China

For its China Report series, Mintel has commissioned exclusive consumer research through KuRunData, a Chinese licensed market survey agent. Online consumer research is run in ten cities, completing 300 interviews per city with a total sample size of 3,000.

  • Our consumer research is based on a random sample of internet respondents from a panel recruited by KuRunData (see more details below).
  • In each wave, we survey four major tier 1 cities ie Shanghai, Beijing, Guangzhou and Chengdu.
  • For tier 2, tier 3 or lower cities, we rotate amongst a selection of cities based on size and economic development (see below).
  • The research is not representative of the population as a whole, and is not being analysed as such. Mintel applies a quota-sampling approach with quotas on age, gender and monthly household income in these cities.
  • Our sample data can only be considered as indicative of urban consumers in those regions rather than representative of China as a whole.
  • For category reports (e.g. chocolate confectionery), consumers will typically be asked about usage, frequency, location of purchase, consumption occasion, brand usage and a series of attitudinal statements about the category.
  • Lifestyle reports cover a broader range of attitudinal and behavioural topics.

Confidence levels

Statistical confidence levels of +/- 2% or 3% can be applied to the data, depending on sample size and percentage of respondents. For example, if 20% of a total sample of 1,000 adults say that they do something, you can be 95% certain that the figure for the population lies between 17% and 23%. For a sample of 3,000 adults, you can be 95% certain that the figure lies between 18% and 22%.

Consumer research is stored in a database supervised by Mintel's statisticians. Additional analysis of information too abundant to be included in published reports can be made available.

Sample sizes by city

When the overall population of a city is large enough (> 20,000), sample size is not determined by the size of population. It is only when the population becomes quite small (eg less than 10,000) that the sample size calculation is affected.

As a result a sample size of 300 per city across all 10 cities in our survey was set. Statistically, this enables us to apply a confidence level at 95% with a margin of error of 5.66%.

Sampling methodology and sampling structure

According to government figures, there are 645 cities in China. These cities are very different in terms of size, economic development, culture, history and lifestyle. To meet Mintel’s client interests (by region, by tier), ten cities are selected in each wave of research based on their geographic coverage and level of economic development (GDP and per capita income).

The table below shows an example of cities covered:

Region  Tier 1 cities  Tier 2 cities  Tier 3 or lower cities  Total 
North China  Beijing  Shenyang    2
East China  Shanghai  Hangzhou  Jinhua
Middle China    Wuhan   
West China  Chengdu    Hengyang 2
South China  Guangzhou  Fuzhou    2
Total  4  4  2  10 

Note 1: Tier 2, Tier 3 or lower cities in the table above are shown as examples only

Note 2: Mintel defines the tier levels of cities in China as follows:

  • Tier 1: Major economic hubs
  • Tier 2: Provincial capital cities and some developed non-provincial capital cities
  • Tier 3 or lower: Other cities beside Tier 1 and Tier 2 cities

Within each city, our sampling structure is presented below:

   

Gender & Age

Monthly household income (RMB)

Tier 1 city

Total

Male

20-29

Male

30-39

Male

40-49

Female

20-29

Female

30-39

Female

40-49

6,000-9,999

10,000-17,999

18K+

Beijing

300

50

50

50

50

50

50

100

100

100

Shanghai

300

50

50

50

50

50

50

100

100

100

Guangzhou

300

50

50

50

50

50

50

100

100

100

Chengdu

300

50

50

50

50

50

50

100

100

100

Tier 2, 3 or

lower city

Total

Male

20-29

Male

30-39

Male

40-49

Female

20-29

Female

30-39

Female

40-49

5,000-8,999

9,000-15,999

16K+

City 1

300

50

50

50

50

50

50

100

100

100

: : : : : : : : : : :
: : : : : : : : : : :
  • Monthly personal income

         We defined 3 different levels for monthly personal income according to different city tier:

          Monthly personal income         Tier 1 cities                            Tier 2, 3 or lower cities

          Low personal income                   RMB2,000 - 5,999                    RMB2,000 - 4,999

          Middle personal income               RMB6,000 - 9,999                    RMB5,000 - 8,999

          High personal income                  RMB10,000 or above                RMB9,000 or above

  • Monthly household income

         We defined 3 different levels for monthly household income according to different city tier:

         Monthly household income         Tier 1 cities                             Tier 2, 3 or lower cities

         Low household income                   RMB6,000 - 9,999                     RMB5,000 - 8,999

         Middle household income               RMB10,000 - 17,999                  RMB9,000 - 15,999

         High household income                  RMB18,000 or above                  RMB16,000 or above

Our research partner - KuRunData

  • Founded in 2006, headquartered in Shanghai, with branches in Beijing and Guangzhou
  • Online panel size - 5,300,000 (by Dec 2018)
  • Owns the interactive panel websites: www.1diaocha.com and www.votebar.com and WeChat survey applet
  • Member of the China Market Research Association (CMRA) and member of ESOMAR
  • Joined ITWP Group in 2017,  the group includes well-known market research companies such as Toluna and harrisinteractive
  • Completed more than 4,000 projects and interviewed more than 2,000,000 samples per year
  • Provides panel data for major multinational research companies, including IPSOS, Lightspeed, TNS, Nielsen, Kantar, Intage, Pulse, SSI, ResearchNow

KuRunData's sampling and quality control

1.Screening

Exclude those who

  • have participated in any survey project in the past three months
  • are working in any sensitive or related industries
  • have participated in previous Mintel surveys
  • have not met sample criteria
  • spend less time than average on answering the survey

2.Sampling

  • System will output all qualified panellists
  • Random sample 30,000 panellists and send out first wave of invitations via SMS or email
  • Random sample another 30,000 panellists and send out second wave of invitations
  • Random sample final batch of 30,000 panellists and send out third wave of invitations

3. Quality control

  • Each panellist has provided KuRunData with his or her own IP address together with all personal information
  • Each panellist needs to use the same IP address and cookie as registered IP address and cookie to participate in any survey project, and he or she can participate in the same survey project only once
  • Each respondent can use the web link they have received once only

A sample will be deleted if the respondent

  • has given an answer to any open-ended question that is judged as of poor quality
  • has failed in any trap questions
  • has given any inconsistent answers, or contradicted his or her registered information
  • has given answers following certain patterns
  • has taken an extraordinary length of time to complete the questionnaire
  • is considered an outlier

Meet the Mintropolitans

Why Mintropolitans?

There has long been unresolved debate as to how to define the “middle class” in China, using various formulae about income levels, educational attainment and ownership of certain key items. However, the notion of “middle class” is very much an invention of 19th century North America and Western Europe, and does not comfortably translate that well into 21st century China.

So, rather than continue to struggle to make the round peg of China fit into the square hole of the “middle class”, Mintel has decided to use a clear and practical definition to define those who not only have higher spending power but can also represent future consumption trends as ‘Mintropolitans’. They had to meet all of the following requirements:

  • They need to have a higher level of income, depending on which city they live in: this being a household income of RMB18,000 per month or over in tier one cities; or RMB16,000 per month or over in tier two, three or lower cities.
  • They need to have a higher level of education: undergraduate or above.
  • They need to own a property – either outright or on a mortgage.
  • Lastly, they also need to demonstrate certain spending attitudes and lifestyles, chosen by Mintel to indicate their spending power, as well as reflecting their pursuit of a quality life.

Broadly, they should represent a sophisticated group of consumers who pursue quality of life rather than just wealth, are well educated and are the potential trendsetters. 

Who are they?

Based on demographic data from the consumer research conducted across multiple Mintel Reports, Mintropolitans account for about 15% of total surveyed households – representing a population of 27 million households who live in China.

Compared to other groups of consumers, besides having a higher income, Mintropolitan are much more likely to be aged 30-39, married with children, have a postgraduate degree and work in foreign companies compared to Non-Mintropolitans.

Further Analysis

Mintel employs a variety of quantitative data analysis techniques to enhance the value of our consumer research.  The techniques used vary from one report to another.  Below show some of the techniques more commonly used.

  • Repertoire Analysis

This technique is used to create consumer groups based on reported behaviour or attitudes. Consumer responses of the same value (or list of values) across a list of survey items are tallied into a single variable. The repertoire variable summarizes the number of occurrences in which the value or values appear among a list of survey items.  For example, a repertoire of brand purchasing might produce groups of those that purchase 1-2 brands, 3-4 brands and 5 or more brands.  Each subgroup should be large enough (ie N=75+) to analyse.

  •  Cluster Analysis

This technique used to assign a set of individual people to groups called clusters on the basis of one or more question responses, so that respondents within the same cluster are in some sense closer or more similar to one another than to respondents that were grouped into another cluster.

  •  Correspondence Analysis

This is a statistical visualisation method for picturing the associations between rows (image, attitudes) and columns (brands, products, segments, etc.) of a two-way contingency table.  It allows us to display brand images (and/ or consumer attitudes towards brands) related to each brand covered in this survey in a joint space that is easy to understand.  The significance of the relationship between a brand and its associated image is measured using the Chi-square test.  If two brands have similar response patterns regarding their perceived images, they are assigned similar scores on underlying dimensions and will then be displayed close to each other in the perceptual map.

  •  CHAID analysis

CHAID (Chi-squared Automatic Interaction Detection), a type of decision tree analysis, is a target group identification method that is used to highlight key target groups in a sample by identifying which sub-groups are more likely to show a particular characteristic. This analysis subdivides the sample into a series of subgroups that share similar characteristics towards a specific response variable and allows us to identify which combinations have the highest response rates for the target variable. It is commonly used to understand and visualize the relationship between a variable of interest such as “interest in trying a new product” and other characteristics of the sample such as demographic composition.

  • Key Driver Analysis

Key driver analysis can be a useful tool in helping to prioritise focus between different factors which may impact key performance indicators (eg satisfaction, likelihood to switch providers, likelihood to recommend a brand, etc). Using correlations analysis or regression analysis we can get an understanding of which factors or attributes of a market have the strongest association or “link” with a positive performance on key performance indicators (KPIs).  Hence we are able to identify which factors or attributes are relatively more critical in a market category compared to others and ensures that often limited resources can be allocated to focusing on the main market drivers.

  • TURF Analysis

Key TURF (Total Unduplicated Reach & Frequency) analysis identifies the mix of features, attributes, or messages that will attract the largest number of unique respondents. It is typically used when the number of features or attributes must be or should be limited, but the goal is still to reach the widest possible audience. By identifying the Total Unduplicated Reach, it is possible to maximize the number of people who find one or more of their preferred features or attributes in the product line.

  • Price Sensitivity Analysis

Price sensitivity analysis shows consumer expectations about pricing of a finished product. Consumers were asked to provide a price point for the finished product. The aggregate price points are then plotted onto Price Maps to indicate Point of Marginal Cheapness (PMC) , Point of Marginal Expensiveness (PME) as well as the Optimal Price Point (OPP).

Statistical forecasting

Statistical modelling

For the majority of reports, Mintel produces five-year forecasts based on an advanced statistical technique known as ‘multivariate time series auto-regression’ using the statistical software package SPSS.

The model is based on historical market size data taken from Mintel’s own market size database and supplemented by published macroeconomic and demographic data from various private and public sources such as the NBS (National Bureau of Statistics of China) and the EIU (The Economist Intelligence Unit).

The model searches for relationships between actual market sizes and a selection of relevant and significant macroeconomic and demographic determinants (independent variables) to identify those predictors having the most influence on the market.

Factors used in a forecast are stated in the relevant report section alongside an interpretation of their role in explaining the development in demand for the product or market in question.

Qualitative insight

At Mintel we understand that historic data is limited in its capacity to act as the only force behind the future state of markets. Thus, rich qualitative insights from industry experts regarding future events that might impact upon various markets play an invaluable role in our post statistical modelling evaluation process.

As a result, the Mintel forecast complements a rigorous statistical process with in-depth market knowledge and expertise to allow for additional factors or market conditions outside of the capacity of the statistical forecast.

[graphic: image 1]

  

The Mintel fan chart

Forecasts of future economic outcomes are always subject to uncertainty. To raise awareness amongst our clients and to illustrate this uncertainty, Mintel has introduced a new way of displaying market size forecasts in the form of a fan-chart.

Next to historical market sizes and a current year estimate, the fan chart illustrates the probability of various outcomes for the market value/volume over the next five years.

At a 95% confidence interval, we are saying that 95 out of 100 times, the forecast will fall within these outer limits, which we call the best and worst case forecast as these, based on the statistically driven forecast, are the highest (best case) and lowest (worst case) market sizes the market is expected to achieve.

Over the next five years, the widening bands successively show the developments that occur within 95%, 90%, 70% and 50% probability intervals. Statistical processes predict the central forecast to fall within the darker shaded area which illustrates 50% probability ie a 5 in 10 chance.

A general conclusion: Based on our current knowledge of given historic market size data as well as projections for key macro- and socio-economic measures that were used to create the forecast, we can assume that in 95% of the time the actual market size will fall within the purple shaded fan. In 5% of all cases this model might not be correct due to random errors and the actual market size will fall out of these boundaries.

Weather analogy

To illustrate uncertainty in forecasting in an everyday example, let us assume the following weather forecast was produced based on the meteorologists’ current knowledge of the previous weather condition during the last few days, atmospheric observations, incoming weather fronts etc.

[graphic: image 2]

  

Now, how accurate is this forecast and how certain can we be that the temperature on Saturday will indeed be 15°C?

To state that the temperature in central Shanghai on Saturday will rise to exactly 15°C is possible but one can’t be 100% certain about that fact.

To say the temperature on Saturday will be between 13°C and 17°C is a broader statement and much more probable.

In general, we can say that based on the existing statistical model, one can be 95% certain that the temperature on Saturday will be between 13°C and 17°C, and respectively 50% certain it will be between about 14.5°C and 15.5°C. Again, only in 5% of all cases this model might not be correct due to random errors and the actual temperature on Saturday will fall out of these boundaries and thus will be below 13°C or above 17°C.

(To learn more about uncertainty in weather forecasts visit: http://research.metoffice.gov.uk/research/nwp/ensemble/uncertainty.html)