Anda di halaman 1dari 30

Electoral Vote Targeting

The Tarrance Group Targeting Grid (TG)2

B.J. Martino
Vice President
The Tarrance Group

201 North Union Street, Suite 410

Alexandria, VA 22314
The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

An Introduction to Electoral Vote Targeting

“Targeting” is a frequently-used term in political campaign, encompassing a number of

related concepts. In a larger sense, targeting is about identifying your vote and your
potential vote, so you can talk directly to voters about specific things that will move and
motivate them. There are many ways to go about this.

Electoral Vote Targeting differs from other types in that it is geographically based and
relies strictly on past election results in order to gauge future performance. Another
different (and pleasant) thing about this type of targeting is that anyone with a computer
and a little time can do it for next to nothing.

Strictly speaking, this type of targeting is the use of aggregate election data (drawn
primarily from previous “like” elections) in order to estimate future voter behavior, set
hard vote goals for the campaign, and select key regions for campaign efforts (persuasion
and turnout).

It is not a replacement for other forms of research, list development, or analysis key to
campaign planning, but it is an important way campaigns can learn to use their resources
most efficiently. In fact, precinct analysis can greatly complement these other methods.

This analysis is rooted in the idea that there is still a great deal of consistency among
groups of voters. While any one voter’s behavior may be less predictable, the total or
aggregate behavior of voters tends to be more consistent.

Thus, past voting behavior can be an excellent indicator of future voting behavior. More
importantly, it allows your campaign to know which precincts to target with persuasion
and turnout efforts, and which to engage in more limited contact with.

While there are numerous sources from which to learn the basics of electoral vote
targeting, this lesson on The Tarrance Group Targeting Grid (TG)2 will provide you with
all the formulae necessary to make the calculations in a Microsoft Excel spreadsheet,
along with additional explanations of how to take the data further.

The History of The Tarrance Group Targeting Grid

The (TG)2 system, used in the mid to late 1980’s at the National Republican
Congressional Committee by then Research Director Brian Tringali, took a significant
step forward from traditional vote goal analyses by borrowing market segmentation
analysis from the business world and applying it to politics.

Not only did the system measure Republican strength in each precinct of each district, but
also a “ticket-splitter” percentage that was used to gauge the relative persuadability of the
voters in each precinct. Each precinct was assigned to one of nine segments based on
average Republican performance and ticket-splitting. Campaigns not only had a sense of

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

how many voters would probably show up on Election Day in each precinct, but very
clear guidelines on how to reach that goal. Over a decade ago, Tringali brought this
system with him to The Tarrance Group, and the system has undergone further
refinements and adaptations since then.

Back in the 1980s, the calculations were either done by hand/calculator or fed into a
mainframe computer. The gap between computer operator, analyst, and campaign
manager was pretty wide. The operator had no concept of the strategic needs of the
campaign, and the campaign manager did not fully understand the numbers that were
being crunched.

Nowadays, there is not a campaign manager who doesn’t launch Microsoft Excel from
time-to-time in order to revise budget numbers or review their media consultant’s buy.
The gap has narrowed almost entirely, while the storage capacity and computing power
of a roomful of circa-1980’s mainframes are in the laptop nestled in the manager’s

A second problem, that is largely becoming a thing of the past, is uneven electoral
reporting. With each cycle, more states are collecting and storing the precinct returns
electronically, and offering them freely on the Secretary of State’s or Election Board’s
website. There are, however, some states that still leave it up to the intrepid data-ogre to
go from county to county, practically begging to rifle through barely legible computer
dot-matrix print-outs, containing the only existing record of precinct-level election

Thus, the system that was once available to only professionals is readily available to all
political professionals with just a little guidance on the home computer.

Democratic Targeting Efforts and The NCEC

While Republicans have traditionally allowed their campaigns to maintain more of an

entrepreneurial spirit when it comes to precinct targeting, Democrats have been providing
their campaigns with highly detailed targeting data for years. The National Committee
for an Effective Congress (NCEC), a Washington DC-based PAC started in 1948 by
Eleanor Roosevelt, has built itself into the definitive Democratic clearinghouse for
precinct targeting information. Rather than simply writing a Democratic campaign a
check, the NCEC gives each campaign a report, as well as the training to use it.

They define themselves as a “full-service political committee,” providing Democratic

campaigns with “Electoral precinct targeting, Demographic precinct targeting, Voter
profile analysis, Get-out-the-vote plans, Candidate scheduling plans, Media market
analyses, Polling sample selections, [and] Vote goal analyses.”

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

In the November 20, 2000 edition of The New Republic, an article about Democratic
GOTV efforts entitled “Knock and Drag” highlighted the NCEC’s importance to
Democratic races:

“…We start with this," Thomas says, tossing me a thick report titled "Electoral Targeting With
Vote Goals." It comes from a Washington-based organization called the National Committee for
an Effective Congress (NCEC), a little-known, left-leaning operation that provides the Democratic
Party and labor unions with electoral data. NCEC's state reports include vital information about
the voting history and trends of every precinct.

In addition, they provide Thomas with a special precinct-by-precinct report on African American
and Latino voting patterns. Using the two reports, Thomas decides which precincts to apply the
base turnout model to; then, she develops a vote goal for each of those precincts on Election Day.
"This is their playbook, their bible," she says… Once targeted districts are identified, Thomas
begins a pre-program consisting of direct mail, phone calls, and visits to voters' homes.”

Contribution records from for the 2006 cycle alone list 221 Democratic
candidates for U.S. Congress who received NCEC data, as well as 24 Democratic
candidates for the U.S. Senate.

Conducting a (TG)2 Analysis

There are four basic steps to conducting a (TG)2 analysis.

1. Collect and clean the data

2. Choose your races and set your goals
3. Segment your precincts
4. Produce your report

The following pages walk you through the steps. Different typeface is used to indicate
different portions of the explanation.

¾ Regular type indicates a discussion of the theory and application of the (TG) 2.
¾ Italic type indicates a focus on an example in a hypothetical district.
¾ [Italic type in brackets] indicates general instructions for manipulating the data in
Microsoft Excel.
¾ [Italic bold-face type in brackets] are specific formulae in Excel to be entered or
specific references to Excel row, columns, or functions.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Collecting and Cleaning Data

This process ranges anywhere from piece-of-cake easy to piece-of-shoe-leather tough.

As noted, there are as many data collection and storage methods as there are states, and in
fact there are more. Incoming Secretaries of State often like to put their mark on things,
and one of the ways is to change the way in which election results are reported. And as
with all things Internetish, the data can vanish without a trace from the site for no good

A few rules for collecting and cleaning data:

• Find A Data-Ogre. Somewhere there is someone geeky enough to spend some time
hunting around online for the data you need, downloading it, pasting it into Excel,
and then manipulating it. That someone might be you, but it might not be.

• Collect and Keep Everything. Your campaign didn’t pay for broadband access and a
computer with a two-hundred gig hard-drive so that you could use it to manage music
on your Ipod. Use it to grab absolutely everything you can find on the election
website. Sure, you may not need it today, but it might not be there tomorrow.

• Paste It All Into Excel. First, I would like to thank the Microsoft Corporation for
dominating the universe. Second, just about everybody has a PC, and just less than
about everybody runs the Microsoft Office suite with Excel. Since we all use the
same spreadsheet, we can all communicate and share the data.

• Make a copy of the file containing all the data, and save it under a different name.
This is the file you will actually mess around with. This way if and when you screw
something up at two in the morning, you won’t have to put on another pot of coffee
and start entirely from scratch.

• Make Sure Your Precincts Match. In regions of high growth or rapid decline, or for
other reasons, precincts tend to be split or merged from election to election. This can
be troublesome. Often, it will take a call to a county board of elections to ask how the
lines were changed. Sometimes, simply looking at the data will guide you.

o New precincts. In the example below, Blank County has two precincts reporting
in 2004. In 2006, there are three precincts. A quick examination of the
registration from those years makes it pretty clear that Precinct B was split in half
for the 2006 elections.

2004 Reg 2006 Reg

Precinct A 500 Precinct A 505
Precinct B 1,000 Precinct B 506
Precinct C 504

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Any raw election returns from Precinct B in 2004 will have to be divided by two
(600 turnout and 300 votes for the Republican in Precinct B in 2004 will be
divided to 300 turnout and 150 for each), thus estimating what Precinct C would
have been like, had it existed as a separate entity in 2004. This is not a perfect
solution, but a workable estimate.

o Merged precincts. In another case, example below, Nowhere County has seen
some budget cuts and reduced the number of precincts from three to two. Once
again, looking at the datasheet will clue you in.

2004 Reg 2006 Reg

Precinct A 500 Precinct A 499
Precinct B 400 Precinct B 696
Precinct C 300

All of the raw elections returns from Precincts B and C in 2004 will have to be
added together, thus precisely calculating what the new combined Precinct B
would have been like.

o Sometimes, only a portion of a precinct is in your district. You will need to know
what percentage of the precinct is covered by your district, and use that number to
multiply all of your registration, turnout and raw election returns by.

For instance, Precinct A has 500 registered voters, but your district only covers
25% of the precinct. You will have to correct that precinct to estimate 125
registered voters (500 times 25%), and correct all the other raw number

o Absentee/early ballots can be a pain to track. Some counties merge the absentee
numbers in with each precinct’s returns, others simply report a total absentee
count by county or race within the county.

If the numbers are reported for the entire county, and your district does not
encompass the entire county, you will need to make a guesstimate (worse than an
estimate) as to how many absentee ballots will come from your portion of the
county, and how many votes you would expect to get (using the same kind of
math you did for any precincts that were partially in your district).

o Those are very clear, easy examples. Reality, as is often the case, can be slightly
more complicated than computer models. Use your best judgment when matching
precincts, but always save a back-up copy in case something goes wrong, and
don’t hesitate to call the county boards of elections and ask their advice.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Choose Your Races and Set Your Goals

Now is the portion of the analysis that is plain old vanilla precinct targeting. If you want
a longer, more detailed description of the basics, the chapter on “Prior Electoral
Targeting” in Daniel Shea’s Campaign Craft is a good place to start.

I will try to sum it up as painlessly as possible, while including all the Excel equations,
then focus more on the segmentation analysis.

The four steps of standard vote-goal targeting are described below:

1. Choose three or more statewide general election races whose average Republican
score is between 52% and 55% within the district in question. This is a good
minimum winning vote goal. They should be pulled from competitive statewide
races (not necessarily close, but no chump races where its only the popular
Republican incumbent versus the perennial Libertarian or Green candidate), and
you should try to use mostly races from “like” years.

The latter means that to set goals for your campaign in 2008, you would primarily
use 2004 statewide races as your guide. However, you may have to pull from
2002 and 2006 in order to set a proper vote goal.

Of the three competitive statewide races, mostly from a “like” year, that you pull:
one should be of a high Republican performer (representing the best a Republican
can do), another should be a low Republican performer (representing the worst a
Republican can do), and others in the middle.

In the fictional 2008 example that runs throughout this article, the three election
results chosen are from statewide candidates Corleone and Barzini in 2004, and
candidate Sollozzo in 2006. All good Republicans, and while Sollozzo lost the
district with just 42%, Barzini won with 52%, and Corleone garnered 62%. Their
average, therefore, is 52%.

2. Estimate what turnout will be in your district in that year. There are a number of
ways to do this. Traditionally, you take the average turnout percentage of the past
couple “like” election years and multiply that by current registration. One
problem with that method can be that turnout might be trending up or down, and
an average is not appropriate. Another problem is that you may not catch that the
state has purged the registration lists, and taking a simple percentage of a smaller
but truer count of registered voters could cause you to underestimate.

Another way to estimate turnout is to track the trend of raw turnout in the past like
elections, while also paying attention to the change from the last two non-like
elections. (In 2006, this method proved very useful, as the spike in turnout in
2004 was echoed in turnout increased in many places in 2006).

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Whatever turnout number you decide upon, and whatever method you choose for
determining it, you then need to apply that method to each precinct.

Registration for 2008 is estimated to be 10,000 voters in our fictional district, and
turnout in the district overall in 2004 was 50%. Knowing that no list purges have
gone on, and feeling that the 2004 turnout percentage is an accurate estimate for
2008, we would calculate the estimate that 5000 voters will turnout on Election

Once we have decided on our district-wide model for estimating turnout (the 2004
race), we apply it district by district. Although 50% districtwide, turnout varied
by individual precinct. We use the spreadsheet to multiple turnout percentage
from 2004 with our registration by precinct (in this example, registration starts
off equal in each of the precincts).

[In Excel, the formula you would

type to create the “2008 t/o
estimate” value for Precinct A of
Blank County would be =C2*D2.
Simply copying and pasting that
formula down the rest of the
column (through Precinct E of
Lake County), will replicate the
formula for the appropriate row.
Then use the sum function Σ on
the toolbar at the bottom of that
column to total up your estimated
turnout for the entire district.]

In Precinct C of Blank County,

there are 1,000 registered voters,
but turnout was only 30%, thus
we estimate 300 voters will
turnout in 2008. Conversely, Precinct B of Lake County also has 1,000 registered
voters, but had fully 90% turnout in 200, for an estimate of 900 voters to turnout
in 2008.

3. Calculate the vote goal percentage, base, high, and ticket-splitting percentages.
Remember the three or more races you picked to achieve a 52% to 55% average
victory? Average these races for each precinct in your district, creating an
“Average Republican Performance” percentage. Then calculate the “base”
percentage in another column (Excel has a handy formula that will do this for
you, discussed in the example), which registers the worst a Republican has done
in each precinct. Also calculate the “high” percentage, which measures the best a

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Republican has done in that precinct. Finally, subtract the high from the base, and
that is your “ticket-splitting” percent.

Let’s take a moment to discuss what these measures mean, and why they are
important. “Average Republican Performance” is a measure of the general
Republican strength in that precinct. A Precinct with an ARP of 65% is better for
Republicans than one with an ARP of 35%. Pretty easy.

The “base” percentage tells you what you are starting with, that is the percentage
of voters you have pretty much automatically, and the “high” tells you what the
best you can reasonably hope to do it. Their difference, the “ticket-splitting”
(T/S) percentage is therefore a measure of persuadability. There are more
persuadable voters in a precinct with a high of 70% and a base of 50% (20%
ticket-splitting) than in an equal sized precinct with a high of 62% and a low of
58% (4% ticket-splitting).

That last example highlights why ticket-splitting is important. While the precinct
with a high/low of 70%/50% and the one with 62%/58% may have about the same
ARP score (excluding any other races used), at 60%, they are fundamentally
different types of precincts. While running a strong campaign at the first precinct
(20% ticket-split) has had a measurable impact in the past, voters in the second
precinct (4%) don’t seem to change their minds a lot. Thus a campaign would be
better suited to target more persuasion at the first, and focus more on ID and
turnout solely in the latter.

Now we have added many more columns to our spreadsheet. Remember our good
Republican candidates Corleone, Barzini and Sollozzo? We insert the results
from their races into the spreadsheet, and calculate ARP, base, high, and ticket-
splitting percentages.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

[In Excel, type the following formula in the “ARP” column =(F2+G2+H2)/3 and
the average for the three races will be calculated. To calculate base, type
=MIN(F2,G2,H2). For the high, type =MAX(F2,G2,H2). And for the “T/S,” use
=L2-K2. Why have I placed the T/S column to the left of the ARP column, when
T/S is the last thing to be calculated? It will prove useful later on when we are
preparing the segmentation portion of the analysis]

The ARP scores per precinct range from a high of 63% (in Precinct E of Lake
County), to 45% (in Precinct A of Nowhere County). In this case, the “base”
scores are all of the Sollozzo scores for each precinct, however this is not
necessarily the case in practice, which is why we ask the spreadsheet to find the
MIN rather than just apply the lowest race across the board. The “high” scores
are primarily Corleone’s, although Barzini tied Corleone in Precint B of Blank
County and overperformed him in Precinct A of Lake County. The MAX formula
picks up on this.

Finally, note the ARP and T/S for Precincts A and B of Blank County. While both
have an ARP of 57.7% and would thus be considered strong Republican
precincts, voters in Precinct A are much more likely to ticket-split than those in
Precinct B. Thus, more persuasive messaging should be targeted towards
Precinct A.

4. Next multiply the ARP, base and T/S percentage by the estimated turnout for each
precinct. This will return your estimated raw vote goal, the raw base vote you can
count on, and give you a count of how many persuadable voters there are in the
precinct (from which you have to add to your base until you reach the vote goal).

Ignoring the component race columns F, G and H so that the spreadsheet will fit
more legibly on the page, we calculate our vote goal, base votes and ticket-
splitters for our race.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

[In Excel, our vote goal for the first precinct is calculated with the formula
=TRUNC(E2*J2), the number of base voters with the formula =TRUNC(E2*K2),
and the number of ticket-splitters with the formula =TRUNC(E2*I2). Then again
use the sum function Σ to total the vote goals base voters and ticket-splitters. The
“Trunc” function removes decimal places, so that your total count isn’t higher
than the true sum of the whole numbers for each precinct]

The spreadsheet returns a total of 2,600 votes needed to reach 52% on the ballot,
given our estimated turnout of 5,000. 2,100 of those votes are base Republican
voters who the campaign simply needs to identify and turn out. In order to garner
the additional 500 votes necessary to reach the goal, the campaign must identify
the ticket-splitters (of which there are just over 1,000 in total).

Note that total ticket-splitting percent [calculated with the formula =O12/E12] is
0.2% higher than what we would calculate it to be if we had simply subtracted the
overall Corleone from the overall Sollozzo, reflecting the fact that Barzini slightly
outperformed Corleone in one precinct.

Traditional precinct vote goal models stop just about there. Very useful stuff. Each
precinct captain can now be handed this information along with the list, to track down
base voters and ticket-splitters, targeting each with the appropriate messages and contact

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Segment Your Precincts

Now for the application of the market segmentation analysis. Using each precinct’s ARP
and T/S percent score, you are going to place that precinct into one of nine categories.
For each precinct you select, Chinese-menu-style, one descriptor from column A and one
from column B.

Column A Column B
High Republican Performing High Ticket-Splitting
Middle Republican Performing Medium Ticket-Splitting
Low Republican Performing Low Ticket-Splitting

This is best not left to guesswork. Using additional formulas in the Excel spreadsheet,
thresholds for high, medium and low ranges in each category can be set, and moved until
a relatively even distribution is achieved.

First of all, it may help to envision the precincts in your district arranged in a scatterplot,
with the positions along the vertical Y-axis determined by the precincts’ Average
Republican Performance scores, and the positions along the horizontal X-axis determined
by the Ticket-Splitting percentage score.

In our example district, the ten

precincts are arranged as on the figure
to the right. Precinct E of Lake
County, for example, is the only ARP
precinct with an ARP exceeding 60%,
so it’s near the top of the scatterplot,
and with a T/S score of 24%, is placed
right of center.

[To create this scatterplot in Excel,

create a highlighted box over all
precinct values in the T/S and ARP
columns (as shown below). Then find T/S
the button at the top of the page that
looks like a small bar chart, the “chart wizard,” and
click on it. If you cannot locate it, simply click on the
“Insert” at the top tool bar, and click on “Chart…”
(which should also have the image of the bar chart next
to it.)

When the chart wizard dialog box opens, select “XY

(Scatter)” from the list of chart options, and click
“Next.” A small view of your chart should appear in the
dialog box, click “Next” again to move on. The dialog
box should now have several tabs at the top, labeled

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

“Titles,” “Axes,” “Gridlines,” “Legend,” and “Data Labels.” Click on the “Legend”
tab, and turn off (uncheck) the “Show legend.” Click “Next” again.

You are almost done. Exciting, ain’t it? Finally, tell the dialog box to place the chart as
a new sheet in your Excel workbook, simply by clicking on the dot next to “As new
sheet:” then click “Finish.” Voila!]

Now, if you are following along this example in your own spreadsheet, you are probably
wondering why your dots are all squished together, and why mine are scattered nicely
throughout the chart. Note that I changed where the Y-axis to begins to 40%. Excel
draws the chart starting at 0%. You can format the lines to start wherever you like.

Think of the scatterplot of the precincts as the “fingerprint” of your district. Like a
fingerprint’s loops, whorls and arches, scatterplots share some patterns, but also like a
fingerprint, each one is distinct. After having seen several scatterplots, you may begin to
notice some trends.

The following charts are of some unspecified state legislative districts in the country,
with a little explanation of what the scatter is revealing.


OK, well sometimes the scatterplot

simply looks like a bug hit your
windshield, but this type of very
scattered scatterplot tells you that
the district has an assortment of all
types of precincts in terms of both
ARP and T/S.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting


A common occurrence in the

scatterplot is a trail of lower
Republican performing, lower
ticket-splitting precincts heading
off from the main body of
precincts. There are Democratic
strongholds, and even the best
Republican campaign made little
headway in these precincts.


“GOP Enclave”

65% The other side of a “Twister” is the

60% collection of hard-core Republican
55% precincts, that have less tendency to
ticket-split. Often a good place to
focus early fundraising, as well as to
work harder if facing a primary.


20% 22% 24% 26% 28% 30% 32% 34% 36% 38% 40%

Now with a sense of the district, the next step is to determine the thresholds for what we
will constitute high, medium and low Republican performing, and high, medium and low

Essentially, we are going to overlay a tic-tac-toe board over the scatterplot, and precincts
that fall within each block get a specific designation. High Republican performing
precincts are given a “1” designation, middle performing a “2,” and low performers a
“3.” Similarly, high ticket-splitting precincts are given a “C” designation, middle ticket-
splitters a “B,” and low ticket-splitters an “A.” The results are nine categories (1A, 1B,
1C, 2A, 2B, 2C, 3A, 3B, and 3C), each of which is given a name to help you remember
the type of district that the code signifies.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

The following chart (ripped off pretty shamelessly from marketing guru George Day and
others) arranges the categories graphically in a traditional market segmentation grid.

(TG)2 Segment Descriptions

Adapted from Day, “Analysis for Strategic Market Decisions” (St. Paul: West, 1986), pg 204; Abell and
Hammond, “Strategic Market Planning Problems and Analytical Approaches” (Englewood Cliffs, NJ:
Prentice Hall, 1979); And Robinson, Hitchens and Wade, “The Directional Policy Matrix: Tool for
Strategic Planning” in “Long Range Planning 11’ (1978, pgs 8-15

There are further descriptions of these categories at the end of the article. But enough
with the pretty charts, let’s get back to our spreadsheet. Some specific formulae are
necessary to set the proper thresholds and create the categories.

The “Grid” column in the spreadsheet below represents which segment the spreadsheet
places each precinct, based on the high/low thresholds for ARP and T/S that we set. Note
the “Grid” segments are not finalized, in that the thresholds represents a first guess.

[In Excel, we are going to create 14 new columns! Five will be used to set the thresholds
that determine which categories the precincts get places into, and nine will be used to aid
us in making sure we are setting the thresholds properly.

Let’s create the five first.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

The first (P) column sets the high ARP threshold. Take a guess at a percentage for a
high ARP precinct and enter it in P1. It’s just an initial guess, and we’ll change that
value later. Immediately underneath it, enter the formula =IF(J2>P$1,1,Q2). Then copy
and paste the cell down the length of the column for all precincts. This formula instructs
the spreadsheet to see if the ARP value is higher than the value in P1. If ARP is higher
than P1, the spreadsheet returns the value “1” into the cell, if not, it is told to enter
whatever value appears in Q2. The “$” in “P$1” locks the formula, regardless of where
you paste it, to make sure it references the first row. When you do a cut and paste on the
cell below your next formula, you’ll notice the formula automatically changes to
=IF(J3>P$1,1,Q3). Knowing you had moved one cell down, it changes the references to
the J and Q columns one cell down, while keeping the P reference locked in place.

The next (Q) column sets the low ARP threshold. Take another guess at a low threshold
and enter it into Q1. We’ll change this later. Then under it, enter the formula
=IF(J2<Q$1,3,2), then copy it down the length of the column. This formula has the
spreadsheet check if ARP is lower than the value in Q1, which only matters if the P
column has already found that the ARP value is lower than P1. If ARP is lower than Q1,
the cell returns the value “3,” and if it is higher, it returns the value “2.”

Once both columns are done, the P column should now represent whether the precinct is
high, medium or low performing, based on the guess thresholds you entered in P1 and
Q1. We’ll come back to these later.

The third (R) column sets the high T/S threshold. Take a guess at a high T/S just to start,
and enter it into R1. Underneath, enter the formula =IF(I2>R$1,“C”,S2), then copy and
paste the cell down the column. The formula checks if T/S is greater than the guess value
in R1, and if so returns the letter “C” in the cell. If not, it is told to check the value in the
adjoining S column.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

The fourth (S) column sets the low T/S threshold. Take your guess, and enter it into S1.
Underneath, enter the formula =IF(I2<S$1,“A”,“B”), and copy and paste it down the
length of the column. If the spreadsheet has not calculated the precinct to be high ticket-
splitting in column R, then this equation determines if it is to be considered medium
(“B”) or low (“A”).

Once these two columns are done, the R column should now return whether each
precinct is high, medium, or low ticket-splitting, based on the guess thresholds you
entered in R1 and S1. We will revisit these guesses later too.]

Lastly, in the “Grid” column (T), simply enter the formula =P2&R2 right under the
“Grid” label, and the cell will return a combination of its ARP and T/S number and letter
designations. Copy and paste this down the length of the column.]

You have made some initial guesses as to what may constitute a high or low Republican
performer, but what thresholds are appropriate? You want there to be enough votes in
each segment that the resulting definitions are actionable, so we have to devise a system
for roughly dividing the number of Republican votes into equal portions.

While dividing it exactly nine equal ways is next to impossible (except under rare
circumstances), we can roughly divide the number of GOP votes into thirds by ARP, and
once again divide into thirds by T/S. The spreadsheet can help you figure this out.

[In Excel, create nine new columns labeled “1A”, “1B”, “1C”, “2A”, “2B”, “2C”,
“3A”, “3B”, and “3C”.

OK, deep breath, the spreadsheet is not as scary as it looks.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Enter the following formula in column U under the “1A” heading,

=IF($T2=U$1,$M2,0). Here we are telling the spreadsheet to check to see if the “Grid”
designation in column T matches the Grid designation at the top of column U. If so, the
raw vote goal value is returned in that cell. If not, the cell gets a goose egg.

The dollar signs are important in this formula, because not only do they allow us to copy
and paste the cell formula down the rest of that column, it allows us to copy and paste it
down all nine columns!

Next, use our friend the sum function Σ in the toolbar at the bottom of each of the nine
columns to sum up the raw number of Republican votes that currently fall into that

Then, under the sum for “1A” (U), divide the sum by the total vote goal, using the
formula =U12/$M12. This will return the percentage of the total Republican vote that
comes from that category. Copy and paste that cell formula under the other eight sums
as well. These percentages should add up to 100%. (Note: the numbers may not appear
as percentages, you may have to format the cells to show percentages).

Under the percentage in “1B” (V), enter the formula =U13+V13+W13. This returns the
percent of the total Republican vote that is currently designed high Republican
performing. Copy this cell formula under the “2B” (Y), and “3B” (AB) percentages,
and the result will add up the middle and low performers as well.

In row 15, column U, enter the formula =U13+X13+AA13, which returns the percent of
the total Republican vote designated low ticket-splitting. Copy and paste the cell formula
in the two cells immediately to the right to calculate the percentage of middle and high
ticket-splitting Republican votes.]

Now, we’ve created a dynamic worksheet. Simply by changing the threshold percentages
at the top of four columns [P, Q, R, and S], the spreadsheet will do all the work in
redefining the “Grid” designation, and allow us to eyeball the percentages in each
segment and tier, until we decide we have a working model.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

[In Excel, play with the threshold values at the top of columns P, Q, R, and S. Districts
will generally have more than ten precincts, and a 1% change up or down will generally
mean a gradual rise or drop in the summary percentages in Rows 13, 14 and 15. Also,
districts with more precincts will enable you, generally, to get much closer to
33%/33%/33% even distribution. In this case, about the best I could do was set the high
ARP threshold at 57%, and the low at 48%, which divided up the GOP votes
45%/30%/25%. I was much more successful with ticket-splitters, by moving the
thresholds to 24% and 10%, I hit a distribution of 32%/35%/33%, which is very good.]

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Produce Your Report

You’ve now done all the calculations, and want to summarize your findings in a useful,
printable report that the campaign team can understand and use. Looking through all the
columns of data you have entered, you must decide which pieces of information are
necessary, and which information that, while interesting, does not contribute to the
usefulness of the report.

To prepare a printout of the spreadsheet, you want to include the precinct identifiers,
current registration, your turnout estimate, the ARP %, the raw vote goal, raw base
voters, raw ticket-splitters, and grid designation.

In addition to the spreadsheet, you’ll want to prepare a cover memo or presentation that
summarizes the process and the assumptions you made (especially concerning turnout
and races used to calculate ARP). There may be members of the campaign team who
know something you do not about why one of your assumptions is faulty, and unless they
scroll through all of your formula, the cover memo will be the only way they can spot

[In Excel, the best way to prepare the summary spreadsheet is to create a duplicate
worksheet, copy the whole sheet, and “paste special” as “values,” then eliminate the
unnecessary columns.

You do this by first

right-clicking on the tab
on which you placed all
your data and your
equations and select
“Move or Copy…”
which opens a dialog
box. Click on “Create
a copy” and click
“OK.” There is now a
duplicate worksheet in
your file with all the
same information.

In the duplicate, right click on that blank square above row 1 and to the left of column A.
Click “copy.” Right click on that same box, then click “Paste special…” which will open
a dialog box. Under the “paste” header, select “values,” then click “OK.”

Look at some of your cells now under the ARP and T/S columns. The formula are all
gone, and only the values that were returned remain in the cells. Now when you cut
away some of the columns that were used to create them, Excel won’t return an error

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Right click and delete columns U, V, W, X, Y, Z, AA, AB, and AC. Right click and delete
columns P, Q, R, S. Note that the “Grid” column (T) will slide over and fall under P
once you do this. Right click and delete K, L, I, H, G, F, and finally D. Only the
columns you see in the spreadsheet above will remain (but will be under A, B, C, D, E, F,
G, H, and I now).

Using “File” and “Page Setup…” commands, you can add page numbers to the
spreadsheet report (while this example will print on one page, most districts will not),
add other headers and footers to the page, and request that Excel print out row 1 at the
top of every printed page so the readers will know what column they are looking at.

You can also resize some of the columns. Often you will have to make the precinct name
column wider, since often a city name is part of the precinct name, and you want to be
able to exactly identify the precinct in the printout. If this pushes the worksheet wider
than the printed page, you can often tighten some of the other columns to make up for it.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Further (TG)2 Grid Descriptions

The following paragraphs provide a longer narrative description of each of the segments.
The names may be a little too cute, but they can be useful mnemonic devices. Low
ticket-splitting segments have an equestrian theme, while medium and high ticket-
splitting segments are named based on their “explosive” potential.

1A- Thoroughbred-- Voters in this segment give statewide

Republican candidates a significant margin of victory, time and
time again. Protect the natural advantage you have with them, and
spend resources turning them out on Election Day. Persuasive
messages have little use.

1B- Bottle Rocket-- The voters of these areas are capable of giving Republicans
a significant victory, but have shown the capacity to shoot a short distance (up
or down). They should both be targeted with some persuasive messaging, then
turned out.

1C- Dynamite-- These voters give Republicans significant victories on

average, but have shown an explosive potential. Target persuasive
messages to drive up the winning percentage and turn them out.

2A- Workhorse-- These voters return a decent percentage for

Republicans, although not necessarily a winning one. Voters in
these areas should be ID’d then turned out. Do not waste as
many persuasive messages here.

2B- Firecracker-- These voters can pop or fizzle, generally returning a respectable (but
not necessarily winning) percentage, with some capacity to be driven up or down.
Target well-crafted persuasive messages combined with ID driven turnout efforts to get
the biggest bang for your buck.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

2C- Old Nitro-- While averaging a decent (not necessarily winning)

percentage among those voters, Republicans must be wary of their
volatility. Target powerful persuasive messages, ID, then turnout.
Be careful, your opponent is doing exactly the same with them.

3A- Mule-- Stubborn and more Democrat, the voters of this region
may be very carefully ID’d and turned out, if the resources are
available. Any turnout should be late and focused.

3B- Dud-- Low Republican percentages are common, the

only question being how low. Defuse your opponent’s
strength here, and look carefully for your voters.

3C- Opponent’s Powder-- These areas are poor Republican

performers, with voters who possess the potential to return
significant victories for your opponent. Throw some water to reduce
its explosive potential. Target messages to drain your opponent’s
strength. You must look carefully for your voters here, as well.

Taking (TG)2 Further

Once you’ve finished applying all the formulae and have finished your vote goals and
segmentation analysis, the data may not be done serving you yet.

Is voting age population (VAP) data by precinct available? By identifying higher

Republican performing precincts, with lower-than-average percent registration, you can
give yourself a few more efficient targets for voter registration drives. (See “Voter
Registration” section)

How about precincts that are higher Republican performing, but have lower than average
turnout? In our example, Precinct E of Lake County has a 63% ARP, but only 40%
turnout (below the 50% average for the district). Of the 600 voters staying home, we
might estimate that 378 are inactive Republicans (63% times 600). Try to find unique
turnout efforts to energize those voters who are staying home.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

How about precincts that have the highest drop-off from Presidential to mid-term or off-
year elections, but are higher Republican performing. In non-Presidential years, can we
get some of those drop-off voters to turn out? In a Presidential year, if the national race
is turning out to be non-competitive, we need to focus on those high Republican drop-off
precincts to make sure they don’t sit this one out.

If in a Congressional or statewide race, rather than just examining the data under a
microscope at the precinct-level, take a step back and conduct the same analysis on the
county level. This sort of analysis is especially useful for helping to plan a candidate’s
schedule, as well as planning television and radio buys. Another Microsoft product
called MapPoint can produce color maps of ARP, T/S and the Grid segments by County,
allowing you to get a telling visual look at the entire state.

Below is an example of a map created several years ago, segmenting the state of
Tennessee by county.

If multiple or a whole slate of legislative races in a particular state need precinct targeting
analyses, it is much easier to first apply the process to the entire state. If you have a
column that defines which district each precinct falls in, you can then sort the whole state
by district, and cut and paste that portion of the state into its own worksheet.

Note that when doing this, you’ll need to reset the ARP score to a 52% - 55% winning
percentage for that district (some districts will have higher than a minimum vote goal,
others lower). You’ll also need to reset the thresholds, and thus the grid segment
designations. A precinct that might be a 2C for a statewide race, might be considered a
1A within terms of its legislative district.

Finally, occasionally there will be local races that may affect turnout one particular part
of your district. Or you may decide that your candidate’s home precinct[s] will be
stronger than the usual vote goal, or that your opponent’s home precinct[s] will be
weaker. The model will not take these kinds of things into account. Use your best

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

judgment to tweak estimated turnout and/or the vote goals up or down to suit your needs.
Make a note of those adjustments in your cover memo, so users will be aware of what
you’ve done.

If done early in the election season, when things may not be as busy, the analysis can be
used to help lay out the groundwork for the entire campaign. If you wait to produce the
analysis until you are gearing up GOTV efforts, a lot of the other benefits of its findings
may be lost.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Voter Registration

The basic voter registration targeting involves finding the highest Republican performing
precincts, and working those precincts. However, if Voting Age Population data (VAP)
is available, then the spreadsheet can be used to help identify more efficient targets.

In our district, I’ve inserted the VAP data into column C, and cleaned up the other
columns, leaving only those needed for this calculation (registration, ARP, and the Grid).
The first calculation is what percentage of the precinct is registered. [In the spreadsheet,
divide the registration by the VAP, type =D2/C2. Then subtract the Reg from the VAP to
indicate how many adults are not registered in each precinct, type =C2-D2. Finally,
estimate the number of possible Republican voters who are not registered by multiplying
the Total Not Registered by the ARP score, type =G2*F2.]

Precinct B in Blank County appears to be the first target for registration drives, followed
by Precinct C of Lake County. Both Tier 1 in the Grid, they are highest in terms of the
efficiency (registering more Republicans than Democrats), and in the raw number of
potential registrants. Note than while Precinct B of Nowhere county has 326 possible
Republican registrants, the efficiency is much lower. A blind registration drive there
will probably register slightly more Democrats than Republicans. Because there is such
a large number of Republicans there, but low efficiency, consider a drive to have
currently registered Republicans in the precinct identify and register friends.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting


You can also identify precincts that are better targets for turnout efforts. For Presidential
year elections, you can calculate the differences in turnout from past Presidential election
years. Those precincts that experience relatively larger percent changes in turnout
between elections, are more likely to be those where turnout efforts can make a
difference. In non-Presidential years, who can apply the same method across non-
Presidential years, OR analyze the change in turnout compared to the last Presidential. A
spike in turnout in a Presidential year can be a precursor to increased turnout in the
following off-year.

In this example, we have registration and turnout data from 2000 and 2004, which have
been inserted into the spreadsheet, along with 2008 registration and turnout estimate,
ARP, the Raw Vote Goal, and the Grid. If you had all this data to begin with, you could
have constructed your initial turnout estimate for 2008, by taking an average of the 2000
and 2004 %s. [Calculate the “turnout swing percent” between past Presidential
elections, calculate the difference between the 2000 and 2004 percentages, typing
=ABS(F2-D2). The ABS function calculates the absolute value of the result, so that the
difference is always a positive number, even if 2000 turnout was higher than 2004. Then
calculate an estimate of the Raw Vote Swing, that is the number of Republican votes that
could be affected by a turnout effort. Type =TRUNC(L2*J2).]

Precinct B of Lake county has a 20% swing and a potential impact of 86 votes, but is a
2C precinct. Blind turnout in this precinct is going to be a mixed bag, so more targeted
efforts are needed in this precinct. Precinct C in Lake county and Precinct B of Blank
county have the next largest potential vote impact, and being Tier 1 precincts can be
targeted for turnout more broadly.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Election Night Results Tracking

Watching returns trickle in on Election Night can be harrowing. As precincts or counties

come in, it is very helpful to have your vote goals handy, to compare the incoming
numbers with the model.

[Keeping the turnout estimate, the ARP, and the Raw Vote Goal, create blank columns
for Votes Cast and Our Vote. Create additional columns for Our Vote %, turnout versus
estimate, and Vote versus Goal. Type =G2/F2 to calculate Our Vote %, =F2-C2 for
turnout versus estimate,
and =H2-D2 for Vote
versus Goal. The cells will
return error messages
(dividing by zero) or
negative numbers at first,
but that will change as you
start inserting the elections

On Election Night, simply

insert the number of votes
cast in your race and the
number of votes cast for
your candidate in the
appropriate columns.
You can then easily
compare if turnout
is generally higher
or lower than the
model, if that
growth in turnout is
happening in better
or worse
performing GOP
precincts, and if
your candidate’s
actual vote percent
is meeting the
minimum goal. You
can also pull out the
precinct for which
you have data and
quickly calculate a

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

Post-Election Analysis

The election is over, and you stopped entering the results after 30 minutes when the A.P.
declared a winner. Once the results are certified, it may be worth your while to enter in
all the results by precinct and do one last analysis.

[In your spreadsheet, include the following columns IN THIS ORDER: Precinct, Votes
Cast, ARP, High, Base, and the Our Vote %. Make sure the precincts are ordered
geographically, by county, city and precinct: this will make identifying trends in the chart
mush easier. With your mouse, highlight the entire dataset, in this case from A1 to F12.
Then click the Chart Wizard button on the Standard Toolbar. Choose the “Stock” chart
type, and the “Volume-Open-High-Low-Close” sub-type.

The Tarrance Group Targeting Grid (TG)2 And Electoral Vote Targeting

If you click finish at that point, the following chart will generate:

For each precinct, the 1,000 80.0%

chart summarizes all 900

five pieces of data. The 800
solid bar from the 60.0%
bottom is the number of
50.0% Votes Cast
votes cast in your race, 600
with the Y-axis on the 500 40.0% High
left show. The thin line 400
30.0% Our Vote %
represents the gap from 300
the High Republican to 200

the Base, and the 10.0%

thicker bar shows the
relationship between 0 0.0%




the ARP score and your




















candidate’s percent. A
white bar means that
your candidate performed better than the vote goal %, and the black bars indicate those
precincts where your candidate performed worse.

You can also play around with the colors and the format of the chart, so it becomes a
little more user friendly. With this chart, you can see that the Republican performed
better than average
1,000 80% in Blank county,
GOP% especially Precinct
900 A, but did not meet
70% the goals in Lake
800 County, especially
Precinct C. This
60% report can provide
useful information
600 55%
for an After Action
50% Report, to see where
campaign efforts
worked, and where
40% they could be
300 improved in the next
200 30%
Blank A Blank B Blank C Nowhere Nowhere Lake A Lake B Lake C Lake D Lake E TOTAL