Talking About the Weather: Understanding Household Demand ...€¦ · “green˙investment˙v5”...

ii

“green˙investment˙v5” — 2019/3/15 — 19:35 — page 1 — #1 ii

ii

ii

Talking About the Weather:

Understanding Household Demand for Green

Investment

March 15, 2019

Abstract

In the summer of 2014, numerous weather events caused catastrophes throughout

Scandinavia. In the wake of these events, some investors came to hold beliefs about

expected future climate change that lie well outside the generally accepted scientific

consensus, dramatically overweighting the probability of unlikely adverse climate

scenarios. These investors traded into mutual funds that were labeled ‘green’ after

2014, but not before. This result is robust to the level of financial and environmental

knowledge, as well as pecuniary and non-pecuniary motives for investing. In sum,

weather calamities drive behaviorally biased investors into investment vehicles that

tout environmental sustainability.

1

ii


ii

ii

1 Introduction

Talking about the weather is often the perfect icebreaker, especially in Sweden, where

the summer of 2014 saw the highest temperatures on record throughout many parts of

Scandinavia. By the end of 2014, the Swedish Meteorological and Hydrological Insti-

tute (SMHI) had recorded new average annual high temperature records at 47 out of 100

weather stations throughout the country. These unusually high temperatures were asso-

ciated with an excessive number of weather-related anomalies, such as droughts, floods

and lightning strikes. Indeed, one of the worst wildfires in Sweden since the 1950s oc-

curred that summer, destroying an area the size of the District of Columbia.

In this paper, we show that these extreme weather events shaped opinions about the

environment, causing some individuals to hold extreme opinions about the likelihood of

future weather events. We then show that these opinions increase the likelihood of mak-

ing investments in mutual funds that claim to invest in an environmentally sustainable

way.

To do this, we administered a survey to almost 4,000 randomly selected Swedish

households in January and February 2018. The survey contains questions about both

environmental knowledge and beliefs about future climate-related calamaties such as

global temperature increases, food shortages and rising sea levels. We then match sur-

vey responses to detailed government registry data on household socio-economic status

and retirement savings choices. This allows us to connect beliefs about calamities along

with broader measures of financial and environmental knowledge and sophistication to

actual investment decisions.

Some respondents think that future climate-related catastrophes are very likely, far

more likely than would be implied by the accepted scientific consensus for worst case.

Their tendency to overweight low probability extreme events suggests that they are well

described by elements of prospect theory (Tversky and Kahneman (1992)). Indeed, we

show that increased exposure to localized weather calamities increases the probability of

holding such beliefs, which in turn demonstrates that these beliefs arise from a recency or

familiarity bias. After the summer of 2014—but not before—investors holding these be-

liefs trade their retirement accounts into positions that are heavily concentrated in mutual

fund holdings that are labeled as environmentally sustainable. The fact that their move-

2

ii


ii

ii

ment into “green” mutual funds occurs only after the calamities of 2014 but not before

suggests that extreme weather events caused certain individuals to overweight the prob-

ability of accelerated global warming, and that these fears then drove them into green

investments.

A natural alternative hypothesis is that investors simply see the risk return tradeoff

of green investments differently after witnessing weather calamities firsthand. To cap-

ture this, we measured alternative motives for holding green funds, such as whether the

respondent thinks green investments outperform, or whether they are driven by moral

considerations. Increased exposure to localized weather calamities does not make these

beliefs more likely. Nevertheless, our results hold even after controlling for these factors.

Moreover, we also measure the financial and environmental literacy of these respondents;

these factors have little impact on investment choices.

Our results add to a growing literature on socially responsible investing in general.

Our paper is especially connected to recent paper by Hartzmark and Sussman (2018),

which studies how the introduction of Morningstar ESG ratings affects mutual fund

flows. Their work focuses on aggregate fund flows in and out of high and low scoring

funds and suggests that investor irrationality is an important component of this effect.

Our work offers an account of the underlying behavioral mechanisms behind this result,

and is the first paper to consider how those interact with green investment decisions.

The remainder of the paper is structured as follows. First we describe the context of

the Swedish heat wave of 2014. This is provided in Section 2. Then we provide the demo-

graphics of our study population and present basic summary statistics for the questions

in our survey. This is presented in Section 3. Section 4 provides our main results. Section

5 concludes.

2 The year 2014: A Catastrophic Summer in Sweden

The weather in 2014 was exceptionally warm throughout the entirety of Sweden. The

Swedish Meteorological and Hydrological Institute (SMHI) recorded new average annual

temperature records at 47 out of 100 weather stations throughout the country.1 The sum-

1SMHI publication, The Year 2017 - Air Temperature, downloaded from www.smhi.se.

3

ii


ii

ii

mer of 2014 saw an extreme heatwave with average temperatures in July being 4 (in the

south) and 6 (in the north) degrees centigrade higher than normal. Figure 1 displays heat

maps obtained from SMHI for the month July, and compares the average temperature

in 2014 to those in 2012 to 2017. There were between 6 to 26 “heat wave days” (defined

as days with temperatures above 25 Centigrades) in the ten most northern weather sta-

tions above the Arctic Circle during 2014. By comparison, the norm would be zero or

only a handful of days in neighboring years. Hot air and humidity also released a record

number of lightning strikes, adding to the dramatic weather conditions.

Figure 1 here

High temperatures in July caused drought in many areas and are thought to be responsi-

ble for the worst Swedish wildfire since the 1950s. The fire eventually destroyed approx-

imately 150 square kilometers (or more than 37,000 acres, similar in size to the District of

Columbia) of forest and took eight months to extinguish. But more extreme weather was

to follow. In southern Sweden, there was a record of 100 millimeters of rainfall in Malmo

on a single day (August 31st), and later in the fall heavy rains in the west side of Sweden

caused severe floods in many cities, shutting down roads and trains and prompting citi-

zens to evacuate their homes. Figure 2 illustrates the severity of this shock by making a

similar yearly comparison for rainfall in the month of August between 2012 to 2017.

Figure 2 here

These events created a tremendous upsurge in public discourse surrounding climate

change. Figure 3 plots the number of average high temperature records for the 100

weather stations throughout the country along with then number of articles in the four

largest Swedish newspapers (Aftonbladet, Dagens Nyheter, Expressen and Svenska Dag-

bladet) that contain the phrase “climate change” in them.2 The spike in news relevance

beginning mid-way through 2014 and persisting through 2017 clearly illustrates that peo-

ple were talking about the weather in Sweden.

Figure 3 here

2Relatedly, Choi, Gao, and Jiang (2018) show that Google searches on climate change spike in the wakeof local high temperatures.

4

ii


ii

ii

Concomitantly, there was rapid growth in the number of financial products describ-

ing themselves as adhering to environmental, social or governance standards (ESG), es-

pecially throughout Europe. In Sweden, the Swedish Pension Authority started to label

ESG mutual funds in 2004, when 7% of around 700 mutual funds in the offering were

marked as such. It took almost ten years for their proportion to double to 13% in 2013,

but it quickly doubled again to 31% in 2016. As Figure 3 illustrates, By the end of 2017,

36% of all funds in the pension system have a pronounced ESG strategy of some sort,

even though there is no set industry standard for these guidelines.

3 Data and Empirical Setting

Our data are collected and matched in four steps. First, we worked in conjunction with

Statistics Sweden to administer a survey in January and February 2018. The survey in-

vitation was sent out by mail to 20,000 respondents, but respondents completed the sur-

vey online. Second, these survey results were matched to each respondent’s registry data

from various sources, including the Swedish Tax Authority, which is maintained by Statis-

tics Sweden. This step allows us to combine their environmental views with a large set

of demographic and wealth characteristics, including in which of the 290 municipalities

where the respondent lives. In the third step, we add the complete transaction histo-

ries—this includes the timing and size of any trades as well as the holdings at a point

in time—from the Swedish Pension Authority (SPA), the details of which are described

below. From the SPA, we also obtain data on fund characteristics, which allows us to

determine whether a fund is labelled as an ESG investment choice. The final step merges

on data from the Swedish Meteorological and Hydrological Institute (SMHI). From the

SMHI, we obtain weather warnings issued between 2010 and 2017. Warnings are issued

at a regional level; there are 21 distinct administritive regions. The warnings are graded

from 1 (some risks and disturbances to transport and other parts of society); 2 (danger,

damage and larger disturbances); and 3 (serious danger, serious damage and major dis-

turbances). We match regional warning data to the municipalities for which we have

survey data, which allows us to provide direct evidence of how exposure to the weather

calamities of 2014 affected public opinion.

5

ii


ii

ii

In the remainder of this section, we explain the Swedish pension system in more detail

and the supply of ESG funds. We then show the data on individual allocations in our

sample and explain our survey measures and results sorted on investor characteristics.

3.1 The Swedish Pension System

In June of 1994, the Swedish Parliament passed legislation that transformed the public

pension system from one based on defined benefit to one based partly on defined contri-

bution. A full account of this transition is beyond the scope of this paper; our purpose

here is to highlight the key features as they pertain to our analysis.3 Prior to 1994, the

system was a flat-rate universal benefit system augmented by an earnings-related sup-

plement. The period between 1994 and 1999 allowed for the accumulation of two types

of accounts. One is a defined contribution plan funded on a pay-as-you-go basis based

on a contribution rate of 16% of labor income, analogous to Social Security in the United

States. An additional 2.5% of labor income was credited to an individual account man-

aged by the pension authority. This operates in a manner similar to a 401(k) plan in the

United States, but as part of the state pension, rather than an occupational pension as is

common with 401(k) plans. Starting from 2000, individuals were allowed to control how

this account was invested by allocating this portion of their account across as many as

five different funds.4 In 2000, there were 456 funds available, a number that has grown

to 892 at the end of 2017. Swedish Pension Authority (SPA) started to label ESG funds in

2004, when 7% of around 700 funds in the offering were marked ESG funds. By the end

of 2017, 36% of all funds in the pension system have a pronounced ESG strategy of some

sort, even though there is not a set industry standard for these guidelines.

3The 1990s were a period of tremendous economic change and upheaval in Sweden more generally. Theexact details of the transition are complicated and are discussed at length in Palme, Sunden, and Soderlind(2007) and Palmer (1995).

4Cronqvist and Thaler (2004) and a follow-up paper Cronqvist, Thaler, and Yu (2018) discuss the roleof nudging people to make a choice or falling into the default fund. Anderson and Robinson (2018) usea similar approach to the one considered here in documenting how beliefs about fees and performanceaffected pension choices.

6

ii


ii

ii

3.2 Green Investment Options in the Swedish Pension System

The SPA maintains complete records of all trades and portfolio balances in the pension

system going back to 2000. We obtain these records for those respondents who completed

our survey. In addition, we manually collect fund characteristics from the catalogue that

is printed and mailed to first time savers as well as listed on the SPA website. The ESG

label was introduced in 2004, and lets fund companies label themselves as investing with

restrictions determined by ethical or environmental considerations (so-called negative se-

lection funds as in Hong and Kacperczyk (2009)). This information is required to be clearly

provided in all information and marketing about the funds, but there is no standard or

minimum requirements given by the SPA to which funds must adhere in order to earn

this label.

Figure 4 here

Figure 4 shows how total pension savings in our survey have evolved over time across

traditional and ESG investments. Prior to 2014, the amount of money trading into ESG

funds was tiny relative to the amount trading into non-ESG funds. For example, in 2012

58 million SEK was traded into non-ESG funds while only 3 million SEK were traded

into ESG funds. After the summer of 2014 the amount flowing into ESG funds grows

dramatically. In 2015, 31 million SEK were traded into ESG funds while 44 million SEK

were traded into non-ESG funds. By 2016 the trading into ESG funds matches non-ESG

funds, and by 2017 the total outstanding balances held in ESG funds outweigh the amount

held in non-ESG funds.

Panel A of Table I shows spectacular growth in the offering of ESG funds from 2010

to 2017, and Figure 3 traces out the full history from 1999 (which are the funds given

in the catalogue in 2000). In 2010, there was 839 funds offered in the system of which

89 were ESG funds - a fraction of almost 11%. Panel B of Table I shows a snapshot of

the porfolio holdings as of December 2017. We have 3,667 respondents after matching

with the pension data, of which 1,193 never made an active choice and so were still in

the default fund as of 2017.5 It is difficult to tell if respondents chose the default fund

because of its ESG profile, so we confine the rest of the analysis to those actively chosing5The default fund technically consists of two funds: an equity fund and a bond fund to which savers are

allocated depending on age.

7

ii


ii

ii

funds within the pension system. The third to the fifth row of Panel B sorts investors

in accordance with their ESG holdings defined for different weights. There were 1,827

investors with a positive share of ESG labelled funds in their pension portfolio, and 647

with no weight. At the threshold of at least 50% ESG funds, there were 1,434 investors

owning ESG and 1,040 not. Finally, 840 respondents held all ESG labelled funds at the

end of 2017, which corresponds to 34% of all those selecting their own funds—a fraction

that is close to the overall offering of funds.

Table I here

3.3 Surveying Swedish Environmental Beliefs

To measure environmental literacy and see how it squares with perceived literacy and

a general understanding of financial matters, we invited 20,000 Swedish households by

regular mail to participate in an online financial and environmental literacy survey.6 A

total of 4,257 respondents completed the survey where 3,993 remain after having deleted

incomplete responses, or almost 20% of the invited sample; this shrinks to 3,667 after hav-

ing matched them to the pension data. The survey is matched to registry data obtained

from Statistics Sweden, from where we obtain standard information on gender, age and

income, but also detailed information about level and subject of study and the area in

which people live (divided into 291 districts/municipalities). The location of individuals

also allows us to match on Green Party election outcomes, which is a control variable used

alongside population density in the regression analysis that follows. Table II provides a

demographic breakdown of the respondents.

Table II here

Relative to the overall Swedish population, we have an over-representation of older,

wealthier, better-educated respondents in our sample. Almost half of the individuals in

our sample went to college and 57% of our respondents are 45 or older, while only 41%

of the Swedish working age population is in this age range. Also, women are slightly

over-represented in our study. Statistics Sweden offers sampling weights that allow us to

6The results of the financial and environmental test and its questions are discussed in Anderson andRobinson (2019).

8

ii


ii

ii

adjust our regressions for these sampling differences, so that our results can be taken as

though they are drawn from a stratified random sample of the population.

Overall, we find that about a third in our sample is in the default fund, which is sim-

ilar to the results in Anderson and Robinson (2018). Default fund investors tend to be

lower income, young females. The default fund has since inception a clear ESG profile,

having an governance policy that not only screen companies, but also one that involves

expressing opinions about sustainable behavior through meetings with management and

shareholder voting. In this respect, one can argue that the default fund is a low cost al-

ternative to obtain an ESG portfolio for most investors, but it is of course also likely that

many investors fall into default because of their lack of interest in investments in general.

For this reason, we focus on the holdings and trades of active investors, which are those

that at some point in time chose a different portfolio. At the end of 2017 there were 3,667

individuals in our sample, of which 2,474 had chosen their portfolio.

The three last columns of Table II shows the fraction in our sample that had some

(a non-zero weight), most (over 50%) or all (100%) of their portfolio invested in an ESG

labelled fund. A majority of 74% had a non-zero portfolio weight in an ESG fund, but

that might not be so surprising given that the menu offering contains over 36% ESG funds

together with the fact that most people hold more than one fund. Increasing the threshold

to be above 50% decreases the share considerably, and about one-third of the sample hold

an ESG-only portfolio. Women and those having studied environmental science are more

likely to invest in these funds. For income, we find that wealthier individuals are more

likely to invest in ESG funds. We find that all in the youngest age group hold ESG funds,

but 98 per cent of these individuals are in the default fund. This suggests caution in

interpreting this number. In the next section we show that environmental concerns are

greater for younger and lower income respondents, but the results in Table II actually

shows that the fraction of all ESG portfolios increases somewhat with age.

3.4 Motivations for Green Investment

Our survey includes four basic types of questions: financial literacy questions, environ-

mental literacy questions, questions on household behavior, and questions that gauge

9

ii


ii

ii

one’s own perceptions of financial literacy and environmental matters.7

The main focus of this study is to measure possible motivations for holding green in-

vestments, where we specifically focus on climate change calamities. Even if temperature

change has been in focus of the policy discussion, its effects on food shortage and sea level

rise has been powerful illustrations of the consequences of extended periods of drought

and melting ice in the arctic region. We ask:

• “In the next 20 years...”

– “The average temperature on earth rises by more than one Centigrade”

– “Food shortage will increase”

– “The sea level will increase by over one meter”

The responses on a five point Likert scale ranging from “Very Unlikely” to “Very

Likely” are displayed in Figure 5 and shows that 80% of the respondents in our sample

finds a steep temperature change likely or very likely, 65% believes that food shortage will

increase and 47% that the world sea level will increase by more than one meter. The en-

vironmental calamity outcomes are obviously difficult to predict, but over a twenty year

horizon, they tend to be pessimistic outcomes compared to current scientific consensus.

Figure 5 here

The fact that almost 40% of respondents find that a one degree temperature change

is very likely within such a short time frame is surprising given current scientific pre-

dictions. According to the United Nations and the Intergovernmental Panel of Climate

Changes (IPCC), it is unlikely that the average temperature would rise by one degree

centigrade in twenty years, since the current temperature increase is measured to be

around a rate of 0.17 Centigrades per decade, and historically have been about one Centi-

grade since the early 1920’s. A one-half centigrade increase in global average temperature

per decade would be considered a worst-case scenario by these estimates.

Likewise, a sea-level increase of one meter in a twenty year period far exceeds con-

sensus estimates for sea level increases. For example, the IPCC report in their worst cace

7The financial literacy test is a modified version explained in Anderson and Robinson (2018) which isbased on Lusardi and Mitchell (2007) and Lusardi and Mitchell (2011).

10

ii


ii

ii

scenarios from 2014 that there is a 95% probability that the sea level rise will be less than

one meter by 2100.

Broadly speaking, global hunger and undernourishment have been decreasing over

the last twenty years. According to FAO, IFAD and WFP (2015), food shortage in many

regions of the world can be closely tied to conflicts and natural disasters, but generally

diminishes with economic growth. Thus, a belief in increased food shortage is also likely

to indicate a pessimistic outlook tied to environmental concerns.

We create a measure of a respondent’s focus on climate calamities by summing the

number of “Very Likely” responses for each of the three questions. These responses are

correlated. Of those finding a one degree temperature rise very likely, 46% and 29% also

foresee food shortages and a sharply rising sea level. Our measure therefore covers a

somewhat larger fraction, 47% of the sample compared to the 39% of those only fearing a

rise in temperature. A more comprehensive analysis of the individual responses to these

questions can be found in Anderson and Robinson (2019), in which the survey method

and results are described in detail.

Apart from fears, there is a range of other motives to invest into ESG. Those previously

considered can be categorized into pecuniary (“Doing well”) and non-pecuniary (“Doing

good”). Pecuniary motives are those that include aspects of behavior that eventually will

lead to better economic outcomes for everyone. The non-pecuniary motives are different

from the pecuninary in that they not necessarily lead to better economic outcomes, but

is a goal worthwhile sacrificing for a better or cleaner world. This set of motivations are

often also referred to as “warm-glow” motivations in that they also may give important

reputational externalities.

We attempt to measure the alternative mechanisms leading to ESG investments by

attitudes towards two statements in our survey, each measured on a five point Likert scale

ranging from “Strongly Agree” to “Strongly Disagree”. We code a dummy equal one for

“Strongly Agree” ( fraction reported within parenthesis) to the following statements:

• Pecuniary (14%)

– “Environmentally sustainable investments generate higher returns in the long

run”

11

ii


ii

ii

• Non-Pecuniary (25%)

– “A clean planet is more important for me than economic welfare”

As within different forms of weather calamities, there is considerable overlap between

the two measures of pecuniary motivations. Almost two-thirds of those thinking that

green investments generate higher returns also thinks that a clean planet is more impor-

tant than economic welfare. Our measure of calamities also overlaps the two other moti-

vations, where about one-third prefers a clean planet over economic welfare and one-fifth

believes in higher returns from green investments.

Table III reports a set of Probit regressions to show how the motivations are related

to demographics and geography. In Columns (1) through (4) use the Calamity dummy

as the dependent variable. Column (1) shows that fears of environmental fears are more

common among the young, among lower-income individuals, and among females. En-

vironmental fears are less common among those who have studied economics and busi-

ness. Also, individuals with higher environmental and financial literacy scores think that

future calamities are very likely.

Table III here

In Column (2) we add the beliefs about the test scores which we solicit immediately

after taking the test. When the beliefs are added, the test scores themselves become in-

significant. This is a result that echoes that of Anderson, Baker, and Robinson (2017), who

find that someone’s beliefs about their knowledge of the domain in question (overestima-

tion) is more important for predicting their behavior and beliefs than what they actually

know. Column (2) shows that self-perceptions of environmental and financial literacy are

related to belief in future calamities.

We next test how weather events may shape climate fears. We use the geographical

variation where respondents live in the cross-section and local weather warnings data to

which they were more likely to be exposed in 2014. It is of course likely that individuals

also react to weather shocks outside of their geographical proximity, either through media

exposure or because personal connections to other parts of the country make weather

events there salient. This would, all else equal, make our results weaker. We separately

include milder Class 1 warnings in Column (3) and more severe Class 2 warnings in

12

ii


ii

ii

Column (4).8 We find that people living in areas of a greater number of Class 2 warnings

are much more likely to hold fears of climate calamities, whereas this is not the case for

the milder Class 1 warnings.

In order to rule out that the weather warnings do not coincide with other motives, we

separately run regressions with Pecuniary and Non-Pecuniary motives for green invest-

ments as the dependent variable in Column (5) and (6). Although these motivations share

many of the characteristics, as for instance being more popular among lower-income,

young females, but having been exposed to weather warnings does not explain holding

these beliefs.

The fact that exposure to a greater number of weather anomalies predicts holding

the belief that future weather calamities are extremely likely, but has no effect on the

other motivations, is important for two reasons. First, it provides support for the key

assumption behind our identification strategy–that the extreme weather of 2014 was a

shock that changed opinions. Second, it provides evidence that these changes in opinion

work through a recency or familiarity bias.

4 Actual Fund Choices

To investigate the propensity to choose ESG funds, we use Probit regressions where the

dependent variable takes the value of one if the holdings or trades correspond to one of

the three ESG specifications—Some, Most, or All—defined in Panel B and C of Table I.

We first present the results for the year 2017 and then contrast panel estimates from 2012

to 2014 with those from 2015 to 2017.

4.1 Fund Holdings and Trades in 2017

In Table IV we present results from Probit regressions on holdings of ESG for all individ-

uals that made a choice (who were not in the default fund at the end of 2017). In Column

(1) through (5), the dependent variable labelled “All” takes the value of one if the fund

portfolio contained only ESG funds. The regression shows that owning ESG funds are

more likely among people living in areas of higher Green party votes, but insignificant

8There are very few Class 3 warnings in the data and none in 2014.

13

ii


ii

ii

for all other independent variables we consider. The specification in Columns (1) through

and (2) is made with our without controlling for the type of fund. Fund controls are the

portfolio weights to four fund type categories: Equity, Mixed, Bond and Target funds.

The climate Calamities dummy is introduced in Columns (3) and is strongly significant,

and remains so when introducing separate dummies for pecuniary (higher green returns)

or non-pecuniary (cleaner planet) motives in Column (4) and (5). The point estimates re-

veal that those strongly agreeing to quick global warming, increased food shortage and

sharp sea level rise are around 5% more likely to hold an all ESG portfolio at the end of

2017.

Table IV here

In Column (6) of Table IV, we re-define the dependent variable with the condition to

have at least 50% ESG labelled funds in the portfolio at the end of 2017. The results are

broadly the same as having a 100% weight, if not somewhat stronger. Finally, when we

use the lowest threshold on defining ESG which is a non-zero weight marked “Some” in

Column (7), the correlation with climate Calamaties are considerably weaker and cut in

half. The overall results therefore point to that weather Calamities is associated with a

conscious choice of a portfolio tilt towards ESG funds.

One possible explanation for the preceding results is that the tilt towards green funds

arises because funds reclassify themselves, not because investors necessarily make active

choices. Given the fact that the number of ESG-labelled funds has grown rapidly since

2014, this possibility bears careful consideration.

To control for this possible explanation, in Table V we use trades (or rebalances) of

portfolios into ESG labelled funds as the dependent variable. (Specifically, the trades

summarized in Panel C of Table I.)9 In Column (1) and (2) we first model the decision to

trade by creating a dummy equal one if the individual traded during the year, and zero

otherwise. The results show that the average trader is an older male with higher financial

literacy. When introducing the pecuniary motives and calamities in Column (2), we find

that they are less likely to hold non-pecuniary beliefs.

Table V here9When creating dummy variables for the corresponding trade categories, we first average the ESG hold-

ings across trades in cases there were more than one trade.

14

ii


ii

ii

In Column (3) of Table V, we only consider the 382 trades made during 2017 and again

find that the climate calamities measure is highly significant. The point estimate implies

that respondents having climate fears are 15.6% more likely to trade into an all ESG fund.

In Column (4), we present the Probit regression for the 50%, and find a weaker effect. In

Column (5), using the non-zero threshold as the dependent variable, we can not establish

a relation between climate calamaties and porfolio trades at all.

To summarize, in this subsection we establish a link between fears of climate calami-

ties and the propensity to both hold and trade ESG funds in the year 2017. In line with the

hypothesis that people indeed act on their beliefs, we find that this effect is much stronger

for portfolios that are concentrated to ESG holdings. We find weak or no evidence for

alternative mechanisms driving this behavior, such as for pecuniary or non-pecuniary

reasons.

4.2 Fund Holdings and Trades Before and After the 2014 Heat Wave

In the next step, we investigate to which extent the survey responses can explain recent

years portfolios and trades compared to those preceding the climate shock in 2014. We

divide the time-series of portfolio holdings into two periods: those occuring between

2015 and the end of sample in 2017, and those in the preceding three years 2012 to 2014.

We control for year fixed effects and cluster standard errors on the individual level in all

specifications.

Table VI display the results on holdings. Columns (1) to (2) shows separate regres-

sions for the time periods before and after the weather event using the widest definition

of ESG fund holdings, “Some” where all portfolios with a non-zero weight is included

in the specification. We find no evidence of a relation between weather calamities and

ESG investors. In Column (3) and (4) we repeat this analysis and change the dependent

variable to be “Most” ESG holdings (above 50%), and in Column (5) to (6) “All” (100%).

The results show that portfolio holdings are not affected by fears of weather calamities in

the years before 2014, but is strongly related to those after the weather event.

Table VI here

In Table VII, we repeat the analysis for trades and introduce the difference in fees in the

15

ii


ii

ii

portfolio that was traded to compared to the previously held portfolio as a control vari-

able. This is to rule out the possibility that our results would be driven by a motivation

to minimize fees. The results are similar, but stronger compared to those in Table VI.

We find no relation between climate fears and trades for low intensity ESG trades, but

a very strong effect for high intenstiy rebalances. The marginal probability estimate for

our Calamity dummy reveals that there is a 10% increased chance that those with climate

fears traded into an ESG portfolio in period after 2014, but close to zero before.

Table VII here

5 Discussion and Conclusion

This paper connects extreme weather events to the financial decisions made by the people

who are exposed to them. It not only provides evidence that extreme weather calamities

affect both the supply and the demand for environmentally responsible financial invest-

ments, but it explores a variety of economic mechanisms by which this might occur.

The increased salience of environmental issues induced by being exposed to extreme

weather events could operate through several channels. One channel is that individu-

als might believe, rightly or wrongly, that in the future the returns to environmentally

sustainable investments will be higher than previously expected. Altering the perceived

risk/return tradeoff of so-called green investments might induce them to increase their

holdings for purely pecuniary reasons, regardless of any inherent concern for the envi-

ronment or their fellow man. Our results provide little support for this explanation.

Alternatively, scenes of environmental devastation might cause a moral awakening,

whereby investing sustainably becomes a moral imperative. Raising social consciousness

could cause individuals to increase their allocations to green investments regardless of

whether the expected financial returns are higher or lower than previously thought. Our

results also provide little support for this explanation.

Instead, our results suggest that some individuals, after being exposed to extreme

weather events, overweight the probability of extremely unlikely future weather events

will occur. It is these individuals, and not others, who are more likely to tilt their re-

tirement portfolios towards green investments. Thus, elements of prospect theory seem

16

ii


ii

ii

important for understanding the demand for green investments. This suggests that be-

havioral channels are likely an important mechanism for understanding how exposure to

extreme events affects portfolio choice.

These findings are important for several reasons. First, as climate change induces

increasing weather volatility, the exposure to extreme weather events is likely to increase

in the future. Thus the scope for overweighting to play into decision-making is likely

to be important going forward. More generally, households and institutional investors

alike are increasingly being asked to make investment choices based not only on standard

pecuniary risk and return tradeoffs but also on the environmental or social performance

of companies and mutual funds. The difficulty in measuring these alternative dimensions

leaves them even more susceptible to the types of behavioral biases that have been shown

to affect financial decision-making in other contexts.

17

ii


ii

ii

References

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 paperNo. 17-15.

, 2019, Knowledge, fear and beliefs: Understanding household demand for greeninvestments, Swedish House of Finance working paper.

Andersson, Mats, Patrick Bolton, and Frederic Samama, 2016, Hedging climate risk, Fi-nancial Analyst Journal 72, 13–32.

Choi, Darwin, Zheyu Gao, and Wenxi Jiang, 2018, Attention to global warming, Workingpaper, SUHK Business School, The Chinese University of Hong Kong.

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.

, and Frank Yu, 2018, When nudges are forever: Inertia in the swedish premiumpension plan, Working paper, University of Chicago.

FAO, IFAD, and WFP, 2015, The state of food insecurity in the world 2015. Meeting the2015 international hunger targets: taking stock of uneven progress., Rome, FAO.

Hartzmark, Samuel M., and Abigail B. Sussman, 2018, Do investors value sustainability?a natural experiment examining ranking and fund flows, Working Paper, University ofChicago.

IPCC, 2014, Climate change 2014: Synthesis report. Contribution of working groups I,II and III to the Fifth Assessment Report of the Intergovernmental Panel on ClimateChange [Core writing team, R.K. Pachauri and L.A. Meyer (eds.)]., IPCC, Geneva,Switzerland, 151 pp.

Lusardi, Annamaria, and Olivia S. Mitchell, 2007, Baby boomer retirement security: Theroles of planning, financial literacy, and housing wealth, Journal of Monetary Economics54, 205–224.

, 2011, Financial literacy and planning: Implications for retirement well-being,in Annamaria Lusardi, and Olivia S. Mitchell, ed.: Financial Literacy: Implications forRetirement Security and the Financial Marketplace . pp. 17–39 (Oxford University Press)Michigan Retirement Research Center, WP 2015-108.

Palme, Marten, Annika Sunden, and Paul Soderlind, 2007, How do individual accountswork in the swedish pension system?, Journal of the European Economic Association 5,636–646.

18

ii


ii

ii

Tversky, Amos, and Daniel Kahneman, 1992, Advances in prospect theory: Cumulativerepresentation of uncertainty, Journal of Risk and Uncertainty 5, 297–323.

19

ii


ii

ii

Table I: Fund Menu, Portfolio Holidings and Rebalances of ESG Funds

This table presents the number of individuals in sample making changes (portfolio rebalances) within the Swedish PensionSystem in 2017. The first row displays the number of individuals in each bucket of number of changes ranging from 0 to 5 and over5 trades, and shows that 427 individuals of 3,667 made at least one trade. The second row shows the same distribution conditionalon that the portfolio change contained at least partially (non-zero weight) an ESG labelled fund. The third row dispalys the samedistribution with the condition that the portfolio change was made to an all ESG labelled portfolio. Data of portfolio changes andfund holdings are obtained from the SPA.

Panel A: Number of funds / YearType of funds 2010 2011 2012 2013 2014 2015 2016 2017All available SPA funds 839 873 854 885 886 881 873 892of which ESG 89 99 99 118 146 187 270 325Fraction ESG 10.6% 11.3% 11.6% 13.3% 16.5% 21.2% 30.9% 36.4%

Panel B: Number of funds in portfolio 2017Holdings (0) (1) (2) (3) (4) (5) ESG>0 TotalDefault fund 1,193 1,193Active 865 453 316 307 533 1,827 2,474Some ESG funds 647 389 343 288 285 522 1,827 2,474Most ESG funds 1,040 389 292 208 193 352 1,434 2,474All ESG funds 1,634 389 240 102 57 52 840 2,474

Panel C: Number of portfolio changes 2017Changes (0) (1) (2) (3) (4) (5) (>5) TotalFull sample 3,285 264 50 21 11 5 31 3,667Changed to some ESG funds 65 231 35 14 7 5 25 382Changed to mostly ESG funds 168 166 27 7 3 3 8 382Changed to all ESG funds 260 109 8 1 2 1 1 382

20

ii


ii

ii

Table II: Sample Characteristics

This table reports sample proportions and summary statistics across some key characteristics of the responses of the surveyquestions about environmental beliefs as well as ESG pension fund holdings. Column (1) shows the sample proportions and Column(2) the corresponding population average for Sweden. Columns (3) through (5) present the proportions of respondents stronglyagreeing to the following three groups of statements: Calamities takes the value of one if agreeing to: “How likely do you thefollowing globals scenarios are in the next twenty year”, followed by the following three statements “The average temperature onearth rises by more than one degree Celsius”, “Shortage of food will increase” and “The seawater level will increase by over onemeter” and zero otherwise. Pecuniary takes the value of one if agreeing to the following statement: “In the long run, environmentallysustainable investments generate higher returns” and zero otherwise. Non-Pecuniary takes the value of one if strongly agreeing to:“A clean planet is more important for me than economic welfare”. Column (6) shows the proportion of individuals in the pensiondefault fund. Column (7) shows the fraction of respondents not in the default fund, holding a non-zero weight in an ESG labelledfund as of the end of 2017. Columns (8) and (9) dispaly the same fraction, but defining the holdings to be over 50% (Most) or 100%(All) of their portfolio in an ESG fund. There are 3,667 individuals in the full sample (first six columns), of which 2,474 actively chosetheir portfolio (the last three columns).

Sample Pop. Green Beliefs Def. ESG holdingsProp. Prop. Calamities Pecuniary Non-Pec. Fund Some Most All

Overall 100.00 100.00 0.47 0.13 0.26 0.33 0.74 0.58 0.34Pop. Wtd. . . 0.49 0.15 0.25 0.41 0.72 0.57 0.34

GenderMen 48.70 51.10 0.45 0.12 0.23 0.30 0.75 0.56 0.32Women 51.30 48.90 0.49 0.15 0.29 0.35 0.73 0.60 0.36

Age18-24 4.23 15.50 0.61 0.11 0.23 0.98 1.00 1.00 1.0025-34 15.19 22.90 0.60 0.16 0.26 0.74 0.57 0.48 0.3035-44 19.55 20.80 0.51 0.16 0.30 0.39 0.74 0.56 0.3345-54 27.02 22.00 0.42 0.12 0.24 0.17 0.77 0.59 0.3255-65 34.01 18.90 0.42 0.12 0.25 0.15 0.73 0.59 0.36

Income0-111 8.45 25.00 0.56 0.16 0.33 0.73 0.71 0.54 0.25111-287 34.14 24.90 0.48 0.15 0.27 0.40 0.70 0.57 0.38287-399 31.20 25.20 0.45 0.13 0.24 0.25 0.74 0.57 0.34399+ 25.63 25.00 0.45 0.10 0.24 0.17 0.78 0.61 0.32

EducationSome school 5.40 17.40 0.49 0.16 0.19 0.32 0.69 0.57 0.34High school 39.19 44.00 0.45 0.12 0.21 0.29 0.70 0.55 0.34College 55.03 38.60 0.48 0.14 0.30 0.35 0.77 0.60 0.34Studied Env/Bio 1.88 . 0.59 0.19 0.35 0.42 0.75 0.60 0.38Studied Econ/Bus 10.31 . 0.41 0.14 0.22 0.25 0.71 0.54 0.32

LocationUrban 34.14 . 0.51 0.14 0.29 0.39 0.78 0.60 0.35Rural 65.86 . 0.45 0.13 0.24 0.29 0.72 0.57 0.34

21

ii


ii

ii

Table III: Motives to be Green: Calamaties, Pecuniary and Non-Pecuniary

This table reports the results of Probit regressions where the dependent variable takes the value of one if the response is“Strongly agree” to one or several statements. Calamities takes the value of one if agreeing to: “How likely do you the followingglobals scenarios are in the next twenty year”, followed by the following three statements “The average temperature on earth risesby more than one degree Celsius”, “Shortage of food will increase” and “The seawater level will increase by over one meter” andzero otherwise. Pecuniary takes the value of one if agreeing to the following statement: “In the long run, environmentally sustainableinvestments generate higher returns” and zero otherwise. Non-Pecuniary takes the value of one if strongly agreeing to: “A cleanplanet is more important for me than economic welfare”. Dependent variables Warnings Class 1 and 2 counts the number of weatherwarnings in each category in 2014, Perceived and Actual financial and environmental literacy denote self-assessed and actual scoreon a five question test. Log income refers to disposal income, Age is divided by ten and Female denotes a dummy equal to one forwomen, zero otherwise. Urban and Green denote population density and the share of green party voters in the municipality of therespondent. University, ECON and ECO student are education indicator variables for subjects having a university degree or havingstudied Economics/Business or Biology/Geography/Environmental science at any level since high school. Sampling weights areused.

(1) (2) (3) (4) (5) (6)VARIABLES Calamities Calamities Calamities Calamities Pecuniary Non-Pec.

Warnings Class 2 0.015*** 0.003 0.008(0.006) (0.004) (0.005)

Warnings Class 1 -0.000(0.001)

Perceived Env. Lit. 0.043*** 0.043*** 0.044*** 0.029*** 0.029***(0.011) (0.011) (0.011) (0.008) (0.010)

Perceived Fin. Lit. 0.001 0.001 0.001 0.006 -0.007(0.011) (0.011) (0.011) (0.008) (0.009)

Env. Lit. 0.022** 0.009 0.009 0.008 0.013* 0.033***(0.010) (0.010) (0.010) (0.010) (0.007) (0.009)

Fin. Lit. 0.024*** 0.015 0.015 0.015 -0.012 0.002(0.009) (0.010) (0.010) (0.010) (0.008) (0.009)

Log Income -0.037*** -0.037*** -0.037*** -0.036*** -0.010* -0.035***(0.009) (0.009) (0.009) (0.009) (0.006) (0.008)

Age -0.048*** -0.055*** -0.055*** -0.056*** -0.011* 0.006(0.008) (0.008) (0.008) (0.008) (0.006) (0.007)

Female 0.053*** 0.067*** 0.067*** 0.068*** 0.035** 0.069***(0.020) (0.020) (0.020) (0.020) (0.014) (0.017)

Urban 0.008 0.008 0.008 0.003 0.005 0.006(0.008) (0.008) (0.008) (0.008) (0.006) (0.007)

Green 0.001 0.001 0.001 0.003 0.002 0.008***(0.004) (0.004) (0.004) (0.004) (0.003) (0.003)

University -0.023 -0.027 -0.027 -0.032 0.005 0.040(0.069) (0.070) (0.070) (0.069) (0.046) (0.063)

ECO student 0.069 0.050 0.050 0.044 0.046 0.045(0.074) (0.073) (0.073) (0.073) (0.057) (0.063)

ECON student -0.071** -0.063** -0.063** -0.062* 0.021 -0.015(0.031) (0.031) (0.031) (0.031) (0.024) (0.027)

Observations 3,667 3,667 3,667 3,667 3,667 3,667Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

22

ii


ii

ii

Table IV: Holding Green Mutual Funds in 2017

This table reports the results of Probit regressions where the dependent variable takes the value of one if the respondent hasmore than 50% of their government premium pension savings invested in ESG labelled funds; zero otherwise. The independentvariables follow those from Table III. The sample represents the 2,277 respondents who were not in the default fund at the endof 2017. Independent variables follow those of Table III. Data of portfolio changes and fund holdings are obtained from the SPA.Sampling weights are used.

(1) (2) (3) (4) (5) (6) (7)VARIABLES All All All All All Most Some

Calamities 0.046** 0.043* 0.045* 0.042* 0.025(0.023) (0.023) (0.023) (0.024) (0.022)

Pecuniary 0.030 0.038 0.022 0.011(0.036) (0.038) (0.039) (0.035)

Non-Pecuniary -0.019 0.018 0.053**(0.028) (0.030) (0.026)

Perceived Env. Lit. -0.001 -0.001 -0.003 -0.004 -0.003 -0.003 0.002(0.013) (0.013) (0.013) (0.013) (0.013) (0.014) (0.013)

Perceived Fin. Lit. -0.016 -0.017 -0.017 -0.018 -0.018 -0.008 -0.014(0.013) (0.013) (0.013) (0.013) (0.013) (0.014) (0.013)

Env. Lit. -0.005 -0.003 -0.003 -0.003 -0.003 0.001 -0.005(0.012) (0.012) (0.012) (0.012) (0.012) (0.013) (0.012)

Fin. Lit. -0.004 -0.003 -0.004 -0.004 -0.004 -0.012 -0.001(0.012) (0.012) (0.012) (0.012) (0.012) (0.013) (0.012)

Log Income 0.013 0.012 0.013 0.013 0.013 0.017 0.017(0.014) (0.014) (0.014) (0.014) (0.014) (0.015) (0.013)

Age 0.014 0.025** 0.027** 0.027** 0.028** 0.036*** 0.041***(0.012) (0.012) (0.012) (0.012) (0.012) (0.013) (0.012)

Female 0.012 0.001 -0.001 -0.002 -0.001 0.007 -0.038(0.024) (0.024) (0.024) (0.024) (0.024) (0.026) (0.023)

Urban -0.004 -0.001 -0.001 -0.001 -0.001 0.015 0.024***(0.009) (0.009) (0.009) (0.009) (0.009) (0.010) (0.009)

Green 0.009** 0.009** 0.009** 0.009** 0.009** 0.004 0.002(0.004) (0.004) (0.004) (0.004) (0.004) (0.005) (0.004)

University 0.016 0.010 0.009 0.007 0.008 0.086 0.002(0.087) (0.085) (0.085) (0.085) (0.086) (0.087) (0.088)

ECO student 0.093 0.083 0.083 0.084 0.086 0.012 0.078(0.094) (0.093) (0.093) (0.093) (0.094) (0.098) (0.065)

ECON student -0.032 -0.033 -0.031 -0.033 -0.033 -0.040 -0.026(0.034) (0.034) (0.034) (0.034) (0.034) (0.038) (0.035)

Observations 2,474 2,474 2,474 2,474 2,474 2,474 2,474Fund controls No Yes Yes Yes Yes Yes Yes

Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

23

ii


ii

ii

Table V: Choosing Green Mutual Funds in 2017

This table reports the results of Probit regressions where the dependent variable takes the value of one if the respondentshas rebalanced their portfolio in 2017. In Column (1) and (2) the dependent variable takes the value of one for the 382 respondentsthat made at least one rebalancing of their portfolio, zero otherwise. Columns (3) to (5) displays displays the same regression focusingon the 382 that made at least one trade. In Column (3) the dependent variable takes the value of one if the respondent rebalancedtheir porfolio to an all ESG portfolio. In Column (4), the indicator variable takes the value of one if the rebalancing into a portfolioof at least 50% ESG labelled funds, and Column (5) if greater than zero; the indicator variable is zero otherwise. and (5) repeats theanalyis of Column (4) but uses more than 50% of their government premium pension savings invested in ESG labelled funds; zerootherwise. The regressions include several sets of controls. Individual Characeteristics an Knowledge controls follow those fromTable III and IV but educational dummies are dropped due to collinearity in Columns (3) through (5). We include year fixed effectsand portfolio weights in four fund categories: Equities, Mixed, Bond and Target funds. Data of portfolio changes and fund holdingsare obtained from the Swedish Pension Authority. Sampling weights are used.

(1) (2) (3) (4) (5)VARIABLES All Trades All Trades All ESG Mostly ESG Some ESG

Calamities 0.005 0.132** 0.113* -0.021(0.010) (0.058) (0.060) (0.044)

Pecuniary -0.000 0.026 0.111 0.028(0.015) (0.091) (0.089) (0.053)

Non-Pecuniary -0.023** -0.097 -0.097 -0.045(0.012) (0.069) (0.082) (0.061)

Perceived Env. Lit. -0.000 0.000 0.023 0.006 -0.013(0.006) (0.006) (0.036) (0.035) (0.027)

Perceived Fin. Lit. 0.005 0.005 -0.041 0.004 0.013(0.006) (0.005) (0.037) (0.037) (0.026)

Env. Lit. -0.006 -0.005 0.002 0.003 -0.007(0.005) (0.005) (0.031) (0.033) (0.023)

Fin. Lit. 0.014*** 0.014*** -0.022 0.017 0.019(0.005) (0.005) (0.032) (0.032) (0.021)

Log Income 0.013** 0.012** -0.093* -0.136** -0.093**(0.005) (0.005) (0.053) (0.059) (0.042)

Age 0.024*** 0.024*** 0.008 -0.007 0.045**(0.004) (0.004) (0.028) (0.029) (0.022)

Female -0.028*** -0.026** 0.027 -0.026 -0.082*(0.010) (0.010) (0.062) (0.065) (0.046)

Urban -0.001 -0.001 -0.034 0.021 0.022(0.004) (0.004) (0.024) (0.023) (0.015)

Green -0.001 -0.001 0.016 0.001 0.005(0.002) (0.002) (0.011) (0.011) (0.009)

University -0.044* -0.044*(0.024) (0.024)

ECO student -0.012 -0.011(0.043) (0.043)

ECON student -0.010 -0.009(0.015) (0.015)

Observations 3,667 3,667 382 382 382Fund controls Yes Yes Yes Yes Yes

Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

24

ii


ii

ii

Tabl

eV

I:H

oldi

ngs

Bef

ore

and

Aft

erth

e20

14H

eatW

ave

This

tabl

ere

port

sth

ere

sult

sof

Prob

itre

gres

sion

sw

here

the

depe

nden

tva

riab

leta

kes

the

valu

eof

one

ifth

ere

spon

dent

sha

sa

prop

orti

onof

thei

rpo

rtfo

lioin

vest

edin

ESG

fund

sin

the

year

sbe

fore

oraf

ter

the

heat

wav

ein

2014

.Th

esa

mpl

e“B

efor

e”co

ntai

nsal

lye

ar-e

ndpo

rtfo

lios

duri

ngth

eye

ars

2012

to20

14,a

ndth

esa

mpl

e”A

fter

”fo

r20

15to

2017

.T

hede

pend

ent

vari

able

inC

olum

n(1

)and

(2)t

akes

the

valu

eof

one

for

the

resp

onde

nts

held

apo

rtfo

liow

ith

apo

siti

veES

Gw

eigh

t,ze

root

herw

ise.

InC

olum

n(3

)and

(4),

the

indi

cato

rva

riab

leta

kes

the

valu

eof

one

ifth

epo

rtfo

lioco

ntai

nsat

leas

t50%

ESG

labe

lled

fund

s,an

dC

olum

ns(5

)and

(6)i

fthe

hold

ings

wer

eal

l(10

0%)i

nES

Gfu

nds;

the

indi

cato

rva

riab

leis

zero

othe

rwis

e.C

ontr

olva

riab

les

follo

wth

ose

inTa

ble

III.

Year

fixed

effe

cts

and

fund

-typ

eco

ntro

lsar

ein

clud

edin

all

spec

ifica

tion

s.D

ata

ofpo

rtfo

liofu

ndho

ldin

gsan

dcl

assi

ficat

ions

are

obta

ined

from

the

SPA

.Sam

plin

gw

eigh

tsar

eus

ed.

(1)

(2)

(3)

(4)

(5)

(6)

VAR

IABL

ESSo

me

Befo

reH

WSo

me

Aft

erH

WM

ostB

efor

eH

WM

ostA

fter

HW

All

Befo

reH

WA

llA

fter

HW

Cal

amit

ies

0.00

50.

013

0.01

00.

045*

*-0

.003

0.04

0*(0

.022

)(0

.022

)(0

.014

)(0

.022

)(0

.009

)(0

.021

)Pe

cuni

ary

-0.0

25-0

.002

0.01

00.

022

0.02

60.

044

(0.0

35)

(0.0

34)

(0.0

24)

(0.0

36)

(0.0

20)

(0.0

34)

Non

-Pec

unia

ry0.

050*

0.05

3**

0.03

7**

0.02

20.

001

-0.0

10(0

.027

)(0

.025

)(0

.019

)(0

.027

)(0

.011

)(0

.025

)

Obs

erva

tion

s7,

213

7,39

07,

213

7,39

07,

213

7,39

0C

hara

cter

isti

csYe

sYe

sYe

sYe

sYe

sYe

sK

now

ledg

eco

ntro

lsYe

sYe

sYe

sYe

sYe

sYe

sFu

ndco

ntro

lsYe

sYe

sYe

sYe

sYe

sYe

sYe

arFE

Yes

Yes

Yes

Yes

Yes

Yes

Stan

dard

erro

rsin

pare

nthe

ses

***

p<0.

01,*

*p<

0.05

,*p<

0.1

25

ii


ii

ii

Tabl

eV

II:T

rade

sB

efor

ean

dA

fter

the

2014

Hea

tWav

e

This

tabl

ere

port

sth

ere

sult

sof

Prob

itre

gres

sion

sw

here

the

depe

nden

tva

riab

leta

kes

the

valu

eof

one

ifth

ere

spon

dent

sha

sre

bala

nced

thei

rpo

rtfo

lioin

the

year

sbe

fore

oraf

ter

the

heat

wav

ein

2014

.The

sam

ple

“Bef

ore”

cont

ains

allt

rade

sdu

ring

the

year

s20

12to

2014

,and

the

sam

ple

”Aft

er”

for

2015

to20

17.T

hede

pend

entv

aria

ble

inC

olum

n(1

)and

(2)t

akes

the

valu

eof

one

for

the

resp

onde

nts

that

mad

eat

leas

ton

etr

ade

into

apo

rtfo

liow

ith

apo

siti

veES

Gw

eigh

t,ze

root

herw

ise.

InC

olum

n(3

)an

d(4

),th

ein

dica

tor

vari

able

take

sth

eva

lue

ofon

eif

the

reba

lanc

ing

into

apo

rtfo

lioof

atle

ast5

0%ES

Gla

belle

dfu

nds,

and

Col

umns

(5)a

nd(6

)ift

hetr

ade

was

into

anal

l(10

0%)E

SGpo

rtfo

lio;t

hein

dica

tor

vari

able

isze

root

herw

ise.

Con

trol

vari

able

sfo

llow

thos

ein

Tabl

eII

I.Ye

arfix

edef

fect

san

dfu

nd-t

ype

cont

rols

are

incl

uded

inal

lspe

cific

atio

ns.D

ata

ofpo

rtfo

lioch

ange

san

dcl

assi

ficat

ions

are

obta

ined

from

the

SPA

.Sam

plin

gw

eigh

tsar

eus

ed.

(1)

(2)

(3)

(4)

(5)

(6)

VAR

IABL

ESSo

me

Befo

reH

WSo

me

Aft

erH

WM

ostB

efor

eH

WM

ostA

fter

HW

All

Befo

reH

WA

llA

fter

HW

Cal

amit

ies

-0.0

49**

0.02

9-0

.012

0.10

6**

-0.0

14**

*0.

083*

*(0

.023

)(0

.038

)(0

.009

)(0

.044

)(0

.005

)(0

.037

)Pe

cuni

ary

-0.0

500.

014

-0.0

030.

108

0.00

30.

068

(0.0

31)

(0.0

59)

(0.0

15)

(0.0

75)

(0.0

09)

(0.0

62)

Non

-Pec

unia

ry0.

085*

*0.

026

0.00

40.

001

-0.0

02-0

.068

*(0

.036

)(0

.047

)(0

.012

)(0

.055

)(0

.005

)(0

.040

)Fe

edi

ff.

0.07

1-0

.071

-0.0

22-0

.210

*0.

001

-0.1

34(0

.054

)(0

.105

)(0

.024

)(0

.116

)(0

.009

)(0

.097

)

Obs

erva

tion

s1,

419

1,04

51,

419

1,04

51,

419

1,04

5C

hara

cter

isti

csYe

sYe

sYe

sYe

sYe

sYe

sK

now

ledg

eco

ntro

lsYe

sYe

sYe

sYe

sYe

sYe

sFu

ndco

ntro

lsYe

sYe

sYe

sYe

sYe

sYe

sYe

arFE

Yes

Yes

Yes

Yes

Yes

Yes

Stan

dard

erro

rsin

pare

nthe

ses

***

p<0.

01,*

*p<

0.05

,*p<

0.1

26

ii


ii

ii

Figure 1: July temperatures in Sweden 2012-2017

This figure display heat maps of differences from average temperatures over Sweden in July for the years 2012 to 2017. Yellow denotes

normal, blue to purple below, and orange to red above average temperatures.The maps are obtained from the Swedish Meteorological

and Hydrological Institute, downloaded from www.smhi.se.

2012 2013 2014

2015 2016 2017

27

ii


ii

ii

Figure 2: August rainfall in Sweden in 2012-2017

This figure display heat maps of rainfall in percentages of normal over Sweden in August for the years 2012 to 2017. Yellow denotes

average (100%), green to blue above, and orange to red below normal rainfall. The maps are obtained from the Swedish Meteorological

and Hydrological Institute downloaded from www.smhi.se.

2012 2013 2014

2015 2016 2017

28

ii


ii

ii

Figu

re3:

The

Emer

genc

eof

ESG

Fund

sO

ver

Tim

e

This

figur

esh

ows

how

ESG

mut

ualf

unds

have

prol

ifer

ated

over

tim

ein

the

Swed

ish

Prem

ium

Pens

ion

Syst

emas

per

Dec

embe

rea

chye

aras

pres

ente

din

the

PPA

fond

broc

hure

.The

bars

show

tota

lnum

ber

offu

nds,

split

upin

perc

ento

fESG

and

non-

ESG

fund

sin

the

offe

ring

.The

ESG

labe

lwas

intr

oduc

edin

2004

.The

grey

line

depi

cts

the

num

ber

ofne

ws

item

sfr

omth

efo

urla

rges

t

new

spap

ers

inSw

eden

(Aft

onbl

adet

,Dag

ens

Nyh

eter

,Exp

ress

enan

dSv

ensk

aD

agbl

adet

)eve

ryye

arco

ntai

ning

the

wor

d“C

limat

ech

ange

”.Th

eda

tafo

rfu

nds

are

colle

cted

from

the

SPA

fund

cata

logu

esan

dw

ebpa

ge.T

heda

taon

new

sco

vera

geis

from

the

Nat

iona

lLib

rary

ofSw

eden

(ww

w.k

b.se

).

100%

100%

100%

100%

100%

93%

93%

94%

92%

91%

91%

89%

89%

88%

87%

84%

79%

69%

64%

7%7%

6%8%

9%9%

11%

11%

12%

13%

16%

21%

31%

36%

0102030405060708090100

0

100

200

300

400

500

600

700

800

900

1000

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

No. of temperature records

No. of funds

Year

Non-

ESG

labe

lled

fund

sES

G la

belle

d fu

nds

Artic

les o

n "C

limat

e Ch

ange

"Hi

gh te

mpe

ratu

re re

cord

s

29

ii


ii

ii

Figu

re4:

Tota

lPen

sion

Savi

ngs

and

ESG

Hol

diin

gs

This

figur

esh

ows

the

sam

ple

hold

ings

fund

sdi

vide

din

toES

Gan

dno

n-ES

Gan

dth

ede

faul

tfun

d(a

lso

anES

Gla

belle

dfu

nd).

The

hold

ings

are

split

upin

the

amou

nts

trad

edin

toea

chca

tego

ry

ever

yye

ar.

2029

3942

3143

5554

6898

137

157

197

252

4871

9598

8011

212

011

417

5

216

277

256

228

236

719

3253

33

6281

79

58

67

5944

38

44

69

1113

10

1426

27

34

39

6413

121

6

247

11

35

4

3

45

3

12

14

31

38

58

0

100

200

300

400

500

600

700

800

900

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

SEKm

Defa

ult

Hold

oth

er fu

nds

Trad

ed in

to o

ther

fund

s

Hold

ESG

Trad

ed in

to E

SG

30

ii


ii

ii

Figu

re5:

Surv

eyR

espo

nses

toC

limat

eC

alam

itie

s

This

figur

esh

ows

the

surv

eyre

spon

ses

toth

ree

ques

tion

sab

out

clim

ate

cala

mit

ies

ona

five

poin

tLi

kert

scal

era

ning

from

Very

Uni

kely

toVe

ryLi

kely

.Th

equ

esti

onis

“In

the

next

20ye

ars”

follo

wed

byth

ree

stat

emen

ts:

i)“T

heav

erag

ete

mpe

ratu

reon

eart

hin

crea

ses

bym

ore

than

one

Cen

tigr

ade”

;ii)

“Foo

dsh

rtag

ein

crea

ses”

;and

iii)“

The

wor

ldse

ale

velr

ises

than

mor

eth

anon

e

met

er”. 0%10

%

20%

30%

40%

50%

Don'

t kno

wVe

ry u

nlik

ely

Uni

kely

Nei

ther

Lik

ely

orU

nlik

ely

Like

lyVe

ry li

kely

Clim

ate

cala

miti

es: "

In th

e ne

xt 2

0 ye

ars.

..""T

he a

vera

ge te

mpe

ratu

re o

n ea

rth

incr

ease

s mor

e th

an o

ne C

entig

rade

"

"Foo

d sh

orta

ge in

crea

ses"

"The

wor

ld se

a le

vel r

ises m

ore

than

one

met

er"

31

Talking About the Weather: Understanding Household Demand ...€¦ · “green˙investment˙v5”...

Documents

Transcript of Talking About the Weather: Understanding Household Demand ...€¦ · “green˙investment˙v5”...