our results suggest that an
important topic for future work is incorporating limited participation, limited financial
sophistication, or else incorporating financial intermediaries who act at the behest of their
constituents, into models in which investors harbor environmental informed preferences.

23

Electronic copy available at: [Link]

References
Allcott, Hunt, and Dmitry Taubinsky, 2015, Evaluating behaviorally motivated policy:
Experimental evidence from the lightbulb market, American Economic Review 105, 2501–
2538.

Ameriks, John, Andrew Caplin, and John Leahy, 2003, Wealth accumulation and the
propensity to plan, Quarterly Journal of Economics 118, 107–114.

Anderson, Anders, Forest Baker, and David T. Robinson, 2017, Precautionary savings,
retirement planning and misperceptions of financial literacy, Journal of Financial Eco-
nomics 126, 383–398.

Anderson, Anders, and David T. Robinson, 2018, Who feels the nudge? Knowledge, self-
awareness and retirement savings decisions, Swedish House of Finance working paper
No. 17-15.

Bauer, Rob, and Paul Smeets, 2015, Social identification and investment decisions, Journal
of Economic Behavior and Organization 117, 121–134.

Beverly, Sondra, Marianne A. Hilgert, and Jeanne M. Hogarth, 2003, Household finan-
cial management: The connection between knowledge and behavior, Federal Reserve
Bulletin pp. 309–322.

Bolton, Patrick, Tao Li, Enrichetta Ravina, and Howard Rosenthal, 2020, Investor ideol-
ogy, Journal of Financial Economics 70, 320–352.

Brière, Marie, and Stefan Ramelli, 2020, Personal values, responsible investing and stock
allocation, Netspar working paper No. DP01/2020-003.

Carlson, Ryan W., Michel André Maréchal, Bastiaan Oud, Ernst Fehr, and Molly J. Crock-
ett, 2020, Motivated misremembering of selfish decisions, Nature commnications pp.
424–428.

Cooper, Michael, O. Dimitrov, and Raghu Rao, 2001, A [Link] by any other name,
Journal of Finance 56, 2371–2388.

Cooper, Michael, Husein Gulen, and Raghu Rao, 2005, Changing names with style: Mu-
tual fund name changes and their effects on fund flows, Journal of Finance 60, 2825–2858.

Cronqvist, Henrik, and Richard H. Thaler, 2004, Design choices in privatized social-
security systems: Learning from the Swedish experience, American Economic Review,
Papers and Proceedings 94, 424–428.

Dahlquist, Magnus, and José Martinez, 2015, Investor inattention: A hidden cost of choice
in pension plans?, European Financial Management 21, 1–19.

FeldmanHall, Oriel, Dean Mobbs, Davy Evans, Lucy Hiscox, Lauren Navrady, and Tim
Dalgleish, 2012, What we say and what we do: The relationship between real and
hypothetical moral choices, Cognition 123, 434–441.

24

Electronic copy available at: [Link]

Friedman, Milton, 1970, The social responsibility of business is to increase its profits, New
York Times Magazine, 32-33, 122-124.
Giglio, Stefano, Matteo Maggiori, Johannes Stroebel, and Stephen Utkus, 2021, Five facts
about beliefs and portfolios, American Economic Review 111, 1481–1522.
Gustavsson, Jenny, Christel Cederberg, and Ulf Sonesson, 2011, Global food losses and
food waste? Extent, causes and prevention, FAO, Rome.
Hastings, Justine S., Brigitte C. Madrian, and William L. Skimmyhorn, 2013, Financial
literacy, financial education, and economic outcomes, Annual Review of Economics 5,
347–373.
Hong, Harrison, and Marcin Kacperczyk, 2009, The price of sin: The effects of social
norms on markets, Journal of Financial Economics 93, 15–36.
Hong, Harrison, and Leonard Kostovetsky, 2012, Red and blue investing: Values and
finance, Journal of Financial Economics 103, 1–19.
Kaustia, Markku, and Sami Torstila, 2012, Stock market aversion? Political preferences
and stock market participation, Journal of Financial Economics 100, 98–112.
Labriet, Maryse, Santosh R. Joshi, Frederic Babonneau, Neil R. Edwards, Philip B. Holden,
Amit Kanudia, Richard Loulou, and Marc Vielle, 2015, Worldwide impacts of climate
change on energy for heating and cooling, Mitigation and Adaptation Strategies for Global
Change 20, 1111–1136.
Lusardi, Annamaria, and Olivia S. Mitchell, 2007, Baby boomer retirement security: The
roles of planning, financial literacy, and housing wealth, Journal of Monetary Economics
54, 205–224.
, 2011, Financial literacy around the world: An overview, Journal of Pension Eco-
nomics and Finance 4, 497–508.
, 2014, The economic importance of financial literacy: Theory and evidence, Journal
of Economic Literature 52, 5–44.
Palme, Marten, Annika Sunden, and Paul Söderlind, 2007, How do individual accounts
work in the swedish pension system?, Journal of the European Economic Association 5,
636–646.
Pastor, Lubos, Robert F. Stambaugh, and Lucian A. Taylor, 2020, Sustainable investing in
equilibrium, Forthcoming in Journal of Financial Economics.
Pedersen, Lasse Heje, Shaun Fitzgibbons, and Lukasz Pomorski, 2020, Responsible in-
vesting: The ESG-efficient frontier, Forthcoming in Journal of Financial Economics.
Riedl, Arno, and Paul Smeets, 2017, Why do investors hold socially responsible mutual
funds?, Journal of Finance 72, 2505–2550.
Schwartz, Shalom H., 1982, Universals in the content and structure of values: Theoretical
advances and empirical tests in 20 countries, in M. Zanna, ed.: Advances in experimental
social psychology . pp. 463–489 (Academic Press: Orlando, F.L.).

25

Electronic copy available at: [Link]

Svenson, Ola, 1981, Are we all less risky and more skillful than our fellow drivers?, Acta
Psychologia 1, 143–148.

Van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie, 2011, Financial literacy and
stock market participation, Journal of Financial Economics 191, 449–472.

26

Electronic copy available at: [Link]

Table I: Sample Characteristics
This table reports summary statistics, average test scores and sample proportions of the key survey questions. The first two
columns shows the sample proportions and the corresponding population average for Sweden. Columns labeled Green Beliefs show
the proportions of respondents strongly agreeing to the three statements: “A clean planet is more important to me than financial
welfare” (Clean planet); “In the long run, environmentally sustainable investments generate higher returns” (High return) and “It is
worth paying higher fees for a mutual fund that only make environmentally sustainable investments” (High cost). Green preferences
is an average of the three questions, where each question is scored from -2 (Strongly disagree) to 2 (Strongly agree). The columns
labeled Green Behaviour show the fraction strongly agreeing to the statements “I am willing to pay more for environmentally
friendly products” (Green products) and “I recycle a great deal more than my neighbors” (Recycles more). The two columns labeled
Literacy Scores present the average score from the financial and environmental tests, where the maximum score is 5. The final column
shows the proportion of respondents holding individual stocks in sample. The sample is compared across gender, age, income and
education where information about the underlying population is provided by Statistics Sweden. In addition, there are indicator
variables for whether the subject having studied Economics/Business or Biology/Geography/Environmental science at any level
since high school. Urban is a dummy variable for those living in any of the 12 main cities of Sweden and Green Party a dummy for
the highest quintile of green voting districts. There are 3,933 individuals in the sample.

Green Beliefs Green Behavior Literacy %

Proportions Clean High High Env. Green Recycles Scores Owns
Sample Pop. planet return cost pref. products more Fin. Env. Stock

Overall 1.00 1.00 0.25 0.14 0.09 0.47 0.28 0.15 3.14 2.27 0.24
Pop. Wtd. . . 0.25 0.16 0.09 0.46 0.27 0.15 2.93 2.23 0.19

Gender
Men 0.49 0.51 0.23 0.13 0.08 0.36 0.25 0.13 3.44 2.32 0.31
Women 0.51 0.49 0.28 0.15 0.10 0.58 0.31 0.16 2.85 2.22 0.18

Age
18-24 0.10 0.15 0.20 0.13 0.08 0.37 0.20 0.13 2.69 2.32 0.12
25-34 0.15 0.23 0.26 0.17 0.13 0.49 0.36 0.14 3.03 2.30 0.18
35-44 0.18 0.21 0.31 0.16 0.11 0.53 0.36 0.14 3.28 2.35 0.26
45-54 0.25 0.22 0.24 0.12 0.09 0.45 0.27 0.15 3.26 2.26 0.25
55-65 0.32 0.19 0.25 0.12 0.07 0.48 0.24 0.15 3.15 2.21 0.30

Income
0-111 0.15 0.25 0.27 0.17 0.10 0.47 0.27 0.15 2.72 2.31 0.11
111-287 0.32 0.25 0.26 0.15 0.09 0.51 0.26 0.17 2.83 2.25 0.17
287-399 0.29 0.25 0.24 0.13 0.09 0.50 0.28 0.13 3.18 2.21 0.23
399+ 0.24 0.25 0.24 0.10 0.08 0.38 0.33 0.13 3.79 2.35 0.44

Education
Some school 0.10 0.17 0.18 0.17 0.08 0.37 0.21 0.15 2.59 2.21 0.15
High school 0.38 0.44 0.21 0.12 0.05 0.40 0.20 0.15 2.84 2.17 0.22
University 0.52 0.39 0.30 0.14 0.12 0.54 0.36 0.14 3.48 2.37 0.29
Studied Env/Bio 0.02 . 0.34 0.18 0.17 0.62 0.52 0.28 3.62 2.65 0.25
Studied Econ/Bus 0.10 . 0.23 0.14 0.09 0.44 0.27 0.17 3.52 2.16 0.32

Location
Urban 0.34 . 0.29 0.15 0.12 0.54 0.35 0.12 3.32 2.34 0.26
Rural 0.66 . 0.23 0.13 0.07 0.44 0.25 0.16 3.05 2.24 0.23
Green Party 0.20 . 0.32 0.14 0.13 0.54 0.36 0.12 3.33 2.35 0.28

27

Electronic copy available at: [Link]

Table II: Key Survey Questions
This table tabulates responses to which extent respondents agree or disagree to statements related to environmental views
and household behaviors. The questions have been translated from Swedish into English. Boldface indicates the definition of
dummies, taking the value of one and zero otherwise. There are 3,993 responses except for the pension planning question, which has
3,978 responses.

A. Environmental Views
• Clean planet. “A clean planet is more important to me than financial welfare.”
1. Strongly agree (1,015, 25.4%)
2. Agree (1,561, 39.1%)
3. Neither Agree nor Disagree (1,016, 25.4%)
4. Disagree (203, 5.1%)
5. Strongly disagree (96, 2.4%)
6. Don’t know/Prefer not to say (102, 2.6%)

• Higher return. “Environmentally sustainable investments generate higher returns in the long run.”
1. Strongly agree (551, 13.8%)
2. Agree (1,198, 30.0%)
3. Neither Agree nor Disagree (1,490, 37.3%)
4. Disagree (312, 7.8%)
5. Strongly disagree (196, 4.9%)
6. Don’t know (246, 6.2%)

• Higher cost. “It is worth paying higher fees for a mutual fund that only make environmentally sus-
tainable investments.”
1. Strongly agree (361, 9.1%)
2. Agree (1,249, 31.6%)
3. Neither Agree nor Disagree (1,232, 31.2%)
4. Disagree (447, 11.3%)
5. Strongly disagree (320, 8.1%)
6. Don’t know (342, 8.7%)

B. Household behaviours
• Green products. “I am willing to pay more for environmentally friendly products.”
1. Strongly agree (1,138, 28.5%)
2. Agree (1,623, 40.6%)
3. Neither Agree nor Disagree (723, 18.1%)
4. Disagree (276, 6.9%)
5. Strongly disagree (161, 4.0%)
6. Don’t know/Prefer not to say (72, 1.8%)

• Recycle more. “Which sentence best describes how much you recycle?”
1. A great deal more than my neighbors (581, 14.6%)
2. Somewhat more than my neighbors (1,144, 28.6%)
3. About the same as my neighbors (1,865, 46.7%)
4. My neighbors recycle somewhat more than I do (223, 5.6%)
5. My neighbors recycle a great deal more than I do (54, 1.4%)
6. Don’t know/Prefer not to say (126, 3.2%)

• Never checks pension. “How often do you review you retirement savings?”
1. Several times per year (658, 16.5%)
2. Once a year (1,371, 51.0%)
3. When I start or change my pension plan (617, 15.5%)
4. Never (1,030, 25.9%)
5. Don’t know/Prefer not to say (302, 7.6%)

28

Electronic copy available at: [Link]

Table III: Holdings of Stocks and Pension Funds
This table display details of stock ownership data obtained from Euroclear and mutual fund data from the Premium Pension
Agency (PPA). Panel A displays market capitalization and industry carbon emissions for 476 firms that appear in the holdings of 976
investors in sample. All firms are matched to two-digit industry Scope 1 carbon dioxide emissions obtained from Statistics Sweden.
The size of the portfolios are reported in quintiles of market value and number of stocks. Panel B displays details of the pension data
at the end of 2017. There were 892 mutual funds available for investment of which 331 labeled environmental or ethical compliant
(ESG). The mutual fund sample contains 3,667 respondents broken up on those who have made an active choice and so are not in the
default fund. ESG-holdings are presented separately for funds with names suggesting that they focus on sustainable investments
(ESG name) where reported fractions corresponds to those having a positive weight in each category.

Panel A: Stock holdings

Firms Mean Q1 Q2 Q3 N
Market cap (SEK m) 18,038 256 1,287 8,407 476
CO2 emissions (kilotons) 463 62 145 795 476

Portfolios
No. of stocks 4.4 1 2 5 976
Portfolio value (SEK k) 597.5 6.8 33.3 121.7 976

Panel B: Pension accounts

Funds All ESG-Labeled ESG-Named

No. of funds 892 331 46
Percent (of All) 100.0% 37.1% 5.2%

Activity All Active Default

No. of investors 3,667 2,474 1,193
Percent (of All) 100.0% 67.5% 32.5%

ESG holdings All ESG-Labeled ESG-Named

No. of investors 2,474 1,827 167
Percent (of All) 100.0% 73.8% 6.8%

29

Electronic copy available at: [Link]

Table IV: Do Green Households Hold Green Pensions?
This table presents the results of Probit regressions where the dependent variable takes the value of one for the respondents
having an all ESG portfolio (100% allocation) under two assumptions regarding the pro-environmental orientation of the default
fund. Columns (1) through (3) classifies the default option as being a non-ESG fund, and Columns (4) through (6) as being an ESG
fund. Independent variables include log of disposable income and age which is scaled and divided by ten. Urban and Green party
measures the population density and share of Green Party votes within the municipality of the respondent. Female. University,
Env./Bio and Bus./Eco. student are indicator variables for females, subjects having at least completed one university course or
having studied Economics/Business or Biology/Geography/Environmental science at any level since high school. Columns (7)
and (8) splits the sample into those that answered all the “Big-3” financial literacy correctly or incorrectly (labeled Low versus High
financial literacy). The survey data is matched to 3,667 accounts obtained from the Premium Pension Agency (PPA). The table reports
marginal probabilities. Robust standard errors within parenthesis where *, ** and *** denote significance at the 10%, 5% and 1% level.

VARIABLES Active ESG Passive ESG

(1) (2) (3) (4) (5) (6) (7) (8)

Clean planet -0.026* -0.032* 0.031 0.009

(0.016) (0.017) (0.020) (0.022)
Higher cost -0.005 0.044
(0.026) (0.031)
Higher return 0.024 0.046*
(0.023) (0.027)
Green preferences 0.001 0.023** 0.010 0.029**
(0.009) (0.011) (0.020) (0.014)
Env. Lit. -0.002 -0.002 -0.003 -0.003 -0.003 -0.002 -0.016 0.006
(0.007) (0.007) (0.007) (0.009) (0.009) (0.009) (0.014) (0.011)
Fin. Lit. -0.000 -0.000 -0.001 -0.030*** -0.030*** -0.030*** 0.002 -0.032**
(0.006) (0.006) (0.006) (0.008) (0.008) (0.008) (0.016) (0.016)
Log income 0.024*** 0.024*** 0.025*** -0.044*** -0.044*** -0.045*** -0.033 -0.050**
(0.009) (0.009) (0.009) (0.014) (0.014) (0.015) (0.020) (0.019)
Age 0.075*** 0.075*** 0.075*** -0.123*** -0.122*** -0.123*** -0.134*** -0.113***
(0.006) (0.006) (0.006) (0.007) (0.007) (0.007) (0.011) (0.009)
Female 0.021 0.021 0.020 0.032* 0.032* 0.029 -0.011 0.039*
(0.014) (0.014) (0.015) (0.018) (0.018) (0.018) (0.031) (0.022)
Urban -0.010* -0.010* -0.010* 0.001 0.001 0.001 0.010 -0.002
(0.005) (0.005) (0.006) (0.007) (0.007) (0.007) (0.012) (0.008)
Green party 0.560** 0.565** 0.544** 1.091*** 1.092*** 1.085*** 0.694 1.249***
(0.276) (0.276) (0.276) (0.351) (0.351) (0.351) (0.632) (0.420)
University -0.008 -0.007 -0.009 0.004 0.004 0.004 0.064** -0.022
(0.015) (0.015) (0.015) (0.019) (0.019) (0.019) (0.031) (0.023)
Env./Bio. student 0.022 0.022 0.021 0.044 0.040 0.043 0.011
(0.054) (0.054) (0.054) (0.062) (0.062) (0.062) (0.069)
Bus./Econ. student -0.001 -0.002 -0.001 -0.053* -0.054* -0.053* -0.022 -0.060*
(0.023) (0.023) (0.023) (0.029) (0.029) (0.029) (0.055) (0.033)

Sample PPA PPA PPA PPA PPA PPA Low High

Fin Lit Fin Lit
Observations 3,667 3,667 3,667 3,667 3,667 3,667 1,228 2,439
Pseudo -R2 0.048 0.048 0.047 0.086 0.087 0.086 0.108 0.072

30

Electronic copy available at: [Link]

Table V: Green Beliefs and Green Pension Choices
This table presents the results of Probit and OLS regressions where the dependent variable equals one for respondents that
at some point made an active choice to opt out of the default fund (Active choice) in Columns (1) through (3), and zero otherwise.
The dependent variable in Columns (4) through (7) is weight in ESG-labeled funds (ESG-Labeled), and in Columns (8) through
(10) the weight in funds with names suggesting that the fund is devoted to sustainable investments (ESG-Named). Column (7)
displays the results for the 1,738 respondents of active choice investors that answered the “Big 3” financial literacy questions correctly.
Independent variables and characteristics follow those in Table IV. Fund controls include fund categories and fees. The survey data
is matched to 3,667 accounts obtained from the Premium Pension Agency (PPA). The table reports marginal probabilities. Robust
standard errors within parenthesis where *, ** and *** denote significance at the 10%, 5% and 1% level.

Active Choice ESG-Labeled ESG-Named

VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Clean planet -0.066*** -0.050** -0.000 -0.015 0.012** 0.010

(0.020) (0.021) (0.017) (0.019) (0.006) (0.006)
Higher cost -0.048 0.031 0.017
(0.031) (0.031) (0.011)
Higher return -0.022 0.036 -0.001
(0.026) (0.024) (0.008)
Green preferences -0.022* 0.015 0.024** 0.009***
(0.011) (0.010) (0.012) (0.003)
Env. Lit. -0.002 -0.002 -0.003 -0.005 -0.005 -0.005 -0.007 0.001 0.001 0.001
(0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.002) (0.002) (0.002)
Fin. Lit. 0.033*** 0.033*** 0.033*** -0.006 -0.006 -0.007 0.001 0.003* 0.003* 0.003*
(0.008) (0.008) (0.008) (0.007) (0.007) (0.007) (0.013) (0.002) (0.002) (0.002)

Characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fund controls No No No Yes Yes Yes Yes Yes Yes Yes
Sample PPA PPA PPA Active Active Active High Lit. Active Active Active
Observations 3,667 3,667 3,667 2,474 2,474 2,474 1,738 2,474 2,474 2,474
Pseudo-R2 0.236 0.237 0.235 0.219 0.220 0.220 0.228 0.029 0.030 0.030

31

Electronic copy available at: [Link]

Table VI: Green Values and Direct Stock Ownership
This table presents the results of Probit regressions estimated on the full sample where the dependent variable equals one
for respondents owning stocks in Columns (1) through (3) and zero otherwise. The dependent variable in Columns (4) through (6) is
an indicator variables taking the value of one if the investor holds energy stocks and zero otherwise conditional on owning stocks.
No students of environmental science hold energy stocks, so this variable is dropped from the regressions. Columns (7) through (9)
reports OLS regressions where the dependent variable is the combined CO2 -emission implied by the portfolio as measured by its
industry composition. Independent variables follow those in Table IV and V. Data for Swedish stock holdings have been obtained
from Euroclear, and industry carbon emissions from Statistics Sweden. Pseudo R-squared is reported in Columns (1) through (6). The
table reports marginal probabilities. Robust standard errors within parenthesis where *, ** and *** denote significance at the 10%, 5%
and 1% level.

VARIABLES Stock Energy CO2

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Clean planet -0.028* -0.021 -0.017 -0.035** -0.006 0.000

(0.015) (0.017) (0.017) (0.017) (0.034) (0.038)
Higher cost 0.008 -0.006 -0.059
(0.025) (0.030) (0.056)
Higher return -0.031 0.093** 0.024
(0.020) (0.045) (0.050)
Green preferences -0.026*** -0.003 -0.003
(0.009) (0.010) (0.018)
Env. Lit. 0.001 0.001 0.001 0.010 0.010 0.010 -0.004 -0.005 -0.004
(0.007) (0.007) (0.007) (0.008) (0.008) (0.008) (0.015) (0.015) (0.015)
Fin. Lit. 0.061*** 0.061*** 0.061*** 0.002 0.003 0.002 0.030** 0.029** 0.030**
(0.007) (0.007) (0.007) (0.008) (0.008) (0.008) (0.014) (0.014) (0.014)
Log income 0.086*** 0.085*** 0.086*** -0.009 -0.007 -0.008 -0.024 -0.023 -0.024
(0.011) (0.011) (0.011) (0.012) (0.012) (0.012) (0.023) (0.023) (0.023)
Age 0.031*** 0.031*** 0.031*** -0.012* -0.011* -0.012* -0.016 -0.017 -0.016
(0.005) (0.005) (0.005) (0.006) (0.006) (0.006) (0.012) (0.012) (0.012)
Female -0.085*** -0.085*** -0.080*** -0.037** -0.037** -0.038** 0.008 0.008 0.009
(0.014) (0.014) (0.014) (0.015) (0.015) (0.015) (0.030) (0.030) (0.031)
Urban 0.006 0.006 0.005 0.005 0.004 0.005 0.002 0.002 0.002
(0.005) (0.005) (0.005) (0.007) (0.007) (0.007) (0.012) (0.012) (0.012)
Green party 0.056 0.042 0.069 -0.065 -0.013 -0.066 0.189 0.248 0.191
(0.272) (0.272) (0.272) (0.318) (0.314) (0.321) (0.592) (0.595) (0.593)
University 0.034** 0.033** 0.034** -0.015 -0.011 -0.015 -0.035 -0.033 -0.036
(0.014) (0.014) (0.014) (0.017) (0.017) (0.017) (0.032) (0.032) (0.032)
Env./Bio. student -0.023 -0.022 -0.022 0.156 0.156 0.156
(0.047) (0.047) (0.047) (0.105) (0.106) (0.105)
Bus./Econ. student 0.039* 0.039* 0.038* 0.033 0.033 0.033 0.004 0.006 0.004
(0.023) (0.023) (0.023) (0.030) (0.029) (0.030) (0.043) (0.043) (0.043)

Sample All All All Stock Stock Stock Stock Stock Stock
Observations 3,993 3,993 3,993 976 976 976 976 976 976
Pseudo-R2 0.094 0.095 0.096 0.034 0.049 0.032 0.011 0.013 0.011

32

Electronic copy available at: [Link]

 Table VII: Environmental Values and Green Engagement
This table presents the results of Probit regressions where the dependent variable equals one for the response is “Strongly agree” to the questions “I am willing to pay more for envi-
ronmentally friendly products” (Green products) in Columns (1) through (3); “I recycle a great deal more than my neighbors” (Recycle much more) in Columns (4) through (6); and if the
response is “Never” to the question “How often do you check your pension saving account(s)?” (“Never checks pension”) in Columns (7) through (9). In columns (10) through (12) the dependent
variable is a dummy for strongly agreeing to “I find environmental matters interesting,” while in columns (13) through (15) the dummy corresponds to strongly agreeing to the statement “I find
financial matters interesting.” The dependent variables take the value of zero otherwise. Independent variables and characteristics follow those in Table IV, V and VII. The table reports marginal
probabilities. Robust standard errors within parenthesis where *, ** and *** denote significance at the 10%, 5% and 1% level.

VARIABLES Green products Recycle much more Never checks pension Environmental interest Financial interest
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)

Clean planet 0.415*** 0.294*** 0.101*** 0.062*** 0.048*** 0.040** 0.405*** 0.296*** 0.030* -0.005
(0.017) (0.020) (0.015) (0.015) (0.017) (0.018) (0.017) (0.020) (0.016) (0.017)

33
Higher cost 0.403*** 0.068*** 0.035 0.253*** -0.013
(0.034) (0.023) (0.026) (0.035) (0.024)
Higher return 0.311*** 0.077*** -0.000 0.371*** 0.144***
(0.027) (0.019) (0.021) (0.026) (0.024)
Green preferences 0.308*** 0.049*** -0.009 0.311*** 0.006
(0.012) (0.008) (0.010) (0.013) (0.009)
Env. Lit. 0.012 0.012 0.014* 0.014** 0.013** 0.015*** 0.005 0.004 0.006 0.023*** 0.023*** 0.026*** -0.007 -0.007 -0.006
(0.007) (0.008) (0.007) (0.006) (0.006) (0.006) (0.007) (0.007) (0.007) (0.008) (0.008) (0.008) (0.007) (0.007) (0.007)
Fin. Lit. 0.027*** 0.029*** 0.024*** -0.004 -0.004 -0.004 -0.029*** -0.029*** -0.028*** 0.029*** 0.032*** 0.025*** 0.051*** 0.052*** 0.051***
(0.007) (0.007) (0.007) (0.005) (0.005) (0.005) (0.006) (0.006) (0.006) (0.007) (0.008) (0.007) (0.006) (0.006) (0.006)
Characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993 3,993
Pseudo R2 0.161 0.240 0.223 0.0298 0.0403 0.0259 0.0766 0.0770 0.0750 0.130 0.189 0.176 0.0466 0.0575 0.0458

Electronic copy available at: [Link]

A Sampling
This Appendix presents the data collection and matching procedure in detail. In early
2018, Statistics Sweden (SCB) mailed out 19,977 invitations to a random sample of Swedes
aged 18-65. The invitation contained information about the purpose of the survey, how
to log on to the response website at SCB, what registry data that was going to be used
and matched to the survey responses if the respondent agreed to participate, and contact
details to SCB and one of the authors in case of questions. Respondents were not offered
compensation for their participation.
In the next step, SCB collected and matched individual pension data and stock hold-
ings to the survey which was supplied by the Premium Pension Agency (PPA) and from
Euroclear Sweden. All identities are scrambled, and the analysis was conducted through
the mainframe computer situated at the SCB from which the authors only can retrieve
and keep the SCB aggregated results and not the data itself.
The procedure followed all standards applied by SCB and the project has been ap-
proved by the Swedish Ethical Review Authority. SCB calibrated the sample to an under-
lying population of 6,097,316 Swedes in the ages 18-65 using gender, age, country of birth,
income, highest level of education, marital status and history of citizenship (details of the
mailer, population weighting procedure and protocol over the data collection is available
from the authors upon request).
Table A.1 summarizes this information, where Panel A gives details about the data
collection and Panel B dislplays the matching procedure.

Table A.1: Sample selection

This table display details of the sample construction. Panel A displays details about the survey invitation and responses, and Panel B
displays details of the matching procedure and deletions due to missing data.

Panel A: Survey invitations

Note Responses % of Total Date
First invitation 1,795 9.0 February 7, 2018
Reminder 1 1,545 7.7 February 22, 2018
Reminder 2 917 4.6 March 13, 2018
Total responses 4,257 21.3
No response 14,819 74.2
Returned mail 281 1.4
Declined 557 2.8
Not found 63 0.3
Total invites 19,977 100.0

Panel B: Matching
Note Responses % of Total Deletions Remark
Total responses 4,257 21.3 0 Initial sample
Deletion 1 4,174 20.9 83 Missing SCB registry data
Deletion 2 3,993 20.0 181 Incomplete responses
Deletion 2 3,667 18.4 326 Missing PPA data

34

Electronic copy available at: [Link]

B Measuring Financial and Environmental Literacy
We use a version of the standard financial literacy test developed by Lusardi and Mitchell
(2007) adapted to the Swedish retirement savings context by Anderson and Robinson
(2018). The test uses the questions used in the National Financial Capability Study in the
US, but we replace the question about mortgages with a question about long-term sav-
ings and the required rate of return for doubling the value of an investment over ten years
because residential mortgages operate differently in Sweden. The first three financial lit-
eracy questions are direct translations from those of the NFCS, and we refer to them as
the “Big-3,” in keeping with the literature. Table I shows that, on average, respondents
got 3.14 questions correct. This is in line with many other studies using these survey
questions. When we apply survey weights to correct for underrepresentation of younger,
lower-income respondents, this number drops to 2.93. Average scores broken out by de-
mographic categories are tabulated in Table I. Univariate comparisons indicate that men
score higher than women, and that younger, less affluent and less educated households
living in rural areas have lower scores. These results coincide with the previous literature,
where Swedes also score higher on the test compared to many other nationalities (see, for
instance, Hastings, Madrian, and Skimmyhorn (2013) and Lusardi and Mitchell (2011)).

Table B.1 here

To measure environmental literacy in a similar manner, we develop a five-question test

intended to capture knowledge of household-related matters that connect to the environ-
ment. The test is designed to offer some simple questions that most individuals would be
expected to know, but to also offer some more challenging questions that require a more
subtle understanding of environmental processes.

Table B.2 here

Two questions focus on energy, three focus on biological and environmental phenom-
ena. Table B.2 summarizes the test. Correct answers are underlined, and response tallies
are provided in parentheses. The first question gauges knowledge about the tradeoffs
surrounding the use of regular versus low energy consumption lightbulbs, “A low en-
ergy lightbulb costs more than a regular lightbulb but uses less energy. About how long
does one last?” Compact-fluorescent (CFL) bulbs last about 10-times longer than standard
incandescent bulbs, whereas led light bulbs last as much as 25 times longer. Most respon-
dents get the correct answer (42.5%) but almost as many respondents (41.6%) think that
low energy bulbs last 100 times longer.9
To test for basic knowledge around one of the key culprits behind global warming, we
asked respondents about the purpose of the ozone layer: “The ozone layer filters what
harmful product?” Almost all respondents (86%) replied with the correct answer (“UV
radiation”), but almost 7% claimed to not know the answer.
The third question measures knowledge about food loss and food waste in the global
food supply chain: “According to the UN, around 30% of the world’s food production
is lost each year. When does this loss occur?” The data behind this question come from
9
This answer is especially puzzling in light of Allcott and Taubinsky (2015) who show that only 28
percent of residential sockets in the US hold CFL’s in 2010, making consumers lose as much as $15 billion
in potential energy savings.

35

Electronic copy available at: [Link]

the United Nations Food and Agriculture Organization (FAO). Over 62% of respondents
incorrectly believe that most food loss occurs at home, after food is purchased from the
grocery store. Although the amount of food is lost in the developed world in this man-
ner is large relative to the supply of food in the developing world, much more food is
lost earlier in the supply chain. In the developing world, the majority of food loss occurs
because of inefficiencies in harvesting, storage and transportation, whereas in the devel-
oped world, transportation systems are more efficient but large amounts of food waste
occurs because cosmetically unappealing, but otherwise perfectly healthy, produce is dis-
carded before it reaches the supermarket. Based on calculations presented in Gustavsson,
Cederberg, and Sonesson (2011), the correct answer is “Most food is lost before it reaches
the supermarket,” but only 11% of respondents answer accordingly. Around 18% report
that it occurs at the grocery store before the produce is sold and over 62% incorrectly
thinks that most food is lost at home. In fact, food loss and waste at the consumer level is
much lower compared to earlier stages in the supply chain even in the developed world.
An inter-connected question relates to global expenditures on heating and cooling:
“Does the world spend more energy on heating homes or cooling them?” In Sweden, as
in much of Northern Europe, home air-conditioning systems are relatively uncommon,
whereas heating is ubiquitous. The correct answer reflects two forces in tension. On the
one hand, many more people in the world live in warmer places than in cooler places,
which militates toward cooling being a larger energy sink than heating. But on the other
hand, air-conditioning is not widely available in these parts of the world. According to
the US Energy Information Administration (IEA), more energy is spent on heating than
cooling, but this relation is expected to change going forward where demand for cooling
to grow 33-fold to 10,000 terawatt hours (TWh) by year 2100—equivalent to roughly half
of the total electricity generated worldwide in 2010.10 Most respondents answered that
cooling was a bigger use of energy (39.9%), while only around 28% respond correctly.
One consequence of global warming that has received attention in the media is the dis-
placement of arctic animals like polar bears from their natural habitats. Our final question
is intended to capture a basic awareness of this process: “Why don’t polar bears eat pen-
guins?” Penguins are native to the Southern Hemisphere, while polar bears are native to
the Arctic region, which makes “None of the above” the correct answer. More than half
(58%) of respondents answer correctly, but a quarter report that they do not know. Over
ten percent of respondents indicate that polar bears do not eat penguins because they
have both been driven from their natural habitats, and just above two percent believe
that Polar bears are vegetarians.
Table I shows that the average score on this test was 2.27. The raw correlation be-
tween the two test scores is around 18%. Measuring this correlation allows us to contrast
between two hypotheses. One is that individuals who have knowledge in one domain
have knowledge in many domains because knowledge comes easily to them. Under that
hypothesis, the two measures would be positively correlated, and environmentally in-
formed views would naturally be expressed in market prices because the environmen-
tally informed would also be financially engaged. The alternative hypothesis is that
through virtue-signaling channel, or simply through selective engagement, individuals
specialize in certain types of knowledge, eschewing others. Contrasting these hypothe-
10
This phenomenon is also reported in an article published in the Guardian “World set to use more
energy for cooling than heating”, October 26, 2015. See Labriet et. al. (2015) for examples of assumptions
and projections.

36

Electronic copy available at: [Link]

ses is useful for understanding whether individuals with high degree of environmental
knowledge are likely to express their environmental preferences through their portfolio
choice decisions. To explore this further, we provide conditional correlations in the form
of regressions in the Appendix in Table B.3. The appendix table shows a low correlation
between environmental and financial literacy, and becomes also much weaker when con-
ditioning the sample on having strong pro-environmental views. In addition, there is a
very strong correlation between financial literacy and reporting a strong interest in per-
sonal financial matters. The correlation with financial interest and environmental literacy
is virtually zero. This supports the idea that environmental and financial knowledge have
little overlap, and particularly the low overlap between green value orientations and fi-
nancial knowledge and interest.
Table B.3 shows that the correlation between environmental literacy and financial lit-
eracy is low. Column (1) shows that the loadings on demographic characteristics for the
financial literacy test confirm what is found in many other studies (see Hastings, Madrian,
and Skimmyhorn (2013)). Higher income older males with university education score
higher on this test. In Column (2) we repeat the regression but only on the sub-sample of
“clean planet” respondents–those who strongly agree to the statement that a clean planet
is more important than financial welfare. The main difference in the columns is that the
positive correlation with income is weaker, as is the positive correlation with having stud-
ied business or economics. Both owe to the fact that there are fewer individuals with these
characteristics in the environmental values subsample.
Table B.3 here
Column (3) of Table B.3 shows that the score of the environmental literacy test is not
well explained with demographic variables. Respondents who studied biology or envi-
ronmental science at university score higher on the environmental literacy test, suggest-
ing that the test questions are well understood by those who have studied a formal cur-
riculum in this area. That correlation is much stronger in Column (4) when we focus on
the subsample of people with pro-environmental values as measured by the clean planet
statement. Columns (5) and (6) of Table B.3 explores the conditional correlation between
the two tests which is around 18%, and when the correlation is conditioned on those hold-
ing strong pro-social values, it drops to a mere 10%. This shows that the relation between
environmental and financial literacy is low, particularly for the environmentally engaged.
Finally, Column (7) and (8) of Table B.3 shows the relation between test scores and
interest in the financial and environmental domain in Probit regressions where the de-
pendent variable takes the value of one if strongly agreeing to the statements “I find
/ personal financial matters / environmental issues / interesting”, and zero otherwise.
The results show that there is a strong relation between higher financial literacy scores
and environmental interest, but no relation between environmental literacy and financial
interest, which further supports the previous finding of low overlap between environ-
mental and financial engagement.
Table B.4 presents the results from Probit regressions on environmental attitudes mea-
sured by strongly agreeing to the three statements “A clean planet is more important
to me than financial welfare” (Clean planet); “Environmentally sustainable investments
generate higher returns in the long run” (High returns); and “It is worth paying higher
fees for a mutual fund that only make environmentally sustainable investments” (High
cost), all zero otherwise.

37

Electronic copy available at: [Link]

 Table B.4 here

Columns (1) through (3) of Table B.4 shows that those valuing a clean planet over
financial welfare are younger, have higher education and income and more likely to be
female rather than male. These views are more likely to be held by people living in Green
party dominated areas, but this effect washes out when controlling for environmental lit-
eracy in Column (3), which also adds to the explanatory power of the regression. The
three measures are indeed highly correlated. Views about a clean planet is strongly re-
lated to both thinking that green investments outperform others in Column (4) and and
a willingness to pay higher fees for green mutual funds in Column (5). The point esti-
mates reveal that those valuing a clean planet are 27% and 19% more likely to agreeing
to the other two statements. Column (6) shows that the willingness to pay higher fees is
not explained solely by thinking returns are higher - a green pro-social value orientation
is strongly related to the willingness to pay higher fees for environmentally sustainable
funds. This shows that preferences for green investments is a mix of both pecuniary and
non-pecuinary motives.

38

Electronic copy available at: [Link]

Table B.1: Five modified financial literacy questions
This table presents the five (“Big-5”) financial literacy questions (the first three are referred to as the “Big-3” in the main text) used
in the study and corresponding frequency responses on each item. Correct answers are highlighted in boldface. The questions have
been translated from Swedish into English. There are 3,993 observations.

1. Compounding. Suppose you had $100 in a savings account and the interest rate was 2% per year.
After 5 years, how much do you think you would have in the account if you left the money to grow?
Please select one.
(a) More than $102 (3,633, 91.0%)
(b) Exactly $102 (68, 1.7%)
(c) Less than $102 (109, 2.7%)
(d) Don’t know (136, 3.4%)
(e) Prefer not to say (47, 1.2%)

2. Inflation. Imagine that the interest rate on your savings account was 1% per year and inflation was
2% per year. After 1 year, how much would you be able to buy with the money in this account?
Please select one.
(a) More than today (199, 5.0%)
(b) Less than today (3,065, 76.8%)
(c) Exactly the same as today (146, 3.7%)
(d) Don’t know (516, 12.9%)
(e) Prefer not to say (67, 1.7%)

3. Diversification. Buying a single company’s stock usually provides a safer return than a stock mutual
fund. Please select one.
(a) True (225, 5.6%)
(b) False (3,261, 81.7%)
(c) Don’t know (458, 11.5%)
(d) Prefer not to say (49, 1.2%)

4. Long-Term Savings. Suppose you were given $10,000 as a gift and wanted to double the amount by
saving the money ten years without having to touch it. What interest rate would you require to
achieve this goal? Please select one.
(a) About 15% annual interest rate (254, 6.4%)
(b) About 10% annual interest rate (1,552, 38.9%)
(c) About 7% annual interest rate (1,764, 44.2%)
(d) Don’t know (353, 8.8%)
(e) Prefer not to say (70, 1.8%)

5. Bond Pricing. If interest rates fall, what should happen to bond prices? Please select one.
(a) They will rise (676, 16.9%)
(b) They will fall (818, 20.5%)
(c) They will stay the same (1,680, 42.1%)
(d) Don’t know (752, 18.8%)
(e) Prefer not to say (67, 1.7%)

39

Electronic copy available at: [Link]

Table B.2: Measuring Environmental Literacy
This table reports the questions and responses of the environmental literacy test where correct choices are underlined. The
questions have been translated from Swedish into English. There are 3,933 individuals in the sample.

1. A low-energy (CFL or LED) lightbulb costs more than a regular lightbulb but uses less en-
ergy. About how long does one last?
(a) About the same as a regular lightbulb (67, 1.7%)
(b) About 10 times as long as a regular lightbulb (1,709, 42.8%)
(c) About 100 times as long as a regular lightbulb (1,662, 41.6%)
(d) Don’t know/Prefer not to say (555, 14.9%)

2. The ozone layer filters what harmful substance?

(a) Acid rain (25, 0.6%)
(b) UV radiation (3,466, 86.8%)
(c) Sewage gas (57, 1.4%)
(d) The Greenhouse Effect (187, 4.7%)
(e) Don’t know/Prefer not to say (258, 6.5%)

3. According to the UN, around 30% of the world’s food is lost each year. When does this loss
occur?
(a) Most food is lost before it reaches the supermarket (445, 11.1%)
(b) Most food is discarded at the supermarket before it is sold (722, 18.1%)
(c) Most food is wasted after it is purchased from the supermarket (2,486, 62.3%)
(d) Don’t know/Prefer not to say (340, 8.5%)

4. Does the world spend more energy on heating homes or cooling them?
(a) More energy on heating (1,109, 27.8%)
(b) More energy on cooling (1,610, 40.3%)
(c) About the same amount on both (559, 14.0%)
(d) Don’t know / Prefer not to say (715, 17.9%)

5. Why don’t polar bears eat penguins?

(a) They have both been driven out of their natural environment (457, 11.4%)
(b) Polar bears do not eat meat (83, 2.1%)
(c) Penguins are only active when polar bears hibernate (133, 3.3%)
(d) None of the above (2,339, 58.6%)
(e) Don’t know/Prefer not to say (981, 24.6%)

40

Electronic copy available at: [Link]

Table B.3: The Demographics of Financial and Environmental Literacy Scores
This table reports results from OLS regressions where the dependent variable is the score of the financial literacy test in
Columns (1) and (2), and environmental literacy tests in Columns (3) through (6), both ranging from zero to five. The dependent
variable in Columns (7) and (8) takes the value of one for those strongly agreeing to the two statements “I find personal financial
matters / environmental issues / interesting”. Independent variables include log of disposable income and age which is scaled and
divided by ten. Urban and Green Party measures the population density and share of Green Party votes within the municipality
of the respondent. Female, University, Env./Bio and Bus./Eco. student are indicator variables for females, subjects having at least
completed on course at the university-level or having studied Economics/Business or Biology/Geography/Environmental science at
any level since high school. There are 3,993 in the full sample (labeled “All”), and 1,015 in the subsample of respondents who strongly
prioritize a clean planet over financial welfare (labeled “CP: Clean planet”). Robust standard errors within parenthesis where *, **
and *** denote significance at the 10%, 5% and 1% level.

VARIABLES Literacy Interest

Financial Environmental Fin. Env.
(1) (2) (3) (4) (5) (6) (7) (8)

Fin. Lit. 0.175*** 0.104*** 0.049*** 0.026***

(0.015) (0.029) (0.006) (0.007)
Env. Lit. -0.006 0.031***
(0.006) (0.007)
Log income 0.097*** 0.054*** 0.025* 0.002 0.008 -0.004 0.000 -0.019***
(0.015) (0.019) (0.014) (0.021) (0.014) (0.022) (0.005) (0.006)
Age 0.062*** 0.037 -0.032** -0.036 -0.043*** -0.040 0.005 -0.000
(0.014) (0.027) (0.012) (0.025) (0.012) (0.024) (0.005) (0.005)
Female -0.659*** -0.668*** -0.115*** -0.081 -0.000 -0.012 -0.084*** 0.052***
(0.035) (0.068) (0.032) (0.061) (0.033) (0.064) (0.013) (0.015)
Urban 0.015 -0.005 -0.014 -0.033 -0.016 -0.033 0.000 0.013**
(0.014) (0.027) (0.013) (0.025) (0.013) (0.025) (0.005) (0.006)
Green party 1.678** 1.872 1.332** 0.843 1.039 0.648 0.071 0.318
(0.695) (1.304) (0.653) (1.199) (0.646) (1.196) (0.263) (0.303)
University 0.689*** 0.638*** 0.160*** 0.063 0.039 -0.003 0.009 0.075***
(0.037) (0.076) (0.033) (0.065) (0.034) (0.067) (0.014) (0.016)
Env./Bio. student 0.199* 0.200 0.282*** 0.346** 0.247** 0.325** -0.119*** 0.233***
(0.108) (0.176) (0.107) (0.149) (0.103) (0.147) (0.038) (0.058)
Bus./Econ. student 0.455*** 0.399*** -0.095* -0.151 -0.175*** -0.193 0.092*** -0.021
(0.055) (0.116) (0.052) (0.123) (0.052) (0.122) (0.024) (0.025)
Constant 1.308*** 2.136*** 2.022*** 2.644*** 1.793*** 2.423*** 0.075 0.287***
(0.203) (0.285) (0.189) (0.308) (0.188) (0.317) (0.068) (0.080)

Sample All CP All CP All CP All All

Observations 3,993 1,015 3,993 1,015 3,993 1,015 3,993 3,993
R-squared 0.183 0.169 0.017 0.013 0.053 0.026 0.047 0.041

41

Electronic copy available at: [Link]

Table B.4: The Correlation in Environmental Attitudes
This table presents the results of a Probit regression where the dependent variable equals one if the answers are “Strongly
agree” to the statements “A clean planet is more important to me than financial welfare” (Clean planet) reported in Columns (1)
through (3); “Green investments generate higher returns in the long run” (“Higher return”) reported in Column (4) and “I am willing
to pay higher fees for mutual funds that only make sustainable investments” (“Higher cost”) in Columns (5) and (6) ); all zero
otherwise. Independent variables follow those in Table B.3. The table reports marginal probabilities. Robust standard errors within
parenthesis where *, ** and *** denote significance at the 10%, 5% and 1% level.

VARIABLES Clean planet Higher return Higher cost

(1) (2) (3) (4) (5) (6)

Higher return 0.089***

(0.016)
Clean planet 0.270*** 0.189*** 0.149***
(0.016) (0.014) (0.014)
Env. Lit. 0.029*** 0.003 0.001 0.001
(0.007) (0.005) (0.004) (0.004)
Fin. Lit. 0.008 0.003 0.000 0.000 0.001
(0.006) (0.006) (0.005) (0.004) (0.003)
Log income -0.022*** -0.023*** -0.023*** -0.009*** -0.002 -0.001
(0.005) (0.005) (0.005) (0.004) (0.002) (0.002)
Age 0.008 0.008 0.009* -0.011*** -0.009*** -0.009***
(0.005) (0.005) (0.005) (0.004) (0.003) (0.003)
Female 0.049*** 0.054*** 0.054*** 0.010 0.003 0.003
(0.014) (0.014) (0.014) (0.011) (0.008) (0.008)
Urban 0.006 0.006 0.006 0.004 0.001 0.000
(0.006) (0.006) (0.006) (0.004) (0.003) (0.003)
Green party 0.467* 0.454 0.426 -0.191 0.246* 0.274*
(0.284) (0.284) (0.284) (0.201) (0.144) (0.140)
University 0.075*** 0.070*** 0.069*** -0.018 0.030*** 0.031***
(0.014) (0.015) (0.015) (0.011) (0.008) (0.008)
Env./Bio. student 0.044 0.043 0.034 0.032 0.029 0.025
(0.055) (0.054) (0.054) (0.043) (0.031) (0.030)
Bus./Econ. student -0.026 -0.030 -0.025 0.011 0.008 0.005
(0.023) (0.023) (0.023) (0.018) (0.014) (0.013)

Observations 3,993 3,993 3,993 3,993 3,993 3,993

Pseudo
R-squared 0.022 0.022 0.026 0.134 0.153 0.175

42

Electronic copy available at: [Link]

C List of variables
This Appendix contains a brief variable description of the variables used in the analysis.

43

Electronic copy available at: [Link]

Table C.1: Variable definitions
This table presents a short description of all key variables grouped into those collected from the survey (Survey), matched to
the Statistics Sweden registry (Characteristics) and matched to the Premium Pension Authority on pensions and Euroclear individual
stock holdings (Decisions). Dummy variables takes the value of one as described and zero otherwise. Variables within each group are
sorted as they appear in the main text.

1. Survey variables
• Clean planet Dummy variable for strongly agreeing to the statement “A clean planet is more im-
portant for me than financial welfare”.
• Higher cost Dummy variable for strongly agreeing to the statement “It is worth paying higher fees
for a mutual fund that only make environmentally sustainable investments”.
• Higher return Dummy variable for strongly agreeing to the statement “Environmentally sustain-
able investments generate higher returns in the long run”.
• Green preferences The average score across the three statements Clean planet, Higher cost and
Higher return grading the responses on a five point scale from -2 (Strongly disagree) to 2 (Strongly
agree).
• Financial interest Dummy variable for strongly agreeing to the statement “I find personal financial
matters interesting”.
• Environmental interest Dummy variable for strongly agreeing to the statement “I find environmen-
tal matters interesting” .
• Green products Dummy variable for strongly agreeing to the statement “I am willing to pay more
for environmental friendly products”.
• Recycle much more Dummy variable if responding “I recycle a great deal more than my neighbors”.
• Never checks pension Dummy variable if the response is “Never” to the question “How often do
you review your retirement savings?”.
2. Characteristics
• Female Dummy variable for females.
• Log income The natural log of disposable income.
• University Dummy variable for post high school education.
• Studied Env / Bio Dummy variable for post high school education in biology/geography/natural
science.
• Studied Econ / Bus Dummy variable for post high school education in economics or business.
• Urban Population density measured at the municipality level (290 areas).
• Green party Election outcome of the Swedish Green Party (Miljöpartiet) measured at the munici-
pality level in 2014.
3. Decisions
• Active choice Dummy variable if the respondent ever made an active choice in the Swedish Pre-
mium pension system and zero otherwise.
• ESG-Labeled The portfolio weight in ESG-labeled funds in the Swedish premium pension system.
• ESG-Named The portfolio weight in funds with names containing the Swedish-language or
English-language words for SRI, ESG, water, sustainable, environmental, ethical, earth, clean or
green.
• Energy Dummy variable if the portfolio holds any position in the Energy sector according to the
two-digit Swedish industry SNI code standard.
• CO2 The portfolio’s industry carbon dioxide emission implied by the Scope 1 industry SNI classi-
fication from Statistics Sweden.
• Stock Dummy variable if the respondent holds individual stocks.

44

Electronic copy available at: [Link]