Transcript Business tendency surveys, part 2

Business statistics and registers
Business tendency surveys,
part 2
Copyright 2010, The World Bank Group. All Rights Reserved.
1
Survey frame
• BTS carried out by trade associations use their
membership list as the survey frame
• The international recommendation for BTS is
to use comprehensive business registers that
are also used for other business surveys
• A reasonably comprehensive business register
is a prerequisite for good business tendency
surveys
Copyright 2010, The World Bank Group. All Rights Reserved.
2
Choice of statistical units
• Many enterprises cannot provide information for
their establishments
• Some enterprises can report for their local units
• Others can report for their kind of activity units
• The choice of reporting unit can influence survey
results
• Basing a survey on establishment units is the best
solution
Copyright 2010, The World Bank Group. All Rights Reserved.
3
Sample design
Three methods:
1. Data collection from all enterprises
2. Judgmental sample
3. Random sample
Copyright 2010, The World Bank Group. All Rights Reserved.
4
Panels
• Original panel based on stratified random
sample
• Using a panel reduces sample variance
• Using a panel has also practical advantages
• Panel must be periodically refreshed, e.g. by
changing 25% of enterprises each quarter
Copyright 2010, The World Bank Group. All Rights Reserved.
5
Sample size
• With random sampling the sample size
depends on the variance of one or two key
variables and on the desired level of precision
• Experience shows that often about 30
reporting units are sufficient to obtain an
acceptable level of precision for each strata
for which data are to be published
Copyright 2010, The World Bank Group. All Rights Reserved.
6
Response management
•
•
•
•
Keep in touch with respondents
Personal contacts enhance response rates
Helping respondents understand questions
Contacts with respondents may help to
improve the questionnaire
Copyright 2010, The World Bank Group. All Rights Reserved.
7
Quality of surveys
• Quality elements include reliability, timeliness,
comparability, transparency and accessibility
• BTS are normally timely because
questionnaires can easily be completed and
the number of questions is small
• Use of harmonized questionnaires enhances
comparability over time and between
countries
Copyright 2010, The World Bank Group. All Rights Reserved.
8
Measurement error
• Measurement errors are generated by
questionnaires and respondents
• Errors caused by questionnaires relate to:
ambiguous phrasing of questions
unclear layout of questionnaire
• Errors stemming from respondents may be
caused by:
insufficient knowledge to answer correctly
lack of motivation to report correctly
Copyright 2010, The World Bank Group. All Rights Reserved.
9
From percentages to single number
• Results obtained must be presented as a time
series
• Multiple percentages are difficult to interpret
• Therefore, results are normally presented as a
single number
• Two methods:
1. Balances
2. Diffusion indices
Copyright 2010, The World Bank Group. All Rights Reserved.
10
Balances method
• In most BTS respondents have three reply
options: up/above normal (+), same/normal (=)
and down/below normal (-)
• (=) replies are often discarded, but may be
presented separately
• To calculate the balance, first transform replies
into percentages
• Then subtract (-) from (+) to calculate the balance
Copyright 2010, The World Bank Group. All Rights Reserved.
11
Diffusion indices
• Balances (B) are calculated as: B = 100 ( P – N )
• Diffusion indices (DI) are calculated as 100 ( P +
E/2 ), where P is the fraction of (+) replies in the
total, N of (-) replies, and E of (=)
• Balances can take values from –100 to +100,
diffusion indices from 0 to 100.
• Balances and diffusion indices have exactly the
same information content
• Balances are more commonly used
Copyright 2010, The World Bank Group. All Rights Reserved.
12
Weighting
Two types of weighting:
1. Sample weights
2. Size weights
• Sample weights are the inverse of the probability
of selection
• Size weights reflect the relative contribution of
enterprises
• Value added is probably the best size weight
• In practice, number of employees is more often
used
Copyright 2010, The World Bank Group. All Rights Reserved.
13
Publication
• As for all statistics, metadata must be presented
for BTS surveys as well
• Metadata must cover information about:
The sampling frame
Statistical units
Sampling
Weighting
Data collection methods
Treatment of non-response
Precision
Comparability
Copyright 2010, The World Bank Group. All Rights Reserved.
14
Interpretation
• Because BTS are subjective, respondents may
interpret questions differently
• Issues of different interpretation are:
What is ‘normal’?
What reference period to use
How to judge ‘capacity utilization’?
Expected ‘business situation’
• Systematic bias is also possible
• Experience shows that most of these factors do
not seriously distort BTS results
Copyright 2010, The World Bank Group. All Rights Reserved.
15