Twaweza.org

Addressing common concerns raised about Twaweza’s opinion poll survey

Much of the discussion in traditional media, on social media and elsewhere regarding our recently published opinion poll findings on Tanzanian politics has focussed on perceived errors in the brief. We want to correct three of the most common misunderstandings.

1. “Totals do not add up to 100%.”

It has been pointed out that several charts and tables in the brief include data that does not seem to add up to 100%. In some cases, the figures add up to 99%, 101% or 102%.

This is purely as a result of rounding the figures to the nearest whole number, a standard practice in opinion poll research. We do this because presenting figures with one or even two decimal points suggests a level of precision in the findings that can be misleading.

The examples below show how this happens, by presenting the unrounded figures alongside the rounded figures that were presented in the brief.

For Figure 2a: Are you aware of the promises made by your MP?

 

unrounded

rounded

Don't know

5.14%

5%

Yes

64.39%

64%

No

30.47%

30%

Total

100.00%

99%

For Figures 5b, 5c, 5d: If the election was held today, which political party’s candidate would you vote for? / Which political party do you feel closest to, if any?

 

2015 MP (Figure 5b)

2015 councillor (Figure 5c)

2015 affiliation (Figure 5d)

 

unrounded

rounded

unrounded

rounded

unrounded

rounded

CCM

60.4%

60%

59.9%

60%

61.9%

62%

Chadema

24.1%

24%

25.5%

26%

25.0%

25%

CUF

1.8%

2%

2.5%

3%

2.3%

2%

ACT-Wazalendo

1.4%

1%

0.6%

1%

0.7%

1%

NCCR Mageuzi

1.1%

1%

1.0%

1%

0.3%

0%

Ukawa*

2.8%

3%

2.6%

3%

2.1%

2%

Other

6.4%

6%

6.0%

6%

2.3%

2%

Don't know / no response

2.0%

2%

1.9%

2%

5.3%

5%

Total

100.0%

99%

100.0%

102%

100.0%

99%

For Figure 6b: Will Ukawa appear on the ballot paper during the 2015 elections?

 

unrounded

rounded

Don’t know

17.6%

18%

No

25.7%

26%

Yes

56.7%

57%

Total

100.0%

101%

2. “Totals for 2013 and 2014 add up to much less than 100%.”

In figures 5a-5d, some totals for years 2013 and 2014 add up to considerably less than 100%. This is a result of some responses being left off the chart because in this years’ survey nobody gave those responses. In particular, the response “I don’t vote for parties, I vote for candidates” was popular in previous years, but it was not mentioned in 2015.

The complete numbers for these four charts are presented below. Figures that were not included in the brief are presented in italics. For 2015, both unrounded and rounded figures are shown.

For Figure 5a: If the election was held today, which political party’s candidate would you vote for? President

 

2012

2013

2014

2015

Unrounded

Rounded

CCM

61%

47%

51%

65.8%

66%

Chadema

30%

30%

23%

21.5%

22%

CUF

4%

4%

4%

0.5%

1%

ACT-Wazalendo

0%

0%

0%

0.2%

0%

NCCR Mageuzi

0%

1%

1%

0.2%

0%

Ukawa*

0%

0%

0%

3.3%

3%

Other

1%

2%

0%

6.3%

6%

Don't know / no response

4%

0%

4%

2.2%

2%

I don't vote for parties but candidates

0%

15%

16%

0.0%

0%

None

0%

1%

1%

0.0%

0%

Total

100%

100%

100%

100.0%

100%

For Figure 5b: If the election was held today, which political party’s candidate would you vote for? Member of Parliament

 

2012

2013

2014

2015

Unrounded

Rounded

CCM

60%

44%

46%

60.4%

60%

Chadema

31%

31%

24%

24.1%

24%

CUF

4%

4%

4%

1.8%

2%

ACT-Wazalendo

0%

0%

0%

1.4%

1%

NCCR Mageuzi

0%

1%

1%

1.1%

1%

Ukawa*

0%

0%

0%

2.8%

3%

Other

0%

2%

1%

6.4%

6%

Don't know / no response

5%

0%

5%

2.0%

2%

I don't vote for parties but candidates

0%

17%

18%

0.0%

0%

None

0%

1%

1%

0.0%

0%

Total

100%

100%

100%

100.0%

99%

For Figure 5c: If the election was held today, which political party’s candidate would you vote for? Councillor

 

2012

2013

2014

2015

Unrounded

Rounded

CCM

61%

45%

47%

59.9%

60%

Chadema

29%

30%

23%

25.5%

26%

CUF

4%

4%

4%

2.5%

3%

ACT-Wazalendo

0%

0%

0%

0.6%

1%

NCCR Mageuzi

0%

1%

1%

1.0%

1%

Ukawa*

0%

0%

0%

2.6%

3%

Other

1%

2%

1%

6.0%

6%

Don't know / no response

5%

0%

5%

1.9%

2%

I don't vote for parties but candidates

0%

17%

18%

0.0%

0%

None

0%

1%

1%

0.0%

0%

Total

100%

100%

100%

100.0%

102%

For Figure 5d: Which political party do you feel closest to, if any?

 

2012

2013

2014

2015

Unrounded

Rounded

CCM

65%

54%

54%

61.9%

62%

Chadema

26%

32%

27%

25.0%

25%

CUF

3%

4%

4%

2.3%

2%

ACT-Wazalendo

0%

0%

0%

0.7%

1%

NCCR Mageuzi

0%

1%

1%

0.3%

0%

Ukawa*

0%

0%

0%

2.1%

2%

Other

2%

1%

2%

2.3%

2%

Don't know / no response

4%

0%

2%

5.3%

5%

I don't have a party affiliation

0%

8%

10%

0.0%

0%

Total

100%

100%

100%

100.0%

99%

3. “A sample size of just 1,848 respondents is not sufficient.”

Some have suggested that a sample size of 1,848 people is not sufficient to give a reliable answer. This suggestion is wrong. What matters most is how those people are selected: if they are selected at random, in a way that is representative of the whole of the country’s voting age population, then even a much smaller sample can give useful findings. In countries with much larger populations than Tanzania - such as the USA - opinion poll surveys are commonly done with a sample size of around 1000. The well-respected Afrobarometer survey has a sample size of 2,400 people in Tanzania. They are still able to make statistically valid conclusions about the entire population’s opinions. In contrast, if participants in a survey are self-selecting (such as with most online surveys) or if the sampling technique makes it impossible for large sections of the population to participate (also as with most online surveys), then even a very large sample size will not give statistically reliable results.

In this Sauti za Wananchi survey by Twaweza, respondents were randomly selected and they are representative of the whole country’s voting age population. Therefore, for this sample size, we are 95% confident that the figures presented are an accurate representation of the whole population to within a margin of error of +/- 2.5%. This sample size, confidence level and margin of error are in line with global best practice for opinion poll surveys.
 

Read more: data elections Sauti za Wananchi

Download

You might also like...

Translate