Transcript Dickey_Riberas

Do Economic and Demographic
Characteristics Differ between Web
and Mail Respondents to the 2005
Census of Agriculture Content Test?
By Nancy J. Dickey
Zulma T. Riberas
National Agricultural Statistical Service
(NASS) Census of Agriculture


Census is conducted every 5 years
Prior to the census a content test is
conducted
Why conduct a Census Content Test?
To evaluate:
• Format and design of the instrument
• New content items
• Changes to question wording
• Procedural changes
Why conduct a Census Content Test?

Results are used to identify modifications to
incorporate into the final design of the data
collection tool
What’s New for Census 2007



NASS was required to provide respondents
the option of completing the census on the
Web
Two paper versions of the questionnaire, a
long form of 24 pages and a short form of 12
pages, and one Web version
Short form used a open table format to
collect data. Respondents wrote in their
crops and/or livestock information
SAMPLE

27,900 cases
–
–
14,995 long form
12,905 short form
Additional Sample


An additional sample of 200 cases received
a pre-survey letter via mail encouraging them
to complete the survey on the Web
These 200 were selected from a population
of respondents who had previously reported
an agricultural survey using a Web
instrument
Survey Details



Initial mail-out was on December 30, 2005
Postcard/reminder was sent on January 17,
2006
Non-response replacement questionnaire
was sent to all non-respondents on February
9, 2006
Survey Details - Continued



Forms were checked-in, scanned for image,
and data were keyed from the electronic
image
Records were not edited for consistency or
missing data
There was only one Web instrument which
was similar in content and format to the long
form paper version
Survey Results



618 respondents submitted their responses
using the Web instrument
12,596 respondents submitted their
responses using the mail version of the
instrument
From the sample of 200, 61 responded via
Web and 91 responded via mail
So, based on the 2005 Census of
Agriculture Content Test did Mail
Respondents differ from Web
Respondents economically or
demographically?
Size of Farm


Average size of the farm for Web
respondents was 1,157 acres. The median
size was 85 acres
Average size of the farm for mail
respondents was 804 acres. The median
size was 110 acres
Graph 1: Size of Farm
Percent
60
WEB (n = 511)
51.47
50
MAIL (n = 11,879)
47.39
40
30
25.24
18.81
20
16.54
10.17
8.48
10
6.07
5.58
3.72
3.33 3.2
300 - 399
400 - 499
0
0 - 99
100 - 199
200 - 299
Acres
500 +
Have Internet access?
Respondents were asked the following
question:
At any time during 2005, did this operation
have Internet access?
___ YES
___ NO

Have Internet access
Use of a Computer for the Farm
Business ?
Respondents were asked the following
question:
At any time during 2005, did this operation use
a computer for the farm business?
___ YES
___ NO

Use of a Computer for the Farm
Business
Age of the Principal Operator


Average age of Web respondents is 52 years
Average age of Mail respondents is 58 years
Age of Operator
Age of Principal Operator
40
WEB (n = 508)
36.81
MAIL (n = 11,012)
35
30.72
30
27.15
28.54
27.97
Percent
25
20
16.53
15
11.22
10.85
10
6.3
5
2.65
0.59 0.65
0
< 25
25 - 34
35 - 44
45 - 54
Age Range
55 - 64
65 +
Other Economic and Demographic
Comparisons




Total Household Income
Type of Organization
Sex of Principal Operator
Spanish, Hispanic, or Latino Origin
Observations

There was no relationship between
household income, type of organization, sex
of principal operator, and whether the
respondent was of Spanish, Hispanic, or
Latino origin and the mode of data collection
Missing Data




There was a large amount of missing data for
some of the items
Early in the data collection process it was
discovered that some items in the web
instrument were not being captured properly
These problems were fixed as they were
discovered
This was the first use of the Web instrument
Missing Data – Type of Organization




52.91 percent of Web respondents did not
answer this question
8.58 percent of mail respondents did not
answer this question
Question was at the end of the form
Respondents may have quit the Web survey
before reaching this question
Internet Access



For Web respondents, 85.76% did not
answer this question
For Mail respondents, 6.80% did not respond
Finding is inconsistent since Web
respondents were filling the form out on the
Web
What went wrong?

Respondent was asked if the operation had
access to the internet. Maybe the operator
has home internet access but not specifically
for the farm. So it could be the wording of
the questionnaire
Summary

Analysis was based on the hypothesis that
Web respondents differ from mail
respondents. We expected Web
respondents to be younger, male, and less
racially diverse than mail respondents. We
expected Web respondents to have larger
operations, and greater income.
What now?


Our findings indicate that age, race, size of
farm operation, internet access, and
computer use for farm business seem to
have a relationship to the mode of data
collection chosen by the respondent
More research is needed to determine the
nature of the relationship between these
variables


It was disturbing to notice the high item nonresponse
These findings stress the importance of
testing the Web instrument before
respondents are given the opportunity to use
it
2007 Census of Agriculture


The same analysis will be done using data
from the 2007 Census of Agriculture
Knowing if certain characteristics of the
population will report via the Web can result
in cost savings, better data quality, and
increased survey efficiency