►
From YouTube: Community Facilities Study Meeting #5 Part 3 of 6
Description
Arlington VA Community Facilities Study Committee presentation on Validating Forecast Methodologies with consultant Richard Grip, Statistical Forecasting LLC. Recorded April 8 2015 at Wakefield High School.
http://commissions.arlingtonva.us/community-facilities-study/
A
B
But
one
of
here
tonight
is
to
talk
about
what
I
learned
in
my
meth,
illogical
review
of
the
school
enrollment
projections
in
the
district,
and
first
I'd
like
to
say,
is
that
the
district
is
employing
something
that
we
all
use
in.
The
industry
called
the
grade.
Progression,
ratio,
method
and-
and
you
may
hear
me
slip
a
few
times
tonight,
because
when
I
got
my
PhD
in
this
area,
we
call
it
the
cohort
survival
ratio
method
so
depend
upon
who
you
talk
to
in
the
field
in
context.
B
You'll
hear
different
things,
but
it
is
the
method
used
by
all
demographers.
That
I
am
aware
of
that
in
this
niche
field
called
school
demography,
but
I
want
to
talk
about
the
method.
A
little
bit,
I
know
as
Bob
and
mentioned
earlier.
You
have
had
talks
about
the
methodology
and
the
pieces
and
the
intricacies
and
I'm
not
going
to
bore
you
with
that,
but
what
I
will
tell
you?
It's
just
some
of
the
research
that,
over
the
years
has
been
compiled
about
this
method.
B
That
I
think
you
may
find
interesting
in
terms
of
its
accuracy
and
reliability
depending
upon
the
different
research
studies
that
have
been
out
there.
The
accuracy
range
is
anywheres
between
one
and
three
years.
You
see
some
that
go
out
as
far
as
seven,
but
I
would
definitely
not
condone
that,
but
impractical
theory,
three
to
four
years
is
about
right,
because
this
type
of
methodology
is
is
a
linear
trend.
It's
really
assuming
that
what
is
ever
happening
in
your
district
in
the
past
is
going
to
continue
into
the
future.
B
So
where
does
this
method
really
start
to
break
down?
Let's
be
higher.
Okay,
where
is
the
method
starts
to
break
down?
Is
in
districts
where
trends
really
start
to
change.
So
I'll
give
you
an
example:
let's
say
that
you
were
in
a
community
not
talking
about
arlington,
but
that
had
very,
very
little
housing
growth
for
the
last
five
to
ten
years
and
all
of
a
sudden,
a
large-scale
development
started
to
come
into
that
particular
community.
Well,
the
methodology
itself
doesn't
really
pick
up
a
lot
of
new
housing.
B
B
All
right
so
again
just
a
really
brief
overview
of
what's
been
happening
here.
The
historical
enrollments
are
collected
in
your
district
by
the
local
attendance
areas,
a
small
unit
of
analysis,
rather
than
looking
at
the
whole
entire
k
to
12
district
you're.
Looking
at
the
enrollments
at
the
elementary
attendance
areas,
the
feeder
patterns
into
the
middle
and
feeder
patterns
into
the
high
school,
and
we
call
that
a
bottom-up
approach.
B
We
wouldn't
do
this,
typically
in
a
very
small
district
when
I'm
talking
about
a
small
district,
I'm
saying
anything
under
about
5,000
students,
because
the
variability
of
movement,
it's
and
a
very
small
grade
levels
can
be
too
much
where
this
this
methodology
won't
work.
But
at
your
district
at
25,000,
students
is
the
perfect
methodology
to
use
in
years
6
to
10
of
the
projection
you
you're,
estimating
burt's
by
a
three-year
rolling
average.
B
Let
me
talk
about
this
for
a
moment,
because
when
I
do
my
projections
for
my
clients,
I
work
for
the
the
largest
school
district
in
the
nation,
I
work
for
the
New
York
City
Public,
Schools
they're,
one
of
the
few
clients
where
I
actually
do
a
ten-year
projection.
I
would
say
about
ninety-five
percent
of
my
clients
were
actually
doing
a
five-year
projection
because
the
difficulty
becomes
in
year
six
to
ten
you're
starting
to
project
kids,
who
haven't
even
born.
B
Yet
so,
let's
say
here
we
are
in
2015
if
we're
looking
at
your
six
of
a
projection
we're
talking
about
the
year
twenty,
twenty-one
well
those
kindergarten
students
in
2021,
if
you
back
out
five
years,
they're
born
in
2016.
Well,
that
hasn't
happened
yet
so
with
a
ten-year
projection
years,
6
to
10
are
very
difficult
because
you're
starting
to
project
birth
rates,
which
then
means
that
your
kindergarten
counts,
are
pretty
fuzzy
and
also
the
elementary
counts
as
well.
B
B
Okay,
the
next
thing
is,
you
use
something
called
student
generation
factors
here,
and
this
is
another
word
that
you're
going
to
hear
me
flip
flop
with
called
student
yields,
and
it's
just
really
the
number
of
children
/
housing
unit
and
what
your
district
is
doing
is
you're.
Taking
and
again
this
is
standing
the
field
looking
at
new
housing
and
trying
to
bump
up
your
baseline
enrollment
projections
to
account
for
from
new
children
from
those
new
housing
types.
B
So
the
internal
review
of
the
projections
does
occur
here
in
the
school
district
and
what
the
school
district
does
is
only
looks
out
one
year,
so,
if
they're
projecting
for
the
1415
year,
which
is
this
current
year-
and
you
did
it
the
year
prior
they're,
just
looking
at
the
one
year,
error
rate
and
they're
doing
that
every
single
time-
and
this
is
from
one
of
your
community
meetings-
I
saw
the
last
11
years.
Your
total
enrollment
has
been
within
plus
or
minus
two
percent.
I
think
it
was
in
10
of
the
11
times.
B
So
how
does
that
fair
in
terms
of
the
accuracy
and
what
school
demographers
in
the
field
believe
is
accurate?
Well,
this
is
a
tricky
one,
because
one
of
the
researchers
goes
back
to
a
paper
in
the
late
80s,
a
demographer
surveyed,
other
educational
planners,
and
felt
that
two-thirds
of
them
felt
that
plus
minus
one
percent
per
year
is
an
appropriate
benchmark
for
accuracy.
B
So
if
you
have
a
10,000
student
district
in
the
first
year,
a
one
percent
error
rate,
a
hundred
students,
if
you're
within
plus
minus
100
great
second
year,
if
you're
within
plus
minus
200
awesome
10
years,
you
have
up
to
a
thousand
students
a
ten
percent
error
rate
that
is
considered
to
be
okay.
But
what
they
didn't
say
in
this
article
is
that
what
kinds
of
districts
are
these
people
coming
from,
where
they
feel
that
a
one
percent
accuracy
rate
is?
B
B
There
are
two
ways
to
compute:
enrollments,
there's
a
top-down
method
and
there's
the
bottom-up
method
for
the
size
of
your
district.
The
bottom-up
method
by
attendance
area
is
the
best
way
to
go,
because
you
have
different
attendance
zones
that
have
different
rates
of
growth
depending
upon
different
kinds
of
housing
and
people
that
are
in
those
zones.
So
to
look
at
the
enrollment
and
project
K
to
12,
the
entire
district
would
really
lose
the
flavor
of
what's
going
on
at
the
individual
attendance
zone,
so
the
bottom-up
approach
is
the
perfect
way
to
go.
B
B
So,
as
bob
and
mentioned,
we
have
recommendations
and
the
recommendations
are
broken
down
into
immediate
and
things
that
should
happen
somewhere
down
the
line.
Now,
when
I
work
with
clients-
and
I
do
presentations
like
this
all
the
time
to
folks
just
like
you-
what
we
usually
do
is
a
demographic
study
as
an
output,
so
you
have
a
comprehensive
study
with
tables
and
charts
and
narrative
for
public
consumption,
and
the
school
district
does
not
really
publish
an
annual
report
like
that.
B
So
that
was
one
of
my
recommendations
is
that
we
need
to
have
something
like
that
out
there
for
for
the
people
to
see,
and
it
would
contain
not
only
the
projections
I
would
recommend
having
some
data
from
the
Census.
It
can
integrate
with
a
lot
of
things
that
Bob
it's
talked
about
before,
and
that
kind
of
comprehensive
material,
I
think
would
be
a
great
bonus
for
the
district
number.
Two
alternative
set
of
projections
never
do
I
just
do
one
set
of
projections
for
a
district,
because
this
is
not
an
absolute
certainty.
B
This
is
what
they
are
their
projections.
A
lot
of
my
colleagues
in
the
field
do
a
low
medium
and
a
high
set
of
projections
based
upon
different
scenarios.
I
know
it's
a
lot
of
extra
work,
but
listen.
This
is
what
it
is.
You
can
change
some
of
the
variables
around
to
see
what
might
happen,
for
instance,
housing.
B
Some
of
the
things
that
you
may
have
in
the
future
that
are
lined
up
in
the
pipeline
builders
may
say
you
know
what
I'm
not
going
to
build
that,
because
the
market
just
doesn't
sustain
it
right
now
and
that's
going
to
change
your
projection.
So
you
can
change
some
of
the
pipeline
information.
I'll
get
more
into
that
in
a
moment.
B
Number
three
perform
a
longitudinal
analysis
of
the
projections
really
simple,
rather
than
looking
just
one
year
out
to
see
how
the
error
rates
are.
Why
don't
you
go
back
five
years
and
take
a
look
and
see
how
the
actual
romans
five
years
out
compared
to
the
projective
so
that
we're
getting
more
than
just
a
one
year,
flavor
what's
going
on
and
that
really
tests
the
method?
B
So
if
we
can
do
that
even
further
back,
maybe
seven
or
eight
years
boy,
then
I
can
even
come
back
here
and
publish
because
then
I
can
see
that
we
have
a
ten-year
projection
that
works
really
well,
but
but
that's
really.
The
idea
is
that
we
want
to
test
the
methodology
by
doing
more
than
just
a
one-year
look
out.
B
Okay,
the
fifth
one
is
a
little
more
technical,
a
little
more
complicated.
It's
just
aggregating
student
generation
factors
by
attendance
area.
So
the
way
it
works
right
now
is
you
have
student
generation
factors
which
is
the
number
of
kids
/
housing
unit
by
housing
type
at
the
planning
unit
level
here
so,
for
instance,
if
you
have
a
elevator
apartments
that
are
going
to
be
scheduled
in
a
particular
planning
unit,
and
if
that
particular
planning
unit
has
never
had
housing
types
of
that
in
the
past,
well,
there's
no
student
generation
factors
that
you
can
use.
B
So
what
they're
currently
using
is
the
countywide
information
for
that
particular
housing
type.
My
suggestion
is
not
to
use
the
countywide
information
but
to
aggregate
the
planning
units
at
the
attendance
area
level.
A
smaller
level
of
geography
to
capture
that
local
flavor,
again
of
what's
going
on
in
in
the
particular
community,
so
it's
sort
of
an
in
middle
step,
an
intermediate
step
of
capturing
the
number
of
students
per
housing
unit.
B
Number
six,
this
one
off
to
talk
about
for
a
moment,
consider
past
home
construction
before
adding
students
in
due
to
new
home
construction.
So
here's
here's
what
I
have
to
get
a
little
bit
of
a
description
of
how
the
great
progression
ratios
work.
So,
let's
say
a
very
hypothetical
situation
in
the
last
five
years
here
that
500
single-family
detached
homes
were
built
and
out
of
those
500
homes.
One
kid
per
house,
500
kids
came
in
to
those
homes
and
then
entered
your
school
district,
so
say
100
per
year,
the
coat
the
come
here.
We
go.
B
The
great
progression
survival
rates,
the
great
progression
ratio
method
does
account
for
those
kids
moving
into
the
system
and
the
ratios
are
inflated
or
increase
due
to
those
kids
moving
in.
So
if
I
have
700
units
planned
in
the
next
five
years,
it
would
not
be
appropriate
to
add
700
children
into
the
projections,
because
you're
already
accounted
for
500
in
the
ratios.
B
You'd
have
to
take
the
difference
and
add
only
200,
so
to
make
a
long
story
short
is
what
I'm
asking
the
district
to
do
is
to
consider
the
past
home
growth
in
each
one
of
these
planning
units
before
you
start
adding
kids
into
the
system.
Now
from
what
I
understand,
there
hasn't
been
too
much
growth.
There
hasn't
been
too
much
historical
housing
and
a
lot
of
planning
units.
B
Well,
if
that's
the
case
well
then
this
this
whole
discussion
is
moot,
and
you
can
just
add
all
those
kids
in,
but
it's
very
important
that
you
consider
the
historical
counts.
Otherwise
you're
going
to
be
double
counting.
Kids
number
7
update
the
APS
website.
I
found
some
information
on
there.
That's
no
longer
pertinent,
no
longer
accurate
that
the
census
data
is
being
used
to
project
enrollments,
we're
from
what
I
understand
it's
not
being
used
at
all.
B
B
What
I'm,
saying
and
I'm
going
to
show
you
this
picture
in
a
moment,
is
to
not
do
it
that
way
and
to
do
it
by
length
of
ownership,
because
what
we
start
to
see
I'll
show
you
a
figure
in
a
moment
is
that
student
yields
are
not
the
same
across
the
board.
Once
you
get
out
to
about
14
or
15
years
of
ownership,
those
kids
they're
gone,
they've
left
the
system
they're
off
in
college,
or
now
you
have
empty
nesters.
B
Ok,
it
will
be
a
lot
more
easy
to
explain
in
a
moment
number
nine
talk
about
with
this
with
Bob
one
of
our
one
of
our
things
that
we
recommend
is
a
collaboration
between
the
county
government
and
the
school
district
in
projecting
Bert's
by
doing
forecasts
of
the
number
of
women
of
childbearing
ages
for
the
next
5
10
15
years.
You
can
then
use
age,
specific
fertility
rates
to
compute
the
number
of
births
for
women
at
different
age
groups,
and
then
that
will
give
us
a
number
of
births
going
forward.
B
It's
a
little
more
scientific
than
using
a
three-year
rolling
average.
What's
going
on
right
now,
and
I
only
recommend
that
recommend
this
type
of
technique
in
large
areas.
This
is
how
we
do
it
in
New
York
City
exactly
this
way,
but
you
have
to
have
a
big
pool
of
people
to
make
this
work
and
I
think
you
have
that,
and
the
last
thing
is
allow
the
the
personnel
from
a
ps2
and
the
school
facilities
area
to
attend
professional
conferences
in
school
demography.
There's
not
many
of
us
in
in
the
Nate
and
a
nation
actually
I.
B
Think
when
I
go
to
these
conferences
is
probably
about
25
or
30
of
us
sitting
in
the
room.
So
if
they
are
able
to
go
to
some
of
these
things,
they'll
get
they'll
pick
up
a
lot
of
ideas
in
a
lot
of
collegiality
between
all
of
us.
Here's
the
figure
that
I
mentioned.
This
is
really
neat.
So
this
is
a
community
in
right
outside
of
Princeton
in
New
Jersey,
it's
a
very
affluent
School
District,
median
single-family
homes
or
median
is
about
six
hundred
thousand
dollars
median
income.
Family
income
is
about
150.
B
This
is
a
district
where
people
are
climbing
over
each
other
when
there's
resales
when
there's
new
homes
to
get
into
now,
their
school
district
came
to
me
and
said:
Richie,
listen,
I
know
that
the
student
yields
that
were
were
using
are
probably
too
low.
How
do
you
come
up
with
an
accurate
generation?
Factor?
Student
yields
very
difficult.
So
what
you're
seeing
here
is
a
compliation
of
a
ton
of
data
and
you're
getting
the
very
end
of
the
output,
but
what
it
is.
B
If
you
look
at
the
bottom,
the
very
bottom
is
the
years
of
ownership
of
a
house.
It
goes
from
0
to
about
3
2
and
on
the
y-axis.
Is
the
yields
per
home
at
each
length
of
ownership?
So
what
I
had
to
do
is
I
had
to
go
into
a
property
database
and
look
at
the
most
recent
sale
for
this
particular
community.
Now
some
homes
never
sell.
Some
people
have
been
in
their
homes
for
40
years.
B
So
they
haven't
been
in
the
home
that
long,
they
moved
in
from
another
community.
If
this
community
had
just
taken
the
average
straight
across
the
board
number
of
total
units
of
detached
single-family
homes
by
the
total
number
of
kids
point,
seven
six
would
be
the
yield
so
I'm
telling
them.
No,
you
really
gotta
look
at
this
a
little
differently.
I
would
consider
if
I
was
them.
B
If
I
wanted
to
figure
out
what
the
yield
is
for
a
community,
just
look
at
the
first
ten
years
of
ownership
and
take
the
average
and
it
was
around
1.1,
which
is
very
different.
So
what's
my
point,
if
you're
a
really
very
good
dis
school
district,
which
I
believe
you
are,
and
people
want
to
come
here,
you're
going
to
have
a
distribution
that
looks
like
this
now
until
you
do
it,
no
one
will
really
know
now.
B
If
there's
a
now,
what
my
research
is
not
done
yet
is
school
districts
that
are
poor
and
lower
socio
economics,
because
my
feeling
is
that
the
length
of
ownership
model
will
not
be
as
stark,
you
might
have.
The
average
might
be
pretty
close
to
the
first
ten
years
of
ownership,
because
I'm
not
sure,
as
many
people
would
want
to
move
in
with
kids.
B
But
the
point
here
is
that,
if,
if
this
is
the
case
here
in
arlington,
then
you're,
probably
under
estimating
a
lot
of
the
student
generation
factors
currently,
okay,
I
want
to
invite
bob
up
for
the
last
slide
here,
which
is
additional
areas
of
collaboration
between
the
two
and
Bob.
Do
you
want
to
lead
on
this?
One.
A
So
you
probably
gathered
that
one
of
the
one
of
the
real
links
between
both
of
our
evaluations
of
the
methods
comes
around
a
demographic
component
that
leads
into
that
migration
and
fertility
question.
So
currently
we're
we're
utilizing
different
data
sources
and
methods
and
both
are
appropriate
for
the
methods,
but
there's
some
links
that
we
can
develop.
Residential
housing
development
is
an
important
link
between
the
two
and
and
that,
hopefully,
is
obvious
that
if
you
don't
aren't
monitoring
that
residential
development
and
its
translation
into
children
coming
into
school,
you're
missing
the
component
there.
A
And
that
leads
to
point
number
three,
which
is
really
the
key
here:
integrating
a
demographic
component
to
the
methodology
that
helps.
You
understand
that
migration
factor
what
it
means
for
changes
in
the
age
distribution
and
what
that
means
in
terms
of
changes
in
the
number
of
women
of
childbearing
age
and
then
the
number
of
births
being
generated
is
where
you
kind
of
maximize
the
two
methods
and
bring
the
two
of
them
together.
B
I
think
for
the
the
main,
the
last
one
I
think
I've
hit
that
one
a
few
times
with
the
fact
that
we
need
to
compute
births
in
a
different
way
and
way
we
can
do
that
is
by
partnering
with
the
county
to
get
the
number
of
women
at
different
years
going
forward.
I
think
that
would
be,
but
I
think
the
tricky
part
is
going
to
be
that
twenty-something
group,
but
you
already
mentioned
that's
that's
going
to
be
interesting.
A
It
it
truly
is,
and
it's
you
know,
one
of
the
things
that
I'd
spent
a
fair
amount
of
time
looking
at,
because
I
was
frankly
kind
of
confused.
I
have
a
cohort
component
method
and
whenever
I
generated
that
migration
pattern
that
you
saw
with
that
high
peak
in
the
early
20s
and
I
threw
that
into
a
projections
model.
Well,
you
know,
40
years
out,
I've
got
this
huge
number
of
65
and
over
population
that
just
seems
totally
unreasonable
and
so
figuring
out.
A
What's
going
on
in
there
I
think
is
just
a
real,
critical
piece
and
the
way
the
method
works
is
simply
migrate.
People
generate
women
five
years
from
now.
How
many
of
them
are
childbearing
age?
You
apply
a
pattern
of
fertility
to
them.
You
generate
a
number
of
births.
You
move
another
five
years
and
do
the
same
thing
so
being
able
to
monitor
that
change
in
fertility
and
change
in
births
is
really
where
you
can
feed
back
into
five
years.
From
now
how
many
of
those
growths
are
going
to
be
coming
back
into
school,
see.
B
I
think
the
thing
we
don't
know
and
that
we
need
more
research
is,
for
instance,
in
in
Manhattan,
the
20-somethings
don't
stay
once
they
get
married
and
they
have
kids,
they
flee
to
the
suburbs.
Now,
if,
if
they
do
stay
here,
your
birth
rates
are,
you
know
they're
going
to
go
through
the
roof,
but
that's
the
question
you
know:
are
they
just
staying
or
are
they
going
to
go?
I
mean
that's
what.
A
If
you
can
imagine
that
graph,
that's
in
the
report
and
that
really
high
peak
in
the
20s
and
then
it
comes
down
and
is
pretty
stable,
just
just
slightly
negative.
If
you
look
at
that
graph
for
Manhattan
has
that
same
peak
in
the
twenties
and
then
what
happens
in
the
30s
boom.
It's
that
low
in
the
negative
direction
so
they're
coming
into
manhattan
and
then
they're
leaving
and
they
aren't
staying
around
to
have
have
births
they're
moving
to
New
Jersey.
A
You
know,
hudson
valley
and-
and
the
question
is
what's
going
on
here
in
Arlington,
because
that's
not
showing
up
they're
coming
in,
but
they
don't
seem
to
be
leaving
in
droves.
But
I
can't
imagine
that
they're
all
staying
and
that's
why
I
lead
towards
what
component
of
that
large
increase
is.
Actually
a
college
population
and
college
population
simply
behave
differently.
They
aren't
subject
to
the
same
fertility
rates.
They
aren't
subject
to
the
same
migration
rates.
We're
probably
going
to
long
are
we.