Methodological considerations for future studies

One of the objectives of this study was to develop a methodology and learn lessons from the application of that methodology in at least three member states. The application of the methodology developed in this study has identified a number of issues, which are discussed below.

Were the sample sizes big enough?

The standard errors of the resulting valuations generally are quite large, particularly when we take account that respondents have provided multiple choice observations. In this report we present both the resulting valuations and their 95% confidence interval because these are the usual standards for academic publications, but perhaps this level of confidence is more stringent than what is required by policy makers in this domain. However, even 90% confidence levels would still remain large.

In addition to having a wide range of possible values, having large standard errors also means that we are less likely to observe significant differences in valuations for specific attributes across different market segments. More precise estimates would mean that studies would be more likely to identify differences in preferences for different segments, for example by age or income group, and understanding such differences may be important for policy makers.

We note that the valuation measures would also be improved with better measures of cost sensitivity, which may have occurred with investigation of a larger price range, and this is discussed further below.

We also consider the impact of sample size on the reliability of the valuation estimates. For this study, we aimed to collect (and collected) 125 business interviews and 350 resident interviews per country. We pooled the business observations across the three countries, because we found that pooling the data did not result in a significant reduction in model fit. However, we developed separate models for SMEs (n=225) and big businesses (n=150) because preferences for services differed significantly between these segments. We did not pool the resident interviews across countries, so each resident model was based on information from 350 observations.

Given the importance of businesses in the postal market, particularly because of the volumes of letters they send, we recommend there should be larger sample sizes in future studies. In comparison to this study, the Accent (2008) study for Postcomm in the UK contained 350 business interviews, which we believe would provide a better basis for quantification of business postal preferences. We recommend that separate quota samples for SMEs and big businesses should be specified. The Accent (2008) study also included 550 resident interviews, which again would lead to more reliably estimated valuations for residents. Again, we recommend that quotes for vulnerable and non-vulnerable residents should be specified separately.

Did we test a large enough cost range?

Examination of the trading at different cost levels indicates that a substantial proportion of resident and business respondents are choosing the most expensive options in the SP choice exercises. This means either that the resulting cost sensitivity may be too low, with a risk that the resulting valuations are high, or that other attributes have dominated the choice experiments (e.g. lost items). Better estimated cost sensitivity would also reduce the standard errors of the resulting valuations. In future studies we recommend it is worth considering testing larger price differences as well as looking at the influence of dominant alternatives, which is discussed further below.

Were the respondents able to deal with the two-class options?

The results from the first and second experiments indicate that businesses prefer next day delivery or a two-class service including next day delivery. Businesses, particularly large businesses, place a reasonably negative value on two-day and three-day services where no next day option is in place. However, speed of delivery seems to be less important to residential consumers. Also, we see that large businesses do not favour non-uniform delivery options, whereby letters or parcels in an urban area may be delivered the next day whereas deliveries to more rural areas may take longer, compared with a single class next day service; residential respondents and SMEs are more ambivalent on this issue.

However, we were somewhat surprised to see that neither business nor residential respondents showed a preference for a two-class service including a next day service option, compared with a next day service. One reason may be that respondents found these options more complex than the one-class options. This could be tested in future studies through qualitative research with groups of business and residential consumers. We also saw this pattern in the pilot survey analysis and at that stage amended the calculation of costs for the two-class options to ensure that we presented options where the two-class costs would be both less expensive and more expensive than the one-class options. Reviewing the costs in the main survey confirms that the costs of the two-class options were indeed sometimes cheaper and sometimes more expensive than the one-class costs.

RAND Europe Conclusions and considerations for future studies

Accent

Swiss Economics

The most successful operators to deal with volume per capita in Europe all have two-class options, so this is an important issue that requires more research.

Was the percentage of lost letters and parcels dominant in the choice exercises? The most important service attribute in the first and second experiments is the percentage of letters and parcels which are lost. We increased the range of lost letters and parcels tested in the choice experiments after the pilot survey to make the choices “more different” after comments from respondents that too many of the choices looked the same. Perhaps in future smaller ranges could again be tested (because we made other changes after the pilot survey, including dropping the Saturday delivery attribute). Alternatively, increasing the prices in the experiments may help to make the lost letters and parcels less dominant, but they will still be important when compared to other attributes.

Do consumers value guaranteed time of delivery?

We see evidence that businesses value delivery by 13:00 (or not later than 17.00) for letters and parcels. In contrast, time of delivery for letters did not seem to matter to residents, except in Poland where residents seemed to prefer deliveries during the day compared with early deliveries. We also see preferences for delivery of packages during the day, in Poland and Italy – in retrospect it would have been interesting to collect information on whether the respondents worked or were at home during the day and if they purchased parcel items where they had these delivered to their home (or perhaps to their work).

Did the survey collect enough respondents who did not have internet access?

Although as part of the model analysis we looked to examine whether internet access influenced consumers’ preferences for postal services, we did not find any significant differences between those respondents with internet access and those without.

However, we also observe that only 2% of business respondents did not have access to the internet at work and 6% of consumers had no internet access at all, with nine out of ten having access at home. Although this varied somewhat across the countries, e.g. in Sweden 96% of consumers had internet access at home, in Italy the figure was 91% and in Poland much lower at 86%, in general the levels of internet access was higher than we were expecting, particularly in Poland and Italy. This may have been because people with internet access were more amenable to undertaking the surveys (because it could be done within one single telephone call). Therefore, if using a phone-fax/post/e-mail-phone approach in future we recommend specifying a quota for respondents who do not have access to internet, which would allow a better chance of identifying differences in postal needs between those with and without internet access.