Detrimental Effects of Daylight-Saving Time on SAT Scores

Abstract

Generating a novel scion to several obliquely related literature streams, this study examined a particular high-profile cognitive outcome of a sometimes controversial government policy, daylight-saving time. Controlling for socioeconomic status by proxy, the principal finding was a surprisingly strong negative relationship between imposition of the time policy in a geographic area and SAT scores of local high school students. The cautious conclusion is that the daylight-saving time policy should possibly be even more controversial for, at minimum, its economic implications.

Full text

Detrimental
Effects
of Daylight-Saving
Time
on
SAT Scores
John F.
Gaski
University
of
Notre
Dame
The
question "What
time
is
it?" is
more
charged or loaded
in some parts
of
the Unitec
States than
others. Most
states and locales
have
wrestled,
at one time
or another, with
the
policy
issue
of whether
to adopt
daylight-saving
time
(DST)-that
is,
clocks
advanced an hour
in rhe
spring, then reset
back in
the fall. In
a few
regions,
the
seemingly mundane
dispute
has
become serious
political
and
social conflict.
Be-
cause of the
purported
lifestyle
and
economic
benefits, all
of the United
States except
the
states of Arizona
and Hawaii
now recognize
DST
for about
7.5 months
per year. (Most
of the
world
does, too. in
some form.
Notable holdouts
are China, India,
and Japan-as
of this writing.)
Before
addressing
the details
of this debate,
and
pertinence
to the reported
research,
some
back-
ground
is
provided.
DST History
Benjamin
Franklin
is
considered
to be the
father
of DST
by virtue of his
1784
essay
pub-
This
anicle was
published
Online First
November
8, 2010.
John F.
Gaski, Department
of Marketing,
University
of
Notre Dame;
Jeff Sagarin,
Sagarin Computer
Sports Rat-
ings,
Bloomington,
IN.
Order of authorship is
alphabetical.
Dr. Gaski
wishes to
publicly
recognize
and salute
co-author
Sagarin, the
re-
nowned
sports statistician,
for his
expert role in
the execu-
tion of this
project.
The authors
thank Anne Brinson
of the
Indiana
Department
of Education fbr providing
raw
data.
Correspondence
conceming this
article should
be ad-
dressed
to John F.
Gaski, Mendoza
College of Business,
Notre Dame, IN
46556. E-mail:
jgaski@nd.edu
Journal of Neuroscicnce,
Psychology, and Econonics
2011, Vol.
.1.
No. 1,.1.1 53
O
2010 American Psychologicrl
Association
I
937'12 lX/10/$l 2.00
DOI: 10. l{}37la00201
l8
Generating
a
novel
scion
to several
obliquely related
literature
streams, this
study
examined a
particular
high-profile
cognitive
outcome
of a sometimes
controversial
government
policy,
daylighrsaving
time.
Controlling for
socioeconomic
status
by
proxy,
the
principal
finding
was
a surprisingly
strong negative
relationship
betwcen
imposition
of the timc
policy
in a
geographic
area
and SAT
scores of local high
school
students. The
cautious
conclusion is
that the
daylighrsaving
time
policy
should
possibly
be even more
controversial for,
at minimum,
its
cconomic implications.
Keywords:
cognition,
aptitude, public policy
Jeff Sagarin
Sagarin
Computer Sports
Ratings, Bloomington,
IN
lished
in
a
Paris
newspaper
(Goodman,
193 1).
His original
rationale,
remarkably
similar
to
present
concerns, was
energy
savings
primarily
from
evening lamp-lighting
economies
(a
mil-
lion
francs worth per year
throughout
France,
by his
estimate).
After the British
Parliament
considered
and
rejected
DST
in 1907,
much
of
Europe
and
the
United States
adopted it
during
the First World
War, again
for the fuel
conservation
motive.
After
the
unpopular
practice
was
abolished
by
the U.S.
Congress in
1918, President
Franklin
Roosevelt reinstituted
DST nationally
from
1942
to the
end of World
War II in
1945. From
the war's
end until 1966,
DST
observance in
the
United States
spread
electively
by individual
state and locality
legislation.
Then,
the Uniform
Time
Act
(1966)
mandated
national
DST
be-
tween April
and
October, subject
to individual
state opt-out
and
enforced
by the new federal
Department
of Transportation (DoT).
Twice
since then,
the DST period
has
been
extended
by law
to
its
current late
March-early
Novem-
ber schedule.
The latest
state
government
to
accept DST,
that is,
to
end its opt-out,
was
Indiana
in 2005.
Related
Empirical
Work
Energy
To
what
extent are the
claimed
advantages
of
DST
supported
by evidence?
The
origin of what
may now
be urban legend
about DST
promoting
conservation
of energy
seems
to have been
a
decades-old reporl
by the U.S. DoT
(1975)

DAYLIGHT-SAVING
TIME AND
SAT SCORES
4f
touting
a
l%o decrease
in electricity
demand
attributable
to
DST.
This
putative result
was
repudiated
a
year later,
however,
in a report
to
Congress
by the
U.S.
National
Bureau
of
Stan-
dards
(1976).
(For
an analysis
ofconceptual
and
interpretational
errors
in the
1975
DoT study,
see Gaski
2008.)
More
recently,
Kotchen
and
Grant
(2008),
using
the
shifting
Indiana
DST
landscape
as
a natural
experimental
setting,
and
with more
than
seven
million
monthly
meter
readings
as data
points, found
the
imposition
and
presence of
DST to
cost
the
Hoosier
state
about
$8.6
million
in annual
electricity
waste.
A
factor cited
to explain
their
results
was the
greater
prevalence
of air
conditioning
now compared
with
the
1970s.
ln other
words, more
evening
daylight'
and
ambient
heat
lasting
longer
into
the
night'
induces
longer
running
air conditioners.
Kellogg
and
Wolff
(2007), using
Australian
data, also
refuted
the
l%a saving
claim,
along
with a similar
estimate
of
marginal
efficiency
by
the Califomia
Energy
Commission
(2001).
Safety
Another
contentious
DST-related
implication
has long
been
the
safety
issue,
especially
whether
the shifted
hour of
sunlight-darkness
has an
impact
on the
number
of
motor
vehicle
traflic
accidents
and
fatalities.
Again,
we find
a
conflict
between
the
same
two
vintage
federal
government
studies.
The U.S.
DoT
(1975)
re-
search
asserted
a
slight
(<l%o)
net
decrease
in
fatal
accidents
due
to
DST.
The U.S.
National
Bureau
of
Standards
(1916)
reversed
this
find-
ing,
reanalyzing
the
DoT
data
and
reporting
no
significant
difference
in overall
traffic
fatali-
ties-but
a
statistically
significant
increase
in
fatalities
among
school-age
children
during
the
DST
period, although
the
bureau
was
reluctant
to
attribute
causalitY.
Physiology
and
Further
Implications
Lately,
the
physiological
effects
of
DST
have
received
empirical
attention.
Coren
(1996b)
found
a JTo
increase
in traffic
accident
rates
associated
with the
spring
shift
to
DST' and
ascribed
it to
disrupted
circadian
rhythms.'
He
also
found
a similar
result
for
(nontraffic)
acci-
dents
more
generally
(Coren,
1996a).
Then,
in a
third
study,
an even
larget
l77o traffic
fatality
effect
was
found
and
persuasively
linked
to the
sleep
deficit
aftermath
of
DST
through
day-part
analysis
(Coren,
1998).
Several
other
researchers
have isolated
circa-
dian
rhythm
abnormalities
as a
result
of
DST
(Kantermann, Juda,
Merrow,
&
Roenneberg,
2007;
National
Sleep
Foundation,
Sleep
and
Teens
Task
Force,
2000;
Wagner,
t999; by
inference,
Terman
&
Terman,
2005),
and
then
seasonal
affective
disorder
(SAD),
in turn,
as a
consequence
of
the circadian
rhythm
problem
(Avery
et
a1.,
2001;
Lewy,
Lefler,
Emens,
&
Bauer,
2006; Terman
&
Terman,
2005; White,
Terman,
Musa, &
Avery,
2005).
Particular
symptoms
of insufficient
morning
daylight
in-
clude
low
energy,
hypersomnia,
and excessive
food
intake
(Avery,
2000).
Finally,
the scholastic
and
academic
perfor-
mance
implications
of SAD
and
deficient
circa-
dian
rhythms
have been
raised
prominently
(National
Sleep
Foundation,
Sleep
and
Teens
Task
Force,
2000).
The
overriding
hypothesis
is
that
DST damages
student
performance
via cir-
cadian
rhythm
disruption
(possibly
aggravated
by
SAD).
Essentially,
forcing
schoolchildren
to
awaken
and
try to
function
an
hour earlier
than
their
biological
clocks
command,
an
hour ear-
lier
than
natural
time,
would be
detrimental
to
scholastic
achievement.
For
the
individual's
daily
routine
to
grossly violate
the
quotidian sun
cycle
does
indeed cause
physiological
and
psy-
chological
harm. Scientific
evidence
in the med-
ical
literature
cited
here
affirms
the connection
(see
also
lnternational
Agency
for
Research
on
Cancer,
in
press; Moore,
2006).
The basic
mechanism
is the
astronomical
light-darkness
cycle
that
regulates
human
biorhythms,
along
with
the derived
lethargy
from extended
morn-
ing darkness-in
contravention
of
the
natural
harmony
and
possibly accentuated
by the
semi-
annual
clock
change-which
undermines
intel-
lectual
function
and
even
emotion.
(There
ap-
pear
to
be dysfunctional
effects
of
the spring-
fall clock
change,
to
be
sure, especially
for
traffic
safety
as cited,
but
the
main
phenomenon
at
issue
is the
long-term
overhang
of
life spent
under
the DST
regimen.)
This thinking,
along
with
the empirical
lineage,
underlies
the
hy-
I
Circadian
rhythm
is the scientifically
established
natural
biological
clock
in humans
that
is attuned
to daily
light
and
darkness
(Kantermann
et al.,
2007;
National
Institutes
ot'
Health.
National
Institute
of
Mental
Health,
1999;
Wagner,
1
999).

46
potheses,
design,
and
presently.
reseiuch
effort
rePorted
Method
To reify
previous speculative
hypothesizing
about
psychological and
academic
effects
of
DST, we accessed
secondary
data on Scholastic
Aptitude
Test
(SAT)
scores
at the
high school
level to be
related to
the
presence
or
absence of
the
DST environment
for the surrounding
area
of the
respective schools.
The data set
incorpo-
rated
mean SAT
score
(combined
math
and
verbal, the
latter
now known
as
"reading")
for
those
tested at
each of
the entire
population of
approximately
350
Indiana
public high schools
over
a
lO-year
period
(199'7-2006). Private,
mainly Catholic,
high
schools,
which
represent
about
47a of the
public
total,
were excluded
on
the
grounds
that
they embody
a
systematically
different
phenomenon and setting
than
the
pub-
lic sector.
Therefore,
the unit of
analysis
was the
high
school, and
the
gross
number
of observa-
tions
was 3,501.
(Number
of
high schools
fluc-
tuates
from
year
to
year primarily because
of
consolidation.
With an average
of about
98 tes-
tees
per
school/year,
and
33,989
total
per year,
an
imputed data
set of 339,893
cases
over the
l0
years
can thus
be construed.)
The score
for each school,
the
dependent
variable,
was the
mean SAT score
for
a
given
year,
with
that average
weighted by
each
school's
number
of test-takers.
Of the
3,501
observations,
583
were from the
Central
(with
DST) time
zone counties
(see
Figure
l, in red),
160 came
from the
Eastern
(with
DST) area
(in
green), and the
remainder
(2,758
cases)
were
Eastern standerd
(no
DST) data
points. As can
be seen.
Indiana
is a state
that
has allowed
ditferent
DST regimens
in different
regional
corners.
The source
for the
SAT data
was the office
of
the Indiana
State Superintendent
of
Education.
Also, the
incidence of
nonlocal or
out-of-region
test-takers
at a
given school
is
a
negligible
noise
factor.
Each
high school
was then objectively
coded
as either
a
DST or non-DST
observation
per
the
county-by-county
map shown
in Figure
l. On
average,
across the
10
years covered,
21.27o of
the schools
were
DST and'78.8%
were non-
DST.
(The
state
changed
its
DST law in
2005,
but the
policy
was
not implemented
until
after
the
2006 measurement
period.)
Weighted
least
squares
linear
regression
analysis
was used
with high school
(that
is,
county)
DST
status
as the
dummy
independent
variable and observations
weighted
by number
of school
SAT
test-takers.
Because
Eastern
DST is clearly a
different
temporal
phenome-
non than
Central
DST,
per
Figure
I
again,
this
variable, that
is,
time
zone,
was controlled
for
by
proxy,
entering
as degrees
of longitude
rel-
ative to
the time
zone meridian,
subject
to the
following
qualifi
cation.
In
view of the
inherent nonlinearity
involving
this
predictor-that
is, 360' of change
returns
to
the starting
point,
thereby
producing nonsense
results of
different effects
for 0"
and 360'-a
first-level
Fourier series
model
(Arfken,
1985)
was used,
with the sine
of the
longitudinal de-
viation
from the time
zone meridian
also enter-
ing as
regressor.
The Fourier
is a method
of
estimation
via trigonometric
functions
such as
sine or cosine.
The
role here is simply
to
include
the corresponding
sine
of a
longitudinal-based
independent
variable.
This amounts
to a way
of
capturing
or finessing
the
nonlinearity
within
the confines
of
linear regression,
although
the
linearity assumption
would have been
a reason-
able approximation
for the
relevant, nalrow
geographic range of data.
That
is, as Indiana
is
only 3'
wide longitudinally,
roughly between
85" and
88o west
longitude, using
a linear
model
produces
almost
the exact
same
results as
the
more theoretically
correct sine
or cosine
functions.
Those
functions are
nearly linear
in a
very small
range of degrees
or
radians such as
Indiana's
longitudinal extent.
Socioeconomic
status
(SES)-of
the
high
school unit
of analysis-was
controlled
for by
an
inverse
proxy: proportion of student
lunches
subsidized
by
the school.
This construct
was
actually
entered
in the
regression as two
sepa-
rate
variables:
proportion
of
lunches completely
subsidized
and
proportion
partially subsidized,
normalized
as a school's
difference
from the
statewide
weighted
(by
enrollment)
average
for
a
given
year.
(Those
norms, across the
l0
years
of
data, are
l5Vo fully subsidized
and
6Vo
partially.)
Race
or ethnicity
was also controlled
via
vari-
ables created
for
four different
groups:
differ-
ence
between a school's
enrolled
proportion of
a
given
ethnic
group
and the statewide
weighted
GASKI
AND SACARIN

DAYLIGHT-SAVING
TIME AND
SAT SCORES
Figure
t
. Geographic
daylight-saving
time
(DST)
segmentation of
Indiana counties'
Darkly
shaded
(red)
areas
in northwest and
southwest: Central
time with
DST. Lightly shaded
(green)
areas
in southeast:
Eastern time
with DST. Unshaded
areas:
year-round
Eastem
standard time.
(by
enrollment)
average.
(Because
control
vari-
able
data were
obtained
from other
sources,
usable
N for some
analyses
was 3,440.)
Aside
from the explicit
use
of
geography
(time
zone-
or longitudinal-based
sine)
as a
predictor vari-
able, that
factor tends to
wash out as
a
potential
bias
substantively
because
mean
latitude and
longitude of the
DST areas
combined are
non-
significantly
distant
from the
same for the bal-
ance of
the state:
39.349 N.
latitude. 86.703
W.

48
GASKI AND SACARIN
longitude
versus 39.937 N.
latitude,86.115 W.
longitude, respectively.
(Each
geographic
data
point,
or a county's
score, is its official
U.S.
Geological Survey
coordinates. The same
com-
parison
based on
latitudeilongitude of the
indi-
vidual schools
yielded
virtually identical
results.)
Indiana, as in the
Kotchen and Grant
(2008)
study,
was considered a
quasi-natural laboratory
because
of the
variation in
phenomena
exam-
ined. The state exhibits
not only multiple time
zones
but
different DST observance
patterns
in
different
regions across the data set.
Table I
shows
Indiana's national representativeness
in
terms of some
common demographics.
Moreover, Indiana is
geographically
repre-
sentative
objectively because the
median center
of U.S.
population
lies within the state.
(This
parameter
is officially defined as the
intersec-
tion of
two
perpendicular median lines such that
half
the
nation's
population resides
east
[and
westl
of
the north-south line, and
half lives
north
[and
south] of the east-west
line; U.S.
Census
Bureau, 2007.)
Formally hypothesized, consistent
with the
preponderance
of tangentially
related literature
as
highlighted, is that
DST would
tend
to dam-
age
SAT
performance
because of the
cognitive
effects
of circadian
rhythm
disruption
or some
similar
physiological
mechanism.
(The
study's
main
purpose
was to establish an
evidentiary
baseline,
not
provide
a
medical diagnosis.) To
Table I
Selected
Demographics: U.S.-lndiana Comparison
(2007
Data)
Variable
Indiana U.S
be clear, the
hypothesized relation is not attrib-
uted
to
the
short-term effect of the semiannual
time change event
per
se, but to the condition of
being out of
phase
with natural solar time and
accompanying biorhythms
for several months
per year
over at least
much
of one's
lifetime. It
is not a matter of temporal
proximity
of the
clock
change occurrence to the SAT test occa-
sion
itself.
The null hypothesis, which also may seem
very
plausible,
is that DST
is
totally unrelated to
SAT score.
Results
The coefficients of
Table 2
(Model
IA) dis-
close a
16.34-point negative effect on SAT
scores associated
with DST. In view of the
public
outcry
often observed in reaction to
slight annual
changes in either intrastate or na-
tional SAT aggregate
performance,
this result
can be considered
substantively significant,
even
staftling and
provocative.
A modest and
crude form of
pragmatic
vali-
dation
is
seen
in Indiana's 6-point statewide
decline
in 2001 SAT score
(Indiana
Department
of
Education,2001), following a 32-point
gain
between
1990 and 2006. This finding is roughly
in
line with what might be expected of a tested
cohort
in
which
a
preponderance
of students
would
have had only a single fractional season's
experience
with DST, in the wake of the newly
mandated uni form
practice.
One
minor and uncontrollable confound is
geo-
graphic
in-migration, that
is,
the
inevitability
that
the DST category
will include some test-takers
who
have not spent their complete scholastic ca-
reers
in
a
DST locale. Likewise, some from des-
ignated
non-DST test sites would have experi-
enced
DST
at some
previous
time in their lives.
However, this
noise would only dilute or attenuate
the
measured effect by artilicially elevating the
comparatively
low DST-area SAT scores
and
re-
ducing
the non-DST-based test
results.
So,
if
any-
thing, the
16.34-point effect size only understates
the true
magnitude
and
the unknown migration
error does
not alter the basic result or implication.
The analysis also
investigated
the effect ol'
time zone
geography. As mentioned,
position
within time zone
was
expressed and coded
with
respect
to
nominal time zone meridian as a high
school location's degrees
of longitudinal dis-
tance from that
meridian
(but
now without con-
Mean income
($)
Median age
(years)
Age'. Vo 65 and over
Race,7c
White
Black
Hispanic/Latino
Two or more race
groups
Unemployment
rate, 7o
Mean household/family
size
High school/college
graduate
or
higher, 7c
Sources. U.S.
Census Bureau
(2008);
Demographics
of
the
United
States
(2008);
STATS
Indiana
(2007).
'State-level
race categories total more than 100o/o because
multiple responses
wers
allowed.
33,616 3rJ,6l
l
36.3
31.9
t2.4 12.6
88.3 74.0
8.9
r3.4
.+.8
14.8
1.1"
2.0
1.5 4.5
2.52/3.06 2.61/3.2
85.2t21.7 84.1/2'7.0

DAYLIGHT-SAVING
TIME AND SAT
SCORES
difference
SES
(proxy)
Reduced rate
lunch
Free lunch
Race
Black
Aslan
Hispanic
Multiracial
l 1.338 0.750
.453
-0.841
-2.35'7
.018
1.146
-
14.089 .000
-0.794 -
13.400
.000
14.842
32.122 .000
0.965
6.646 .000
5.651
8.274 .000
Table
2
Regression Stetistics:
Effects on SAT
Scores
Independent
variable
b
t
Regression
model IA
DST
-16.331
-4.941
.000
Time zone
(proxy): sine of
longitude
-
meridian
mathematical
time
zone, Central,
but
with nearly
all
of
its
territory
assigned
to another
zone. This
longitudinal
anomaly
is one
way in
which Indiana
is t1ol
geographically representative.)
The result of
this analysis,
again
controlling
for the
other referenced
variables
including
DST, indicates
a
nonsignificant
0. 197-point
SAT score
decrement
for every
degree
of west-
ward deviation
from
time
zone meridian
(be-
cause
the
positive
coefficient
is associated
with
positive, or eastward,
change
in the
variable; see
Table
2-panel II).
At 15" of
longitude
per
time
zone,
the incremental
effect
that accrues
from
Indiana's
presence in the
Eastern time
zone as
opposed
to the
Central
time zone
would there-
fore be
-2.95
SAT
points, had the
result been
significant.
Because of
the severe
multicolinearity
be-
tween
DST and the
geographic control
variables
(longitude
and
its sine,
respectively;
see
Table 3)
in
the
two
regressions, the
analysis
was
repeated
without
the
geographics. Although
suboptimal
conceptually,
this approach
should
be
tolerable
and
functional
operationally
because
of
near-
linearity within
the
relevant, serviced
range
of
15" of 360".
The
results,
which essentially
ratify
the
prior ones
(b
:
-14.159
for DST
effect),
are shown
in Model
IB of Table
2.
The R'
values
remain strong
throughout,
in the
range of
.56.
Although there
is a
(methodologically) lon-
gitudinal
dimension
to the
data set
in the
form
of the dependent
and SES
control
variables,
independent
variables
DST and
time
zonel
geographic
longitude are
invariate
over time at
the
school
level.
Nevertheless,
the
Durbin-
Watson
statistic
was computed
and
found to
be
no less than
a tolerable
1.63 across
regressions.
Alternative
Analysis
DST is a
phenomenon unto itself,
featuring
its semiannual
temporal
disruption.
Yet,
what
may be
called absolute
time
(within
our
physi-
cal
frame of
reference)
can also be
considered
a
distinct
phenomenon, Einstein's
relativity
not-
withstanding.
In other
words,
Indiana's
Eastern
DST-observing
counties experience
a different
time regimen
than the
Central
DST
pockets.
Both
have
lived on different
times,
over the
course
of the
year,
than
the
Eastern standard
area throughout
the
data collection
period.
(For
DST
SES
(proxy)
Reduced
rate
lunch
Free lunch
Race
Black
Asian
Hispanic
Multiracial
Regression
model IB
-14.159
-0.845
1.757
-0.788
t4.832
-0.942
5.620
8.98.t
-
2.368
11.251
-
13.419
32.tt6
-6.641
8.241
.000
.018
.000
.000
.000
.000
.000
Regression
DST
Time
zone
(proxy):
longitude
-
meridian
difference
SES
(proxy)
Reduced
rate lunch
Free lunch
Racc
Black
Asran
Hispanic
Multiracial
model
II
16.331
-4.945
.000
0.197
0.750
0.841
2.351
-1.746
-
14.089
-0.194
-
13.400
14.842 32.122
-0.965
-6.646
5.651
8.211
.453
.01 8
.000
.000
.000
.000
.000
Note. SAT
:
Scholastic
Aptitude
Test; DST
:
daylight-
saving
time; SES
:
socioeconomic
status.
R'
:
.563;
Durbin-Watson
statistic
-
1.638. Degrees
of freedom
for r
statistics
in
Model IA: 3,431;
Model lB: 3,432;
Mode'
II: 3.431 .
verting
to
its sine, as
before
when used
as a
control).
In
Indiana,
all
geographic
points
would
receive
negative
scores as
lying to
the
west of
the
Eastem time
zone
meridian,
which
bisects
the
states
of
Pennsylvania
and
New
York.
For
instance, a
point in the center
of
Indiana
is at 86.2"
west
longitude
and
is
thereby
scored
-11.2,
as
the Eastern
zone
meridian
is
75'.
(Indiana
happens
to be
the only
state
posi-
tioned
entirely
within
one
geographic, theoretical,

50
Table 3
Omnibus
Correlation
Matrix
GASKI
AND SAGARIN
Variable
I. SAT
2. DST
3.
Time zone
proxy-sine
4. Time zone
proxy
5. Reduced
rate lunch
6.
Free lunch
7. Black
8. Asian
9. Hispanic
10. Multiracial
1.0
.185
1.0
-.t'74
.860
-.174
.860
-.302 -.078
.553
.090
-.438
.
l 78
.363
-.05
I
.245
.261
.084 .096
1.0
.999 1.0
-.089
.089
.103
.103
.226 .226
-.036
-
.036
.328
.328
.091 .091
1.0
.587
1.0
. r03 .648
1.0
-.162
-.082
.tI8
l.0
.21]t
.423 .213 .0'71
1.0
.100 .082
.080 .2lt .112
1.0
Note. SAT
-
Scholastic
Aptitude Test;
DST
:
daylighrsaving
time
part
of the
year,
however, each
DST zone has
had a common time
with the standard
zone.)
For these
reasons, recognizing these
conditions,
and to bolster
the earlier
regression findings
for
time zone
(longitude),
we devised
a supplemen-
tary analysis
combining
DST
and
time zone into a
single
time dimension,
which served as an
inde-
pendent
variable.
This
analysis
compared the
basic SAT effects
across the three
time-
geographic
groups,
controlling
for the SES
proxies.
SES
was
trichotomized
per
the follow-
ing rule:
A
school
with higher than
the state-
wide
norm
proportion
of
free
(or
partially
free)
lunches
was
so classifled
as
in
that
category.
The / test
results are shown
in Table
4. The
essential
finding remains
that DST is detrimen-
tal,
but that the underclass
is hit even
harder in
terms
of damage to SAT
scores.
The
column-
wise SAT
means for the DST
zones are signif-
icantly lower than the corresponding
standard
time
scores,
with
the
effect more
pronounced
within the two
paid
lunch conditions.
This is
highlighted by the
EST-(weighted average)
DST difference scores
in Table
4, revealing
greater
SAT
decrement as SES declines
(22.69
and
48.89,
respectively, vs. 1.61).
Although
race/ethnicity was
not controlled in
this
analysis because
of cell
proliferation con-
cerns,
from the
Table 3 correlation
results, it
appears
that SES
is
a
fair
(not
ideal)
proxy
for
both
variables.
And
this
is but an augmentative
analysis.
Nevertheless,
with
an
extra measure of
ana-
lytical triangulation achieved
via two
regression
approaches
along with the between-groups
/ test
comparison,
natural concern over
the inescap-
able multicolinearity
may be
mitigated. Com-
bining
variables, as
in
the
final analysis
here, is
indeed a customary tactic
in dealing with such
colinearity
(Nie,
Hull, Jenkins, Steinbrenner, &
Bent, 1975,
p.3al).
Conclusion
and
Interpretation
Regardless of how
generalizable
the results
are, at
least
one
major U.S.
jurisdiction,
the
State of
Indiana, may wish to consider
an ap-
parent
detrime^ntal
effect of
its recently enacted
DST
policy.'
The nature of these
SAT-
impinging
effects
may
be understood
as
genu-
ine cognitive
damage at
least
as
much as related
to scholastic
performance
because
the SAT
exam
is more of an
aptitude than an achieve-
ment test. Starkly
expressed, DST appears to
cause brain damage.
In this
less-than-truly
ex-
perimental design, of course, causality
cannot
be definitely
inferred. The
reverse
causation
re-
mains
possible,
that is, that the cognitively
de-
ficient or
poor
SAT
populations will be more
likely to embrace
DST
(a
finding
that
would be
rather
notable in its own
right). Yet, the intui-
tively
more likely causal
sequence has been
represented
here as the default.
The third
pos-
sibility, that
DST and SAT are
mutual effects of
a third
variable, seems exceedingly
remote.
Again,
no
explanatory
diagnostics
are
prof-
fered beyond
the established empirical
record,
which would
nominate circadian
rhythm and
derived
SAD depression
problems resulting
cu-
mulatively
from the seasonal time change(s).
2
Other
regions that
have struggled with the DST
issue in
recent
years
are Florida,
Maine, Ohio, Puerto
Rico,
and
Australia.

DAYLIGHT,SAVING
TIME
AND
SAT
SCORES
Table 4
Composite
DST/Time Zone
EJfects on SAT Scores
(Controlling
Jbr
Socioeconomic
Stcttus)
Socioeconomic
status scgment
Variable
Free lunch Partially
paid Paid lunch
Total
population
Central
DST
EST
Eastern
DST
Weighted average
(mean)
SAT score
for combined
DST zones
Mean
diflerence
(EST
DST)
908.90
(l0l)
966.
l 9"
(396)
9ss.89
(22)
917.30b
48.89
946.98
(30)
913.44"
(ts4)
e62.91
(9)
950.7-5"
22.69
994.33
(439)
es8.76
(s70)
998.42"(2.r60)
986.39'(2,710)
978.55
(129)
970.17
(160)
990.75b
7.6'7
Note.
DST
:
daylight-saving
time; SAT
-
Scholastic
Aptitude
Test; EST
:
Eastem
standard time. Number of
cases
per
cell are shown
within
parentheses.
'EST
mean
is significantly
different from each
of the other two within-column
(socioeconomic
status segment) entries
at
p
=
.OI.
h
Significantly
diffcrent
from other two within-row
means at
p
=
.01. Degrees
of freedom for
reported I test
comnarisons:
l0
=
d.f.
<
733.
(The
SAD
issue
would be
grounded in DST's
abnormally
extended
morning darkness
for
much
of the school
year.)
Serious
medical con-
cern
has
been
adduced
for circadian
rhythm-
based
sleep deprivation
and
its
consequences
for the
adolescent
population in
particular
(Na-
tional Sleep
Foundation,
Sleep and
Teens Task
Force,
2000;
Papatheodorou &
Kutcher,1995),
which is also the
SAT-taking
population. In
view of this
very
broad
and deep
scientific
lit-
erature
on the cognitive
and
psychiatric
detri-
ment
of SAD
(see
Terman &
Terman, 2005,
for
a review),
perhaps
the
magnitude of the
DST-
SAT
effects found
here is not so extreme.
(For
that
matter, cognitive
dysfunction
caused by
DST has even been
manifested
in lower stock
market
returns
following the
annual time-
change
weekends;
Kamstra,
Kramer, &
Levi,
2000.)
Economic
Extension
One other
aspect of
any induced decrement
in
cognitive
capacity
qua
SAT
performance is
the
long-term economic
ramification applying
to
the
afflicted
parties-individuals or a
state en-
tity.
Murnane,
Willett,
Duhaldeborde, and
Tyler
(2000)
found a
l-point difference
in national
standardized
aptitude
test scores to
be associ-
ated
with about a
l.5%o difference
in annual
individual earnings
(measured
at ages
2l and
3l).'Converting
the scale
used
(mean22.2)
to
the
SAT
norm
(900
mean
when data collected,
before
recentering)
equates
to a 0.0317a earn-
ings
increment associated
with a
point
of SAT
score.
(This
result
is highly congruent
with a
nonrefereed finding of
between 0.03l%o
and
0.044Vo in earnings
increment
per
SAT
point
[Cheslock
& Hilmer,
2001,
p. 23]. Tsui
ll998l
found
a 0.0487o
income
effect of SAT
point within management
occupations.)
The 16.34-point
result excavated
here therefore
produces
a 0.605Va
impact on
personal
earn-
ings,
relying
on
the Murnane
et al. magnitude.
Applied to
mean
per
capita
Indiana annual
in-
come
of
$33,616
(as
of 2007), the
16.34-point
SAT score effect
translates into a
$203.38
indi-
vidual income decrement.
Extended over
Indiana's high school
popula-
tion
cohort of
499,346 as of the
2007-2008
academic
year,
at the empirical
16*
points per
student
(irrespective
of
whether all actually take
the
SAT), that
is
an aggregate
annual
loss of
$101,556,980.
Hypothetically extended
to the
state's entire
population
of 6,345,289
(per
the
2007 census estimate),
assuming
no statewide
population
or average
income
growth
but
that
all
would either attend
high school at some
time
or
need cognitive capability
otherwise, the
eco-
nomic
loss would be
$1.291
billion annually.
Extrapolated on
the same terms
to the DST-
observing
U.S.
population
(i.e.,
all states
except
Arizona and Hawaii) of
296,202,503
per
latest
estimate, at 0.6057o
of average
2007
per
capita
national
income of
$38,611,
or
$233.60
per
person,
the total
yearly
economic
loss would be
3
In that study, the standardized
test was not the SAT
but
a comparable instrument
of the Educational
Testing
Service.

52
GASKI AND SAGARIN
$69.192
billion as an
apparent outcome of
DST's
SAT
impact
alone.
Finally,
extrapolating further
across an ex-
pected
4O-year working lifetime,
conservatively
assuming constant income, Indiana's
DST-
induced earnings loss for
the full
population
would
be
$51.64
billion, and the
projected
na-
tional
(non-Arizona,/Hawaii)
loss is
$2.768
tril-
lion. If the weaker
-
14.105
coefficient of Table 2
(Model
IB) were substituted
throughout this anal-
ysis,
each estimate would simply
be 86.347o of
those shown-for example,
a national
long-term
effect of
$2.390
trillion instead
of
$2.768
trillion.
These
projections,
of course, do not
even consider
the
quality-of-life
implication
of
DST-based
cog-
nitive
damage.
Another reason
why minimal interpretation is
appended here is
that the reported numbers
largely would
seem to speak for themselves. No
public policy prescription
is
advanced other
than to
quote
Kamstra,
Kramer, and Levi
(2000,
p.
l0l0). Citing
available evidence of economic
damage attributable
to
DST,
those authors of-
fered,
"an
obvious
policy
implication
is to do
away with the time change altogether."
To
be
fair, for
one nation or state
to abandon
the widespread
worldwide DST
practice
could
have
some offsetting negative economic impact,
possibly
stemming from commercial inconve-
nience.
Given the results represented here, how-
ever, the Kamstra
et al.
(2000)
suggestion may
be something for world
governments
to
ponder
collectively-unless
those
govemment
officials
have
already suffered
the cognitive debilitation
of too much DST.
References
Arfken,
G.
(1985).
Fourier series. In H.
J.
Weber
&
G. B. Arfken
(Eds.),
Mathematical methods
for
physicists (3rd
ed.,
pp.
760-793). Orlando, FL:
Academic Press.
Avery,
D. H.
(2000).
Bright
light, dayvn simulation,
and a reg,ular light-dark
cycle in the treatment oJ
SAD.
Working paper,
University
of
Washington.
Avery, D.
H., Eder, D. N., Bolte, M. A.,
Hellekson,
C. J.,
Dunner,
D. L., Viticllo, M. V.,
&
Prinz, P.
N.
(2001).
Dawn
simulation and bright light in
the
treatment
of SAD: A controlled study. Bioktgical
Ps.\t hiatn. 50. 205-216.
California Energy
Commission.
(2001).
The effects of
davlight saving
time on California electricity use
(Publication
No. 400-01-013).
Sacramento,
CA:
Author.
Cheslock, J. J., & Hilmer, M.
J.
(2001).
How college
enrollment strategies
affect student labor market
success. Working
paper,
Cornell Higher Education
Research Institute,
Cornell University.
Coren,
S.
(1996a).
Accidental death
and the shift to
daylight
savings Iime. Perceptual
and Motor
Skills, 83,921-922.
Coren, S.
(1996b).
Daylight
savings
time and traffic
accidents. New England
Journal of Medicine, 331,
924-925.
Coren, S.
(1998,
December).
Sleep
defcit,
fatal
acci-
dents, and the spring
shift to daylight savings
time.
Presented
at INABIS
'98-Fifth
Intemet World
Con-
gress
on Biomedical
Sciences at McMaster University,
Canada.
Retrieved
from http://www.mcmaster
.calinabis98/occupationaVcoren
Demographics
of the United States. Denographics.
(2008).
http://en.wikipedia.org/wiki/Demograph-
ics
Gaski, J. F.
(2008).
Indiana
Time ktw and
its detri-
mentoL efJects: Time
to
get
it right-this
time.
Working paper,
University
of Notre Dame.
Goodman, N. G.
(Ed.). (1931).
Benjamin
Franklin's
essay on daylight
saving: Letter to
the editor ofthe
Journal o.f Paris, 1784.
In The ingenious
Dr.
Franklin.
Selected scientifc
Letters
(pp.
17 22).
Philadclphia:
University of Pennsylvania
Press.
Indiana
Departmcnt
of
Education.
(2007).
SAT
scores
decline overall in Indfuna. Retrieved
fiom http://
www.doe.
state. in.us/reed/newsr/2007/09-August/
SAT
lnternational
Agency for
Research on
Cancer.
(ln
press).
Shifrwork,
painting
and
fire-fighting.
In
IARC monogruphs
on the evaLuution
of carcino-
genic
risks
to humans. Lyon, France:
Author.
Kamstra, M.
J., Kramer, L. A.,
&
Levi,
M. D.
(2000).
Losing slcep
at the market: The
daylight saving
anomaly. American Economic
Review, 90, 1005-
l0l l.
Kantermann,
T., Juda, M., Merrow,
M., & Roen-
neberg, T.
(2007).
The human
circadian clock's
seasonal adjustment is
disrupted by daylight
sav-
ing
time. Current Biobgv, 17,
1996 2OOO.
Kelfogg, R.,
&
Wolfl
H.
(2007).
Does
extendin,s
duylight saving
time save energy-? Evitlence
.fiom
an Australian
experiment
(IZA
Discussion Paper
No. 2704, Working
Paper 163). Bcrkeley,
CA:
Center fbr the Study
of
Energy
Markets.
Kotchen,
M. J., &
Grant,
L. E.
(2008).
Does
tlctl-lig,ht
saving time save energy? Evidence.from
a natural
e.rperiment
in Indiana. Working paper,
Univcrsity
of Califbmia
at Santa Barbara and National
Bureau
of Economic Research.
Lewy, A. J., Lefler, B.
J., Emens,
J. S., &
Bauer,
V. K.
(2006).
The circadian
basis of winter
depres-
sion. Proceedings
of the National Academl,
of Sci-
ences. USA. 103. 7414-7419.

DAYLIGHT.SAVING
TIME
AND SAT SCORES
53
Moore, S.
(2006,
September
1). Starting
too early.
OpinionJournal.com.
Retrieved
from http:/i
www.opinionJoumal.com
Murnane.
R.
J.,
Willett,
J. B.,
Duhaldeborde,
Y., &
Tyler,
J. H.
(2000).
How
important
are the
cogni-
tive
skills
of teenagers
in
predicting subsequent
earnings?
Journol
of
Policy AnuLysis
and
Manage-
ment.
19.547-568.
National
Institutes
of
Health,
National
Institute of
Mental
Health.
(1999).
How biological
clocks'
worft
(NIH
Publication
No.
99-4603).
Bethesda,
MD:
Author.
National Sleep
Foundation, Sleep
and
Teens
Task
Force.
(2000).
Adolescent
sleep
needs and
pat-
rerns.
Washington,
DC:
Author.
Nie.
N. H..
Hull, H. C., Jenkins,
J.
G., Steinbrenner,
K.,
&
Bent,
D. H.
(1975).
Sttttistical
package
Jbr
the
social
sciences
(2nd
ed.).
New
York: McGraw-Hill'
Papatheodorou,
G.,
& Kutcher,
S.
(1995).
The effect
of adjunctive
light therapy
on
ameliorating
break-
through
deprcssive
symptoms
in adolescent-onset
bipolar
disorder. Journal
of
Psychiatry
and
Neu-
ro.\( i
('
nce. 20, 226
-232.
STATS
Indiana.
(2001).
Intlianapolis,
IN.
Retrieved
fiom
http://www.stats.indiana.edu/profi
les/
Terman,
M., &
Terman, J. S.
(2005).
Light therapy
for seasonal
and nonscasonal
depression:
Efficacy,
protocol, safety,
and
side effects.
CNS
Spec'
trums.
10, 647-663.
Tsui, L.
(1998).
The
effects ofgender,
education, and
personal skills self-confidence
on
income in busi-
ness
management.
Se-r
Roles, 38, 363-373.
U.S.
Census
Bureau.
(2007).
Position
oJ the
geogruphb
center
of areo,
mean aml
median centers
of
population:
2 000.
Retiev ed
fiom
http ://www.census.
gov/geo/
www/cenpop/cntpop2k.html
U.S. Census
Bureau.
(2008).
Stzrrlsrical
abstract
of
the
United States
(l27th
ed.).
Retrieved
fiom
http://www.census.
gov/statab/www/
U.S.
Department
of
Transportation.
(
1975). The
Day-
Iight Saving
Time
Studr-:
A report
to Congress
from
the
Secretary
of
Transportation.
Washington,
DC:
Author.
U.S.
National
Bureau of Standards.
(1976).
Review
and
technicul
evaluation
of the
DoT Daylight
Sav-
ing
Time Stur11,.
Prepared
for the
Chairman,
Sub-
committee
on
Transportation
and
Commerce,
Committee
on
Interstate and
Foreign
Commerce,
U.S.
House of
Representatives.
Washington,
DC:
U.S. Govemmcnt
Printing
Office.
Uniform
Time Act.
(1966).
Public
Law 89-387.
Wagner,
D. R.
(1999).
Circadian
rhythm
sleep disor-
ders.
Current
Treatment
Options
in
Neurology,
l,
299-308.
White,
T. M.,
Terman,
M., Musa,
G. J., &
Avery'
D. H.
(2005,
July).
Incitlence of
winter depression
varies
within time
zones.
Prcsented at
the Society
fbr
Light Treatmcnt
and
Biological
Rhythms
lTth
annual conference.
Eindhoven.
the
Netherlands.
d Psve,TESTS
Make
your
tests
available to
other
researchers
and students;
get
wider
recognition
for your
work.
"PsycTESTS
is
going
to
be an outstanding
resource
fbr
psychology," said
Ronald F' Lc:vtnt,
PhD.
"
I was ttmong
the
first
to
provide
some
of my tests
and
was happy
to do so.
They will
be
available
for
otlrcrs
to use-and
will
relieve me of
the admbistrative
tasks of
providitrg
them
to
individuals."
Yisit
htto://wvwv.opq.org/pubs/dotoboses/psyctests/collJor-tests.aspx
to learn
more about
PsycTESTS
and
how
you can
participate.
s?
Call
'l-800-374-2722
or
write to
Not since
PsycARTICLES
has a database
been
so
eagerly
antiqPatgd!