1 Internet Survey by ESC, NCCU prepared by Chia-hung Tsai 2012.3.21.

1

Internet Survey by ESC, NCCU

prepared by Chia-hung Tsai

2012.3.21

2

What is internet survey

Asking questions by mailing people through internet interface

It’s like mail survey; response rate is an issue (non-sampling error).

Respondents may be from telephone surveys or big databank.

Could be biased toward young, educated internet-users.

3

Some examples

Harris Interactive

YouGov

4

What is NOT internet survey

Entertainment polling: CNN(Quick vote), Yahoo…

Voluntary-panel: 優仕網Self-selected web surveys: government

5

Cost of Internet Survey

No cost of space, supervisor, interviewers, equipments, and telephone fees.

Need to collect Emails, maintain respondents, and reward respondents.

6

Advantages of internet surveys

Interesting interfaces. More elements: video, audio, and

experiments.Easy to track respondents’ attitudesNo boundary of space and time

10

Disadvantages of internet surveys

Not representative of population; young, educated, office people

Self-selection biasNot sure about who answers the

questionsNot sure about how many people will

respond.Not able to ask knowledge questions

11

2007 experimental research for RDEC

We successfully contacted 3,000 people out of more than 15,000 previous respondents through telephone.

Over 90% of people can log on internet.57% of people said they often surf the

internet.

12

(continued)

Only 1,000 people agreed to join the panel.

About 300 people responded to the survey each time.

We offered $2.5 dollars for each interview plus some rewards for loyal participants.

Spending too much to build up a panel.

male male male male

Difference in Gender,4 waves

0.0

1.0

2.0

3.0

ele. high univ.

Difference in Education, N=325

-20

010

2030

20-29 40-49 60+

Difference in Age, N=325

-10

-50

5

N1 Tao Yun Yi

Difference in areas, N=325

-10

12

34

14

Switching to New Method

We recruit people when we are doing surveys beginning from 2009.

We’re able to recruit 20-25% of respondents each survey.

About 7,000 panelists are in our pool.7 surveys have been conducted.

15

20071201 20091202 20100303 20100802

Number of Panelists Recruited from Projects

01

00

20

03

00

40

05

00

60

07

00

16

May Oct.

Last 4 Surveys' Res. Rates

%

02

46

81

01

21

4

May Oct.

Last 4 Surveys' Obs.

01

00

20

03

00

40

0

17

Male Female

Gender, N=3631

-0.4

-0.2

0.0

0.2

0.4

18

20-29 30-39 40-49 50-59 60+

Age, N=3631

-10

-50

5

19

ele. junior high college univ.

Education, N=3629

-30

-20

-10

01

02

03

04

0

20

N1 N2 Tao Chung Yun Kao Yi

Areas, N=3622

-20

24

6

21

DPP Ind KMT NP PFP TSU

Party Identification, N=3631

Diff

ere

nce

, %

-5-4

-3-2

-10

1

22

Strong Uni. Depends Ind. Non

Unification/Independence Issue, N=3631

Diff

ere

nce

, %

05

10

15

23

Taiwanese Both Chinese Non

Taiwanese identity

Diff

ere

nce

, %

-6-4

-20

24

68

24

Ma's Popularity

0-1

0 s

cale

'10.1 '10.2 '10.5 '10.10 '10.10

4.2

54

.50

4.7

55

.00

25

DPP Chairperson Tsai's Popularity

0-1

0 s

cale

'10.1 '10.2 '10.5 '10.10 '10.10

5.2

55

.50

26

New Phrase: Website

• We built facebook and website in 2011.

• Facebook: PollcracyLab

• Website: 線上調查實驗室• We announce our results on facebook and

website.

27

What can we do?

• We send out questions to around 1500 respondents about every 2 weeks.

• Each wave of survey gets 500 completions.

• Every participant can collect 20 points for every completion. 600 points can make NTD$600 7-11 gift card.

28

29

30

Ma’s popularity

31

32

33

Conclusions

It’s difficult to get low-educated, senior people.

Respondents may be more pro-blue than other surveys.

Hot issues may increase sample size.We should keep comparing internet

surveys and others.should develop better matching skills.