 |
Hang Ups and Beeps: Random Assignment, Non-Random Assignment, and Selection Effects
| |
| | Unformatted Document Text:
1
H
ANG UPS AND BEEPS
:
R
ANDOM
A
SSIGNMENT
,
N
ON
-R
ANDOM
A
SSIGNMENT
,
AND
S
ELECTION EFFECTS
John E. McNulty
## email not listed ##
Department of Political Science
Binghamton University (SUNY)
A
BSTRACT
: In the 2002 general election a field experiment was conducted in San Francisco.
5% of the call list for a GOTV mobilization drive was segregated into a control group; the rest were called as planned. Following the election, a comparison of the treatment and control groups revealed that the GOTV calls had virtually no effect on voter turnout. However, not everyone in the treatment group actually received the treatment – i.e., answering the telephone and listening to the message. Looking at only successfully completed calls, there was an apparent large effect, because these people voted at much higher rates than either the control group or members of the treatment group who did not receive the call. This is, of course, a selection effect – it is why we randomize. People who are available and willing to receive a GOTV call are different from people who are either not available or not willing. Indeed, there are even differences among those people who did not receive the call. I have data on the disposition of each call. While people who were successfully contacted voted at the highest rates, this is not the only distinction. The unwilling people vote at the next highest rate – people who answered the phone but hung up on the caller. Other distinctions exist as well. Among people who apparently were not home, people who had answering machines voted at higher rates than people who did not. This would be unsurprising if the callers left messages, but in fact they did not. There was also a very large distinction between wrong numbers and disconnected lines. This paper provides these data and explains some of the underlying theoretical reasons for these distinctions. It explains what the selection effect is, why it exists, and how experimental controls deal with it. It also discusses the methodological reasoning for analyzing data from both a random and non-random perspective to maximally leverage the information at one’s disposal. Paper prepared for delivery at the 2005 Annual Meeting of the American Political Science Association, Washington, DC, September 1-4, 2005.
|
| |
| |
|
|
1
H
ANG UPS AND BEEPS
:
R
ANDOM
A
SSIGNMENT
,
N
ON
-R
ANDOM
A
SSIGNMENT
,
AND
S
ELECTION EFFECTS
John E. McNulty
## email not listed ##
Department of Political Science
Binghamton University (SUNY)
A
BSTRACT
: In the 2002 general election a field experiment was conducted in San Francisco.
5% of the call list for a GOTV mobilization drive was segregated into a control group; the rest
were called as planned. Following the election, a comparison of the treatment and control
groups revealed that the GOTV calls had virtually no effect on voter turnout. However, not
everyone in the treatment group actually received the treatment – i.e., answering the telephone
and listening to the message. Looking at only successfully completed calls, there was an
apparent large effect, because these people voted at much higher rates than either the control
group or members of the treatment group who did not receive the call. This is, of course, a
selection effect – it is why we randomize. People who are available and willing to receive a
GOTV call are different from people who are either not available or not willing. Indeed, there
are even differences among those people who did not receive the call. I have data on the
disposition of each call. While people who were successfully contacted voted at the highest
rates, this is not the only distinction. The unwilling people vote at the next highest rate – people
who answered the phone but hung up on the caller. Other distinctions exist as well. Among
people who apparently were not home, people who had answering machines voted at higher
rates than people who did not. This would be unsurprising if the callers left messages, but in fact
they did not. There was also a very large distinction between wrong numbers and disconnected
lines. This paper provides these data and explains some of the underlying theoretical reasons
for these distinctions. It explains what the selection effect is, why it exists, and how experimental
controls deal with it. It also discusses the methodological reasoning for analyzing data from
both a random and non-random perspective to maximally leverage the information at one’s
disposal.
Paper prepared for delivery at the 2005 Annual Meeting of the American Political Science
Association, Washington, DC, September 1-4, 2005.
|
|
 Convention | | Convention is an application service for managing large or small academic conferences, annual meetings, and other types of events! | | Submission - Custom fields, multiple submission types, tracks, audio visual, multiple upload formats, automatic conversion to pdf. | | Review - Peer Review, Bulk reviewer assignment, bulk emails, ranking, z-score statistics, and multiple worksheets! | | Reports - Many standard and custom reports generated while you wait. Print programs with participant indexes, event grids, and more! | | Scheduling - Flexible and convenient grid scheduling within rooms and buildings. Conflict checking and advanced filtering. | | Communication - Bulk email tools to help your administrators send reminders and responses. Use form letters, a message center, and much more! | | Management - Search tools, duplicate people management, editing tools, submission transfers, many tools to manage a variety of conference management headaches! | | Click here for more information. |
|
|
|
| |
|
|
|
Page 1 of 21 
