National Travel Survey 2022: Technical report chapter 1
Published 30 August 2023
Applies to England
© Crown copyright 2023
This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: [email protected].
Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.
This publication is available at https://www.gov.uk/government/statistics/national-travel-survey-2022-technical-report/national-travel-survey-2022-technical-report-chapter-1
Chapter 1: Fieldwork approach in 2022
The coronavirus (COVID-19) pandemic began in early 2020 and caused an unprecedented shock to data collection in social research surveys. As a random probability survey with a single mode of data collection (face-to-face interviewing), the National Travel Survey (NTS) was particularly impacted, with the data collection methodology being changed significantly.
The NTS remains the DfT’s major source of information on personal travel for people living in England. Given this, it was vital that data collection continued during this period whilst minimising the impact on data quality and the time series as much as possible.
Although the pandemic eased over the course of the 2022, initially COVID-19 remained a major concern at the beginning of the year particularly with the emergence of the Omicron variant in November 2021. As a result, the NTS methodology was still affected. This chapter outlines the steps taken by the teams managing the NTS to respond to ongoing COVID-19 concerns in 2022 and summarises the associated limitations of the data.
1.1 Background
Since its inception in 1965, the National Travel Survey (NTS) has operated using a face-to-face (F2F) design. Under the F2F approach, initial contact with the selected addresses was made by an advance letter, sent through the post. This was then followed-up by a visit to the address by a trained interviewer who attempted to encourage participation and complete a F2F in-home survey and 7-day travel diary, which was then collected by the interviewer from the respondent’s home.
Following the onset of the COVID-19 pandemic, the government introduced public health measures which restricted non-essential travel and interpersonal contact in March 2020. As a result, the F2F approach was no longer feasible on the NTS, and fieldwork was officially paused on 18 March 2020. Given this, February and March work was curtailed and fieldwork for April was closed in its entirety for the 2020 survey.
As government restrictions and uncertainty around the pandemic remained throughout the rest of 2020, NTS fieldwork transitioned to a telephone only methodology known as ‘push-to-telephone’ (P2T). The lower response rate associated with this methodology ultimately impacted the statistical and analytical power of the data for the 2020 survey.
In 2021, with COVID-19 remaining a concern, the NTS had to alter the way in which it collected data in order to offset the decline in response and data quality from the 2020 P2T approach. The survey transitioned to a ‘knock-to-nudge’ (K2N) methodology – interviewers contacted households by visiting their home address maintaining social distancing, but they completed the interview over the phone. The travel week was set in all cases to begin the day before the placement interview (known as the ‘rolling travel week’ approach), and then interviewers populated the diary on behalf of the respondent via telephone. For any points (also known as assignments) that could not be covered by an interviewer the P2T methodology was used instead. These approaches combined allowed for higher levels of response and for the sample profile to return to levels closer to the pre-pandemic period.
For 2022, the intention was to return to the traditional F2F methodology. However, the rise of the Omicron variant at the end of 2021 meant that return to F2F was delayed until April 2022 (the start of the quarter two fieldwork period) and K2N and P2T remained in place for quarter one. From April, the survey returned to F2F with a phone back-up (PB) for interviewers to make use of in selected households where COVID-19 was a concern. P2T remained in place for any points where it was not possible to allocate to an interviewer.
The rest of this chapter will outline the details of the approach used in 2022.
1.2 Mixed mode approach 2022: Returning to face-to-face fieldwork
For 2022, the objective was to return to F2F as soon as feasibly possible. Whilst data collection had continued in 2020 and 2021, the limitations of the methodology used during that time had impacted the quality of the data. In particular, the surveys were impacted by lower level of response and an uneven distribution of the travel weeks start dates, both across the days of the week and dates within the month.
However, the rise of the Omicron variant in November 2021 meant that a return to F2F interviewing from the beginning of 2022 was no longer viable. Working with the DfT, NatCen delayed the return of the F2F approach to April, ensuring that the 2021 methodology continued for January to March fieldwork (quarter one). From April onwards (quarters two, three and four), F2F returned with a phone back-up (PB).
1.2.1 January to March: Knock-to-nudge
For quarter one 2022, the same 2021 K2N fieldwork approach was rolled over, however initial contact was made using a revised advanced letter. This revised letter stated that the interview could be conducted in the respondent’s home or by telephone and that this would depend on the current guidance regarding COVID-19. The letter also indicated that an interviewer would visit the household to make contact and encourage participation, whilst maintaining social distancing on the doorstep. If a household was willing to take part, the interviewer scheduled a telephone interview. All the initial placement interviews (and pick-up interviews) would then be conducted over the phone, while travel diary entry was also completed over the phone with interviewers completing the diaries on behalf of the household. The ‘rolling travel weeks’ approach from 2020-2021 was retained. This meant that the travel week started the day before the placement interview took place. Relaxed proxy rules from 2020-2021 were also continued into quarter one 2022, enabling one individual household member to complete the interview on behalf of the other people living in the household.
1.2.2 April to December: Face-to-Face (with phone back-up)
For quarter two onwards, the NTS returned to the traditional F2F methodology. As with the K2N method from quarter one, interviewers issued the advance letter to make initial contact. The same letter from quarter one was used. The interviewer would then make a visit to the sampled address, encouraging the household to take part on the doorstep. If a respondent was willing to take part in-home, the interviewer would complete the initial placement interview with everyone in the household, with the stricter limitations on proxy interviews reintroduced from the pre-pandemic period. The self-complete section of the interview also returned to being completed by the respondents rather than being administered by interviewers. Unlike the K2N approach, interviewers were also instructed to return to using ‘fixed travel weeks’ (that is, allocating the travel weeks as they had been done prior to the pandemic). Following a reminder call from interviewers (to remind respondents that their travel weeks were starting), the diaries were then completed by respondents in their assigned week and checked by interviewers during a mid-week checking visit or call. The pick-up interview was once again completed face-to-face, and the finalised diaries were checked and picked up at the same visit.
Interviewers were expressly briefed to prioritise F2F interviewing wherever possible. However, interviewers were able to make use of the PB and complete the interview over the phone if:
- respondents clearly stated they did not want interviewers to enter their home due to reasons related to COVID-19 (for example, shielding or fear of infection)
- the household failed the interviewer’s COVID-19 risk assessment
- the household was not willing to complete the interview F2F (used as a last resort only)
The PB process was similar to the K2N process. Having made contact with a household at their home address in person and having established that a phone interview was necessary, interviewers booked a telephone appointment with the household. After completing the placement interview over the phone, interviewers would allocate the fixed travel weeks (in the same manner as F2F interviews). Prior to the beginning of this travel week, interviewers would make a reminder call to the respondent over the phone, and then make two mid-week checking calls during the travel week (ideally on day two and then again on days four or five). During these checking calls, interviewers would populate a couple of days of the respondents’ diaries on their behalf. At the end of the travel week, the pick-up interview was completed over the phone and the outstanding days of the diary populated.
For both quarter one (K2N) and quarters two to four (F2F with PB), some sample points were not covered by NatCen field-based interviewers. A total of 756 sample points were selected overall. However, the challenges of re-building fieldwork capacity for the reinstatement of face-to-face interviewing after the height of the COVID-19 pandemic meant that it was not possible for interviewers to cover all selected points in 2022. In total, 99 points were not covered by field-based interviewers and were worked via the push-to-telephone (P2T) approach, following the same method to the one developed in 2020 and used in 2021. Initial contact was made using an alternative version of the advance letter which invited households to opt-in to the study through any of the following means:
- calling a freephone number where a trained interviewer collected their contact details
- logging into a bespoke online portal to leave their contact details
- emailing a central NTS inbox
Once a household had opted-in, they were then contacted by trained NatCen interviewers to conduct the interview and populate the diaries on the household’s behalf over the telephone.
Taken together, these approaches enabled the NTS to continue collecting data through the final stages of the pandemic whilst building on the methodologies from 2020 and 2021 in a number of ways. As a result, only minimal changes to the survey programme were required, meaning that there was no need to re-design multiple questions or reduce the length of the questionnaire, therefore ensuring consistency with the prior years and across modes.
Although the phone interviewing was different to the pre-pandemic methodology, the interviews continued to be interviewer-administered, rather than resorting to self-administered methodologies. This approach, combined with the return of F2F interviewing from quarter two, therefore minimised the potential for measurement effects.
As with 2020 and 2021, the standard NTS sampling approach was used for the survey. This meant that the random probability design could be retained. Furthermore, the return of F2F interviewing allowed response rates to get closer to pre-pandemic levels (see section 1.3.1), and the continuation of interviewer involvement meant that improvements in the sample profile seen in 2021 continued into 2022 to some extent (see section 1.3.2). Interviewer involvement from 2021 onwards also permitted random dwelling unit and household selection to be re-introduced, as per pre-pandemic protocols, helping to improve the quality of the data collected.
The 2022 approach allowed for methodological flexibility, which meant that the study could pivot from the K2N approach to the traditional F2F mode (albeit with PB) within the survey year. The return to F2F fieldwork for quarter two onwards was also accompanied with the reintroduction of ‘fixed travel weeks’. This approach, taken from the pre-pandemic era, was brought back to ensure that travel weeks would be evenly spread across the days of the week and the dates within the month, and therefore helping to ensure (from quarter two onwards) that the data would be more representative of the country’s travel across the year and able to account for seasonality more effectively.
The presence of PB alongside the F2F method also ensured that hard-to-reach groups that may have been difficult to recruit with F2F only (for example, people that are shielding from COVID-19) were able to participate in the survey, which once again helped to maximise the data quality.
1.2.3 Summary of changes for 2022
Due to the variety of data collection methods used in 2022, several changes were made to the standard NTS design. These changes are outlined below.
From April 2022 onwards, the K2N method that had been in place for 2021 and January to March 2022 was removed and the F2F interviewing that had been in place prior to the pandemic returned (albeit with a phone back-up option). For more information, see sections 1.2.1 and 1.2.2 above.
An updated version of the advance letter was used for 2022. The letter from 2021 was edited for the 2022 survey to make it flexible for both phone and in-home interviewing. This letter was used by all interviewers during 2022 fieldwork (with the exception of P2T points). For any points (also known as assignments) that could not be covered by an interviewer, a separate advance letter was used for push-to-telephone cases based on the P2T letters previously used during the pandemic.
For both K2N and F2F/PB cases in 2022, the random selection of households or dwelling units (in instances where the sampled address contained more than one household or dwelling unit) was retained. Consistent with how the NTS has been run in previous years, the sample frame for the NTS was again the Postcode Address File (PAF) containing a list of delivery points (that is, addresses). One of the known issues of using the PAF as a sample frame is that there could be more than one dwelling unit or household at any given delivery point, although such cases are in the minority. In these instances, interviewers have previously been able to conduct a random selection of a dwelling and/or household, as per the pure F2F methodology that was in place pre-pandemic.
Given that in-home contact was prohibited at times during the pandemic, this household and dwelling unit selection process was not used in 2020. This process was brought back in 2021 and retained for 2022. P2T cases were not subject to this selection process at any point during the pandemic (including 2022) as interviewers did not visit addresses in person. The selection for these cases was therefore left to chance out of necessity, and so the first household which responded was effectively the one selected.
The return to pre-pandemic proxy rules for quarter two onwards. Before the pandemic, all interviewers aimed to interview each respondent in the household in-person as often as possible, thereby minimising proxy interviews and improving data quality. Following the outbreak of COVID-19 and the subsequent pausing of in-home F2F interviewing, the proxy rules were relaxed for 2020, 2021 and for quarter one of 2022. This meant that for households participating via a phone interview they could, if necessary, be completed by one person acting as a proxy for the other household members. From quarter two 2022 onwards, the stricter pre-pandemic proxy rules were put back in place for all F2F cases, however the relaxed telephone interviewing proxy rules remained in place for PB and P2T cases.
The return of allocated travel weeks from April onwards. Historically, the NTS design has used a ‘fixed travel week’ approach, whereby interviewers have provided a household with a specific week in which to record their travel in the travel record (diary). This allocation method follows a methodical process in which the next available travel week is provided to the household from a randomly selected list of travel weeks for the individual point. Via this method, the allocated travel week for a given household could have sometimes occurred several days, perhaps two weeks or more, after their initial placement interview. Likewise, following this method it would have also been possible for the allocated travel week to have started before the placement interview, but by no more than two days. (Note: The two-day maximum back-dating rule aims to reduce any bias caused by respondents being unable to correctly recall travel they did in the past).
During the pandemic when telephone interviewing was the sole interviewing mode (2020, 2021 and quarter one 2022) the NTS used a ‘rolling travel week’ approach. This meant that travel weeks always started the day before the date of the placement interview. This was to allow completion of the first day of the travel diary during the initial placement interview and to allow the flexibility to switch the contact approach (from K2N to P2T) if an interviewer was unable to cover a point at short notice. However, from quarter two 2022 the fixed travel weeks were reinstated, regardless as to whether the interviews were completed F2F or using the PB. This was done to improve data quality, whilst attempting to bring the data back in line with that collected using the traditional NTS methodology and to maintain the time series. For practical reasons though, the rolling travel weeks remained in place for P2T cases and for any cases where the allocated travel weeks had passed (for example, if an interviewer was reassigned a point towards the end of the fieldwork period).
Respondent completed diaries for F2F interviews from April onwards. During the pandemic when telephone interviewing was the sole interviewing mode, the travel records were completed by interviewers on behalf of the respondents. This was done during the mid-week and pick-up telephone calls by interviewers prompting respondents to give their travel details to them, with the interviewers then populating the diaries on the respondents’ behalf. With the return to F2F interviewing from quarter two onwards, diaries were once again completed by the respondents themselves. For practical reasons, however, the interviewer-populated diaries remained in place for all telephone interviews (PB and P2T).
The P2T cases retained a different incentive to K2N cases in quarter one and F2F/PB cases in quarter two onwards. Throughout the pandemic whenever P2T has been used, households recruited via this approach have received a £20 household-level incentive (shopping voucher) for completing the study in order to offset the lower response rates associated with the P2T methodology.
Prior to the pandemic, the conditional incentive was £5 per individual household member in fully productive households. This incentive strategy was re-introduced for K2N fieldwork during 2021 and continued into 2022 for all K2N and F2F/PB cases. Additionally, for F2F cases the interviewer handed out the incentive vouchers at the end of the pick-up interview, whereas households completing via telephone were sent the incentive from the central NatCen logistics office.
After the return of the F2F interviewing, it was possible for the self-completion section of the interview to be completed by respondents. Historically, the placement interview has included a module requiring interviewers to hand their laptop over to respondents to answer some attitudinal questions (the CASI section of the interview). For telephone interviewing that has been in place since the COVID-19 outbreak, this approach was clearly not possible, so the module was instead interviewer administered. However, from April 2022 onwards the CASI section of the interview reverted to being respondent completed for interviews conducted F2F, although for telephone cases the interviewer-administered approach remained in place.
Minor edits were made to the CAPI questionnaire to reflect the DfT’s updated needs for 2022. The questionnaire remained largely the same during the pandemic and into 2022 with a small number of minor changes. These included adding in a new religion question and broadening out any mention of ‘bicycle’ to refer to ‘pedal cycle’ instead.
Additional codes added to the diary data. In 2022, the diary remained the same, however there were additional travel mode codes added to the dataset for ferries and mobility scooters. The additional travel mode codes as well as the codes added in 2021 (e-scooter and e-bike) were added to the existing StageMode and MainMode variables. This involved revising the existing imputation routines for the updated variables.
1.3 Impact of the switch in modes
As with 2020 and 2021, the change to the fieldwork approach in 2022 has had an impact on both the level of response and non-response bias. Consequently, this has had an impact on the survey estimates.
1.3.1 Impact on levels of response
The initial observable impact on the achieved sample related to the sample size. Prior to the pandemic, the response rate achieved each month varied from around 50-60% (resulting in around 480-600 fully productive households each month).
The response rate for 2022 was 31% for the year as a whole, which was a decrease from 38% in 2021. For context, there was a decline in the response rates from the end of quarter three of 2021, and this lower response rate stabilised and then continued into quarter one of 2022 (averaging out at 28% across January to March 2022). The response rate immediately increased for April (start of quarter two) to 40% and then fluctuated between 29% and 34% for the remainder of 2022, which is lower than pre-pandemic levels.
Table 1.1 shows the number of addresses issued each month, as well as the number of productive addresses (broken down by total productive cases and fully productive cases) and the response rate. The number of total productive addresses includes both fully and partially productive cases. The table also demonstrates the increase in the response rate in April, and then the relative stabilisation of the response rate for the remaining months of the year.
Table 1.1: Addresses issued and number responding, by month
| Month | Addresses issued | Total productive addresses | Fully productive addresses only | Response rate (fully productive) |
|---|---|---|---|---|
| January | 1,105 | 333 | 286 | 28% |
| February | 1,003 | 280 | 246 | 27% |
| March | 1,105 | 319 | 277 | 28% |
| April | 1,054 | 444 | 380 | 40% |
| May | 1,088 | 374 | 299 | 30% |
| June | 1,071 | 400 | 318 | 33% |
| July | 1,088 | 365 | 314 | 32% |
| August | 1,088 | 368 | 313 | 32% |
| September | 1,037 | 365 | 307 | 33% |
| October | 1,071 | 348 | 292 | 30% |
| November | 1,054 | 321 | 274 | 29% |
| December | 1,088 | 378 | 340 | 34% |
| Total | 12,852 | 4,295 | 3,646 | 31% |
Note: Response rates are calculated for fully productive cases only, and by excluding (1) ineligible cases, and (2) a proportion of the cases with unknown eligibility (based on the known eligibility rate).
Table 1.2 shows a monthly breakdown of the number of issued addresses P2T in 2022 compared to the number of addresses issued either K2N or F2F, as well as the proportion of addresses issued as P2T. For the year, 13% of all addresses were issued P2T, with the proportion noticeably falling from 22% in quarter one to 10% from quarter two onwards.
Table 1.2: Addresses issued and number responding, by data collection type and by month
| Month | Total addresses issued | Addresses issued K2N/F2F | Addresses issued P2T | Proportion of addresses issued P2T |
|---|---|---|---|---|
| January | 1,105 | 816 | 289 | 26% |
| February | 1,003 | 799 | 204 | 20% |
| March | 1,105 | 901 | 204 | 18% |
| April | 1,054 | 1,054 | 0 | 0% |
| May | 1,088 | 969 | 119 | 11% |
| June | 1,071 | 1,003 | 68 | 6% |
| July | 1,088 | 986 | 102 | 9% |
| August | 1,088 | 952 | 136 | 13% |
| September | 1,037 | 901 | 136 | 13% |
| October | 1,071 | 935 | 136 | 13% |
| November | 1,054 | 935 | 119 | 11% |
| December | 1,088 | 935 | 153 | 14% |
| Total | 12,852 | 11,186 | 1,666 | 13% |
Overall, the response rate was considerably lower for P2T (9%) than it was for K2N and F2F combined (35%), as can be seen in Table 1.3 below. This demonstrates that the interviewer involvement in recruitment has had a substantial impact on increasing the level of response.
Table 1.3: Response rates: K2N/F2F Vs. P2T
| Data collection method | Number of cases issued | Number of fully productive interviews | Response Rates | |
|---|---|---|---|---|
| K2N/F2F | 11,186 | 3,528 | 35% | |
| P2T | 1,666 | 118 | 9% | |
| Total | 12,852 | 3,646 | 31% |
Note: Response rates are calculated for fully productive cases only, and by excluding (1) ineligible cases, and (2) a proportion of the cases with unknown eligibility (based on the known eligibility rate). For P2T cases, the ineligible rate from 2019 is used to account for P2T interviewers not visiting the addresses in person to establish eligibility. See Chapter 4 for further information.
As noted above, the overall response rate fell from 38% in 2021 to 31% in 2022 (31%). The unweighted fully-productive sample size also decreased, from 4,429 households (9,971 individuals) in 2021 to 3,646 households (8,087 individuals) in 2022.
An assessment of the outcome classifications for each survey year suggests that the drop in response was mainly driven by two factors. Firstly, an increase in the partially productive rate, from 3% in 2021, which is historically low, to 6% in 2022, which is more in-line with the historical trend. Secondly, an increase in the refusal rate from 27% in 2021 to 30% in 2022. However, the changes in the data collection approaches across the two survey years (outlined in section 1.2) make it challenging to unpick the reasons behind these differences, making direct comparison problematic.
The response rate for NTS 2021 was uneven across the year, as it was noticeably higher in the first half of the year than the second half. This resulted in a productive sample that was skewed towards the first half of the year. As such, when weighting the 2021 data it was necessary to split the non-response model into quarters, and the calibration into halves. In contrast, the level of response was much more even across months in 2022 as shown in Table 1.1. This, along with other differences apparent in the 2022 data, most notably less bias in the sample and a lower proportion of P2T cases, meant that the weighting efficiency for 2022 was much higher than in 2021. For further information, see the NTS 2021 Technical report.
For instance, in 2022 the fully productive sample consisted of 3,646 households (8,087 individuals). The corrective weights have an 85% efficiency, which means that the effective sample size is 3,116 households. By comparison, the 2021 fully productive sample was larger with 4,429 households (9,971 individuals), but the weighting efficiency was 62%, which means that the effective sample size is 2,766 households.
This means that while the unweighted productive sample size was lower in 2022 than it was in 2021 (by around 18% at the household level), the sample has less bias, and the effective sample size is higher (by around 13%). As such the 2022 data is more robust and has more statistical power when running analyses.
1.3.2 Impact on non-response bias
The changes to the methodology since the outbreak of COVID-19 had further impacts on the achieved sample which are highlighted in this section.
Investigations by NatCen into the achieved sample profile showed differences in the distribution of certain characteristics year on year. The introduction of K2N in 2021 moved the sample profile closer to the pre-pandemic distribution. This trend largely continued in 2022 following the introduction of F2F interviewing.
Household-level distribution
Given that the NTS is a household-level survey, the profile differences across the survey years were initially observable in the distribution of household characteristics, such as tenure. The tenure profile for 2021 was closer to 2019 (the last pre-pandemic year) than the 2020 survey, which had over-represented homeowners and under-represented renters. The profile of 2022 was also closer to 2019 than 2020, however some differences between 2021 and 2022 were apparent.
Most notably, 2022 deviated a little further from the 2019 benchmark by over-representing outright homeowners more than in 2021 (39.8% in 2019, 41.6% in 2021, 43.8% in 2022). In contrast, however, 2022 moved slightly closer to the 2019 benchmark than 2021 for home buyers (with a mortgage) and renters. Home buyers (with a mortgage) made up 26.8% of the sample in 2019, 29.1% in 2021, 25.8% in 2022. Renters made up 31.7% in 2019, 27.4% in 2021, 28.1% in 2022.
Table 1.4 compares the unweighted sample profiles for productive households (both full and partial completes) achieved in 2022, 2021, 2020 and 2019.
Table 1.4: NTS 2022 Vs. previous years: Tenure
Base: productive sample of full and partial completes, household level, unweighted
| Tenure | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| Own outright | 43.8% | 41.6% | 49.0% | 39.8% |
| Buying it with the help of a mortgage or loan | 25.8% | 29.1% | 26.8% | 26.8% |
| Part own and part rent | 0.9% | 0.7% | 0.8% | 0.7% |
| Rent it | 28.1% | 27.4% | 22.4% | 31.7% |
| Live here rent-free | 1.1% | 1.0% | 0.5% | 0.8% |
| Squatting | 0.0% | 0.0% | 0.0% | 0.0% |
| Non-response (don’t know or refusal) | 0.2% | 0.2% | 0.5% | 0.3% |
| Base | 4,295 | 4,796 | 3,050 | 6,789 |
As in previous years, the changes in methodology were reflected in geodemographic measures for 2022. For instance, Table 1.5 compares responding profiles by IMD deciles (using IMD 2019 for the 2022, 2021 and 2020 samples and IMD 2010 for the 2019 sample). The door-step recruitment approaches (K2N and F2F/PB) led to an increase in the proportion of the sample achieved from the three most deprived groups (deciles 1 to 3) since 2020, bringing the sample profile closer to that seen in 2019 (26.1% in 2019, 19.9% in 2020, 24.7% in 2021, and 25.1% in 2022).
Table 1.5: NTS 2022 Vs. previous years: IMD Rank
Base: productive sample of full and partial completes, household level, unweighted
| IMD decile | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| 1: most deprived 10% | 7.8% | 7.5% | 6.2% | 8.4% |
| 2 | 7.7% | 8.2% | 6.5% | 8.8% |
| 3 | 9.6% | 9.0% | 7.2% | 8.9% |
| 4 | 10.0% | 9.6% | 8.4% | 9.0% |
| 5 | 9.7% | 9.6% | 9.8% | 10.3% |
| 6 | 11.4% | 10.6% | 11.2% | 9.5% |
| 7 | 12.7% | 12.6% | 11.5% | 11.3% |
| 8 | 10.0% | 11.2% | 12.7% | 10.1% |
| 9 | 10.4% | 11.2% | 12.9% | 9.8% |
| 10: least deprived 10% | 10.8% | 10.6% | 13.5% | 9.9% |
| Base | 4,295 | 4,796 | 3,050 | 6,789 |
Note: Figures for NTS 2022, 2021 and 2020 are based on IMD 2019; figures for NTS 2019 are based on IMD 2010.
The profile for Acorn classification followed a similar pattern over the same period. As Table 1.6 shows, the proportion of productive households in the ‘Affluent Achievers’ group was 24.9% in 2019, which increased to 32.4% in 2020, and then moved progressively closer to the 2019 benchmark with 28.2% of the sample in 2021 and then 27.9% in 2022.
Table 1.6: NTS 2022 Vs. previous years: Acorn Group
Base: productive sample of full and partial completes, household level, unweighted
| Acorn category | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| Affluent Achievers | 27.9% | 28.2% | 32.4% | 24.9% |
| Rising Prosperity | 9.2% | 8.9% | 9.7% | 9.0% |
| Comfortable Communities | 29.2% | 29.1% | 29.8% | 27.9% |
| Financially Stretched | 19.2% | 19.2% | 16.9% | 20.9% |
| Urban Adversity | 14.0% | 14.1% | 9.9% | 16.2% |
| Not Private Households | 0.5% | 0.3% | 1.2% | 0.7% |
| Unclassified | 0.0% | 0.0% | 0.0% | 0.4% |
| Base | 4,295 | 4,796 | 3,050 | 6,789 |
Individual-level distribution
The NTS is a household-level survey which aims to interview and receive travel records from all household members (fully productive households must meet this condition). However, since 2020, non-response at the household level has also had an impact on the sample profile across certain individual-level measures.
The biggest impacts were apparent at the beginning of the pandemic in 2020, as the productive sample was older, more educated, and less ethnically diverse than that achieved in 2019. However by 2021 the impacts were reduced, with the distribution of the sample profile more closely resembling the pre-pandemic figures from 2019. In 2022, the sample profile was similar to that seen in 2021, but had deviated slightly further away from the 2019 benchmark on certain measures.
In 2021 and 2022, the sample was older than in 2019, but younger than in 2020. Table 1.7 shows that the proportion of the unweighted sample aged 60 or over had increased from 27.5% in 2019 to 34.2% in 2020. By 2021 this figure had reduced to 29.3%, but rose slightly in 2022 to 31.2%.
Table 1.7: NTS 2022 Vs. previous years: Age
Base: productive sample of full and partial completes, individual level, unweighted
| Age band | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| 0 to 16 years | 18.1% | 18.7% | 17.4% | 20.7% |
| 17 to 20 years | 3.4% | 3.5% | 3.3% | 3.7% |
| 21 to 29 years | 8.6% | 8.7% | 8.1% | 9.5% |
| 30 to 39 years | 12.1% | 13.2% | 10.6% | 13.2% |
| 40 to 49 years | 12.4% | 12.1% | 11.6% | 11.8% |
| 50 to 59 years | 14.2% | 14.6% | 14.7% | 13.7% |
| 60 to 69 years | 13.7% | 12.8% | 15.8% | 12.1% |
| 70 years or older | 17.5% | 16.5% | 18.4% | 15.4% |
| Base | 9,642 | 10,936 | 6,814 | 15,953 |
he achieved samples since the outbreak of COVID-19 have been more educated than the 2019 benchmark for those respondents that either have an educational qualification or a professional, vocational or other work-related qualification, as demonstrated by Table 1.8. Additionally, 28.2% of respondents in 2022 did not have an educational qualification or any professional, vocational or work-related qualifications which resulted in a certificate, as compared to 32.2% in 2019.
Table 1.8 shows that 33.5% of the 2019 sample had a degree-level qualification or higher. This rose to 37.7% in 2020, fell to 36.6% in 2021, then rose again in 2022 to 37.4%. 2022 also had the largest proportion of higher degree and post-graduate respondents of all the survey years since 2019, representing 14.5% of the sample.
Table 1.8: NTS 2022 Vs. previous years: Education
Base: all cases that reported having an educational qualification or a professional, vocational or other work-related qualification, from the productive sample of full and partial completes, individual level, unweighted
| Education category | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| Higher degree or postgraduate qualifications (for example M.A., MSc., M.Ed., Ph.D. etc.) | 14.5% | 13.7% | 13.8% | 12.4% |
| First degree level qualification Degree, or degree level equivalent (for example BA, BSc) | 22.9% | 22.9% | 23.9% | 21.1% |
| Diploma in higher education; HNC; HND; Nursing or Teaching qualification (excluding PGCE) | 12.2% | 11.5% | 13.2% | 11.9% |
| A level; AS level; NVQ level 3; GNVQ Advanced; or equivalent | 18.0% | 18.4% | 16.8% | 18.1% |
| GCSE grade A* to C; O level; CSE grade 1; NVQ level 2; GNVQ intermediate; or equivalent | 20.6% | 21.8% | 20.9% | 22.1% |
| GCSE grade D to G; CSE below grade 1; NVQ level 1; GNVQ Foundation level; or equivalent | 4.9% | 4.4% | 4.2% | 5.1% |
| None of these | 6.5% | 6.6% | 6.6% | 8.9% |
| Non-response (don’t know or refusal) | 0.4% | 0.6% | 0.6% | 0.4% |
| Base | 6,919 | 7,917 | 5,045 | 10,812 |
Table 1.9 shows that the achieved samples from 2022 and 2021 were more closely aligned with the 2019 benchmark on ethnicity and more ethnically diverse than 2020. The proportion of respondents that identified as ‘(White) English, Welsh, Scottish, Northern Irish, or British’ rose from 80.5% in 2019 to 85.6% in 2020. This proportion fell to 81.4% in 2021 and then fell again to 80.0% in 2022, slightly below the 2019 benchmark.
Table 1.9: NTS 2022 Vs. previous years: Ethnic group
Base: productive sample of full and partial completes, individual level, unweighted
| Ethnic group | 2022 | 2021 | 2020 | 2019 |
|---|---|---|---|---|
| English, Welsh, Scottish, Northern Irish, or British | 80.0% | 81.4% | 85.6% | 80.5% |
| Irish | 0.8% | 0.3% | 0.5% | 0.5% |
| Gypsy or Irish Traveller | 0.1% | 0.0% | 0.0% | 0.0% |
| Any other White background | 4.8% | 5.4% | 3.6% | 5.7% |
| White and Black Caribbean | 0.5% | 0.3% | 0.2% | 0.4% |
| White and Black African | 0.1% | 0.2% | 0.3% | 0.2% |
| White and Asian | 0.8% | 0.5% | 0.3% | 0.5% |
| Any other mixed or multiple ethnic background | 0.5% | 0.7% | 0.8% | 0.5% |
| Indian | 3.3% | 2.9% | 2.1% | 2.8% |
| Pakistani | 2.0% | 1.6% | 1.8% | 2.2% |
| Bangladeshi | 0.7% | 1.3% | 0.4% | 0.6% |
| Chinese | 0.7% | 0.6% | 0.5% | 0.5% |
| Any other Asian background | 1.3% | 1.5% | 0.9% | 1.0% |
| African | 2.0% | 1.7% | 1.7% | 2.2% |
| Caribbean | 0.5% | 0.5% | 0.4% | 0.8% |
| Any other Black, African, or Caribbean background | 0.4% | 0.2% | 0.1% | 0.2% |
| Arab | 0.4% | 0.2% | 0.1% | 0.3% |
| Any other ethnic group | 0.9% | 0.8% | 0.5% | 1.1% |
| Non-response (don’t know or refusal) | 0.1% | 0.0% | 0.2% | 0.0% |
| Base | 9,642 | 10,936 | 6,814 | 15,953 |
Note: The categories shown in Table 1.9 represent the options available to the respondents, as shown on the question showcard. Note that the categories deviate slightly from the Census 2021 list of ethnic groups. Firstly, the category ‘Roma’ was not available as an answer option in the 2022 NTS. Secondly, the showcard category ‘Any other Black, African, or Caribbean background’ is represented by a reworded category in the Census 2021 list, namely ‘Any other Black, Black British, or Caribbean background’.
In summary, whilst some differential non-response (differences in response levels between groups within a population) is apparent in all surveys, the survey methodology that was used since the outbreak of COVID-19 has had a noticeable impact on the achieved samples. The move away from P2T methodology to a K2N approach in 2021 helped to move the profile closer to the pre-pandemic profile produced by traditional F2F methods. In 2022, with the transition back to F2F fieldwork (with a telephone back-up), this trend continued in some areas, but stagnated or regressed in others to some extent.
However, due to the improved weighting efficiency mentioned in section 1.3.1, the 2022 sample can be considered to have less bias than the 2021 sample. For instance, for 2022 the weighting efficiency of the interview sample is 88% compared to 63% in 2021. This degree of weighting efficiency is much more in line with pre-pandemic levels, for example 92% in 2019. Furthermore, differences between the distribution of the 2021 and 2022 sample profiles for individual characteristics are likely to be driven by random variation and are unlikely to be statistically significant.
The combined evidence suggests that mode-specific biases in the data largely reduced over time as the methodology progressively became more similar to that used historically on the NTS. However, even with the return to F2F, some of these biases remained apparent in 2022.
1.4 Additional analysis of F2F cases
As noted above, the sample profile achieved during 2022 (and in part due to the return to F2F fieldwork) was broadly in line with that seen in 2021 and also the general trend of moving closer towards pre-pandemic levels, at least on certain measures. Additionally, the reintroduction of the ‘fixed travel weeks’ allocation method also had a role to play in improving the quality of the data in 2022. At the same time though, the potential implications of having run a telephone back-up approach in tandem with the F2F fieldwork also needs to be considered. The following section summarises some fieldwork measures that indicate how the data was impacted in 2022.
1.4.1 Distribution of travel week start days
The challenges of running the survey during the pandemic and, for the most part, relying on telephone interviews, meant that changes to the traditional design were introduced for practical reasons intended to ensure the continuity of the study. Changes to the travel week allocation method was one such example.
In quarter one of 2022 (as in 2021 and 2020), the travel weeks were assigned so that they started the day before the placement interview. This approach, applied to all cases during quarter one (and also any P2T cases conducted at any point during the full survey year), was known as the ‘rolling travel weeks’ allocation method. The timing of the travel week start days was therefore driven by when interviewers could schedule the placement interview with a responding household. This had implications for the distribution of travel week start days across the week, and consequently the data collected.
Under the traditional method of ‘fixed travel weeks’ used before the pandemic, the spread of start days would be expected to be relatively even over (a) the dates of the month and (b) the days of the week, due to the random element that was central to the allocation method. For each point, a list of fixed travel weeks was randomly selected for the given fieldwork period, spanning approximately four weeks. Interviewers were instructed to allocate travel weeks to addresses from their list according to a strict, methodical process designed to enhance randomisation and minimise bias that might be introduced by either the respondents or the interviewers influencing when the travel week would start. This allocation method was reintroduced for all F2F (and PB) cases from quarter two onwards.
For fieldwork using the ‘rolling travel weeks’ method (that is, K2N and P2T), bias was more likely to occur. While the spread for quarter one of 2022 was relatively even for travel weeks starting Monday, Tuesday, Wednesday, Thursday and Sunday (with slightly fewer travel weeks starting on Thursdays, and slightly more starting on Sundays), travel weeks starting on Fridays and Saturdays were notably under-represented. This was because interviews on Saturdays and Sundays were less common than interviews taking place on the other days of the week.
Figure 1.1 demonstrates that when fieldwork returned to F2F (with the PB) from April onwards the spread of travel week start days was relatively even overall, as would be expected using the ‘fixed travel weeks’ approach. The figures for travel weeks starting Friday and Saturday were noticeably improved as compared to quarter one, and more closely resembled the other days of the week. It should be noted though that for any cases that were conducted P2T (at any point during the survey year) the ‘rolling travel week’ approach was used and could explain why the figures for Friday and Saturday did not fully recover during quarters two to four to match the other days of the week. Additionally, the proportion of travel weeks starting Sunday were slightly lower than average for quarters two to four, in contrast to quarter one when Sundays were the most frequently occurring travel week start day.
Figure 1.1: Percentage of travel week start days distributed over the days of the week, split by K2N fieldwork period (quarter one) and F2F and PB fieldwork period (quarters two to four), figures inclusive of cases conducted by P2T
Base: productive cases (full and partial completes), household level, unweighted, including cases with missing weighting information
(Alt text for Figure 1.1) This bar chart shows how the travel week start days were distributed over the seven days of the week. Each day of the week (Monday to Sunday) is represented by two bars on the horizontal axis: one bar for the K2N fieldwork period (quarter one) and one bar for F2F/PB fieldwork period (quarters two to four). The vertical axis represents the percentage of cases for each fieldwork approach. The chart shows that for the K2N fieldwork period there were noticeably fewer travel weeks starting on Friday and Saturday, with travel week start days relatively evenly spread over the other days of the week. For the F2F/PB fieldwork period, travel week start days were spread evenly with little variation across the days of the week.
When fieldwork returned to F2F from April onwards the spread of travel week start dates over the days of the month was relatively even, as would be expected using the ‘fixed travel weeks’ approach, whereas quarter one (K2N) was less evenly spread.
Given the more even spread of the travel week start days and dates for quarter two onwards, it would suggest that the return to F2F helped to improve the quality of the data for the 2022 survey.
1.4.2 Split of F2F Vs. telephone back-up interviews
As highlighted previously, F2F interviewing was introduced at the beginning of April 2022. However, acknowledging that COVID-19 remained a concern for some members of the public, fieldwork during this period allowed for a telephone interview option as a back-up (known as the phone back-up, or PB). As noted above, interviewers were expressly briefed to prioritise F2F interviewing wherever possible. However, interviewers were able to make use of the PB and complete the interview over the phone if the household met the criteria outlined in section 1.2.2.
Therefore, in the final three quarters of the 2022 survey there was scope for mode effects to impact the data with the two modes running in tandem. Table 1.10 shows the proportion of productive cases (both fully and partially productive) in each month that completed the survey F2F and those that completed via the phone back-up (PB).
Table 1.10: Proportion of cases completing the survey F2F and on the phone
Base: productive sample (full and partial completes), household level, unweighted, including cases with missing weighting information, for April to December
| Month | F2F | PB |
|---|---|---|
| April | 83.9% | 15.9% |
| May | 82.5% | 17.5% |
| June | 88.1% | 11.7% |
| July | 86.1% | 13.6% |
| August | 87.0% | 12.4% |
| September | 88.6% | 10.0% |
| October | 90.4% | 9.6% |
| November | 89.8% | 10.2% |
| December | 88.1% | 11.9% |
| Total | 87.0% | 12.7% |
Note: in some rare cases (0.3% overall) interviewers did not record the mode used in the expected manner, so figures in the above table may not total to the full 100%.
Most cases completed the survey F2F (87.0% overall for quarters two to four), with the proportion of F2F cases generally increasing as the year went on. The telephone back-up was used most in May (17.5%) and fell to the lowest point in October (9.6%). This suggests that any mode effects caused by the multiple methodologies were most likely minimal.
1.5 Change in weighting design
Despite a lower response rate than pre-pandemic, the overall sample profile has moved closer to the profile from before the pandemic compared to 2020 and to some extent 2021. Due to the balance of K2N, F2F/PB, and P2T data collection methods, and differing response rates by quarter, some challenges for weighting the 2022 data needed to be addressed. The weighting design for 2022 was based on the historical one, with some updates to account for:
- the two concurrent data collection methods in 2022 (K2N and F2F/PB running concurrently with P2T) each with very different response rates and sample profiles
- the uneven and non-random distribution of modes (K2N, F2F/PB, and P2T) across the year, as the modes were not randomly distributed geographically or across the year, which biased the achieved sample
- the lower response rate in quarter one
- the continued use of rolling travel weeks for some cases
The 2022 weighting design was updated to account for this to ensure that the overall weighted sample profiles aligned with population estimates.
1.5.1 Interview and fully responding weights
To account for the changes in the interview and fully responding weights, additional measures were taken. Firstly, the non-response model for NTS 2022 was split by mode in order to account for differing response rates and proportions between K2N, F2F/PB, and P2T cases. The resulting weights were scaled to their respective sample sizes within each mode. Secondly, a calibration to 25% of the weights per quarter was done to account for differing response rates across the year.
The level of corrective weighting required determines the sizes of standard errors and confidence intervals. Due to less bias in the sample, the lower proportion of P2T cases as well as the adjustments to the weighting scheme above, for 2022 the efficiency estimates were much higher than 2021. To illustrate this point:
- in 2022 the interview weights have 88% weighting efficiency compared to 63% in 2021 and 92% in 2019 (pre-pandemic)
- in 2022 the effective sample size for the interview weights was 3,801, compared to 3,011 in 2021 and 6,239 in 2019, meaning that the effective sample size has risen for 2022 compared to the previous year, despite having a lower response rate overall
1.5.2 Diary weights
The continued use of rolling travel weeks throughout 2022, albeit to a lesser extent than 2021 and 2020, resulted in an uneven distribution of diary start days across the week. In order to adjust for this potential source of bias, the diary drop-off weighting incorporated rescaling so that each start day of the week had approximately 14% of the total. Prior to this rescaling, the range was 11.0% to 16.3% between start days of the week, as fewer diaries began on Fridays and Saturdays. This adjustment was also used in previous NTS years when rolling travel weeks were in use.
These adjustments reduced the efficiency of the diary weights. In NTS 2022 this was 84% for the journey-level diary drop-off weights, compared with 56% in NTS 2021 and 91% in NTS 2019.
1.5.3 CASI weights
For the CASI weights, it was possible to largely return to the pre-pandemic weighting approach. That is, weights were not calibrated separately by quarter or half of the year. To even the sample out across the year, quarters were instead included in the non-response model, and the sample was subsequently calibrated as a whole to the adult population estimates for age by sex and region (with the adult population defined as those people aged 16 years or over).
The NTS 2022 CASI weights have 51% efficiency and an effective sample size of 2,084. This compares with 40% efficiency and an effective sample size of 1,135 in NTS 2020 and 55% and 3,565 in NTS 2019.
These limitations need to be kept in mind when analysing the data.
Chapter 4 – Fieldwork Procedures and Response Rate
Instructions for printing and saving
Depending on which browser you use and the type of device you use (such as a mobile or laptop) these instructions may vary.
You will find your print and save options in your browser’s menu. You may also have other options available on your device. Tablets and mobile device instructions will be specific to the make and model of the device.
How to search
Select Ctrl and F on a Windows laptop or Command and F on a Mac
This will open a search box in the top right-hand corner of the page. Type the word you are looking for in the search bar and press enter.
Your browser will highlight the word, usually in yellow, wherever it appears on the page. Press enter to move to the next place it appears.
Further information
National Travel Survey statistics
Email [email protected]