A comparative study of national travel surveys in six European countries

ABSTRACT This paper aims to show how changes in survey design, supervision, and data collector affect travel survey data. The results can be used by policymakers and travel behaviour researchers when evaluating or choosing survey design. A document study of public documentation of national travel survey (NTS) methodology, and research using NTS data material from six European countries were conducted to compare the survey design of their respective NTSs. The results show that all countries included in this analysis have some sort of quality challenges. However, the countries have handled them differently, at different stages, and the transparency of the data collection process varies. Transparency in the data collection process and processing of data is essential for improving the survey design, evaluating quality, and ensuring that the time series is intact.


Introduction
National travel surveys (NTSs) monitor and describe mobility.The practical approach (sampling, recruitment, data collection tool, weighting, etc.), varies between countries.A country's institutional context can affect the continuity and scope of its NTS (Kunert, Kloas, and Kuhfeld 2002).Armoogum et al. (2014) divide NTSs into household (NHTS) and individual travel surveys (NITS).Countries with an NHTS include England, Germany, Spain, Switzerland, and Belgium, while Norway, Sweden, Denmark, Finland, Italy, and the Netherlands have NITS (Armoogum et al. 2014).In NITSs, information about the household is collected through background questions.Some NHTSs collect travel behaviour data from all members.Some NTSs, e.g. the French NTS, sample households and use the household as a statistical unit, but only collect trip data from one member (Armoogum et al. 2014).NTS survey methods are shaped by local traditions, adapting to the conditions for data collection and the available technology (Wittwer et al. 2018).This makes NTS comparison challenging, although there have been attempts, e.g.Kunert, Kloas, and Kuhfeld's (2002) comparative study of NTS practices and the SHANTI project (Armoogum et al. 2014).
Although NTS survey designs vary, combining data collection methods is increasingly popular.Multi-mode surveys are being used to improve data quality and reduce costs (Morris and Adler 2003;Paskota 2006).The problem with combining data collection methods is the risk of having mode effects, i.e. the data collection method affects the responses (de Vaus 2014).On the question of using new technology (e.g.GPS devices, smartphones) in travel surveys, Murakami, Morris, and Arce (2003) underlined the importance of data comparability, and that implementing new technologies requires that questions, definitions, and coding are consistent.Furthermore, the technology of data collection must not affect travel behaviour.Bonnel (2003) discussed face-to-face, postal, and telephone survey data comparability, finding certain groups more and less inclined to participate, depending on the survey mode.This mode effect also affects collected travel behaviour.Bayart and Bonnel (2015) compared web, telephone, and face-toface respondents, and identified differences in socio-demographics and travel behaviour.Some travel behaviour researchers have adopted new sources of data to study travel behaviour, such as e.g.cell tower data (Lee and Sener 2020) or mobile phone apps (Hubrich et al. 2020;Raturi et al. 2021).However, such a radical switch could have negative consequences on representativity (Livingston et al. 2021) and participation (Svaboe, Tørset, and Lohne 2021).Furthermore, data comparability and continuing the time series are important.Thus, NTS methodology has remained quite traditional.
Regarding the quality of travel survey data, most of the literature addresses one problem at a time.Fewer take an overview approach, looking at several indicators at once or comparing countries.A study of the Norwegian NTS identified challenges connected to survey design choices (Svaboe, Tørset, and Lohne 2024).This inspired looking into methods and quality indicators of NTSs in similar countries, to identify ways to improve future NTS performance.Thus, we have studied how different countries have taken measures to minimise errors within their NTS, aiming to learn from their experiences to advise on future NTS survey designs.In this paper the following research questions are addressed: (1) How do countries conduct their NTSs?(2) What challenges do they experience concerning quality?
(3) What measures can be envisioned to mediate the challenges?
To answer these, a document study was conducted into the NTSs of Norway, Sweden, Denmark, England, France, and Germany.The aim is to study quality development and different countries' choices in the face of challenges over time.

Background
The travel behaviour researcher must deal with a trade-off between quality, quantity, and cost (Richardson, Ampt, and Meyburg 1996).If the budget is fixed, one must choose between either higher quantity, lower-quality data or lower-quantity, higher-quality data.Currently, there are no official international standards of quality for NTSs.From a survey methodology perspective, quality is usually understood as minimising errors (Groves et al. 2009;Dillman, Jolene, and Christian 2014;Richardson and Lawton 2013).According to Richardson (2000), there are three main sources of systematic error that the mobility researcher needs to be aware of when conducting a travel survey: non-response (the respondent does not participate), non-reporting (the responses for a respondent are incomplete) and inaccurate reporting (the responses are deemed incorrect by the researcher).Traditionally, non-response has received the most attention since it can cause non-response bias.Underreporting effects and other response biases should be taken seriously because representativity is challenged in a biased sample (Svaboe, Tørset, and Lohne 2024) and missing or incorrect information can lead to the loss of an observation (Wilmot and Adler 2003).In NTSs, the known extent of influence is based on public reports and documentation.
Stopher and colleagues carried out significant work on guidelines, standards, assessments of quality, and standardisation in the 2000s (Stopher and Jones 2003;Stopher et al. 2006).Here, we study the response rate and two transportation-specific measures of quality (trip frequency and share of immobiles/no-trip respondents).For this paper, it is necessary to get insight into the following terms: response rate, transportationspecific measures of quality, and representativity in travel surveys.

Response rate
The response rate is the most used quality indicator in surveys.A low response rate is not a problem if those who respond are representative, and the response rate should not be used as the sole estimator of quality.The problem arises if the non-respondents travel differently than the respondents, i.e. sample selection bias.It is necessary to have a plan for non-response issues from the beginning of the survey process because nonresponse is connected to the overall systemic approach of the travel survey (Richardson, Ampt, and Meyburg 1996).The probability of having non-response bias decreases with higher response rates (Stopher et al. 2006).Different survey methods can give rise to different ways of estimating the response rate (Richardson, Ampt, and Meyburg 1996).
According to Richardson, Ampt, and Meyburg (1996), respondents and non-respondents are different, both when it comes to sociodemographic characteristics and travel behaviour characteristics.Furthermore, the authors found that respondents who participate at first contact and those who respond after reminders and follow-ups travel differently.Paskota (2006) found that non-response was higher in urban areas than in rural areas.

Transportation-specific measures of quality
Transportation-specific measures of quality include, among others, the proportion of immobiles in a survey and the average trip rate per person or household (Stopher and Jones 2003).These variables are considered temporally and spatially stable, and thus should not vary much between surveys (Stopher et al. 2006).A prevalent assumption is that a high percentage of immobile respondents indicates poor survey technique (Stopher et al. 2006;Madre, Axhausen, and Brög 2007).If a respondent reports no trips to reduce their respondent burden, it can be understood as a soft refusal (Madre, Axhausen, and Brög 2007).Further, mobile respondents reporting immobility leads to underreporting of trips.Nevertheless, some respondents will always be immobile on the day of reporting.According to Madre, Axhausen, and Brög (2007), between 8-12 per cent of immobile respondents should be expected in a single-day travel diary survey.Stopher et al. (2006) found that non-mobile rates should be around 20 per cent at the person level and 1 per cent at the household level.Higher rates indicate belowaverage quality and lower rates indicate higher quality.The share of non-mobile respondents is a problematic quality indicator because travel behaviour can vary between countries.
Trip item non-response, meaning failing to obtain correct information from respondents on their trips, is a common challenge in travel surveys (Wolf et al. 2003).According to Brög et al. (1982), there are three reasons trips go unreported: (a) respondent exhaustion, (b) respondent finding certain trips redundant or forgetting them, and (c) deliberate omittance of trips.Richardson, Ampt, and Meyburg (1996) later added that some trips go unreported because respondents deem them too short, 'unimportant', or not worth mentioning because they are walking or bicycle trips.The non-reported trips are mainly short trips, by non-motorized modes, with a discretionary nature.
According to Sammer et al. (2018), up to 30 per cent of trips per day for passenger traffic are underreported, and the survey method is a part of the cause.Based on a pilot comparing GPS with traditional travel diary survey methodology conducted in Victoria, Australia, respondents generally underreported trips, underestimated the distance, and overestimated the trip duration in traditional travel surveys (Stopher and Greaves 2010).Kelly et al. (2013) found similar results when reviewing papers on GPS-tracked travel and self-reported trips.They found that respondents underreport trip frequency in self-completion questionnaires but overestimate the duration of the trips.The underreporting of trips mainly concerns short trips, with the modes of walking, cycling, or car, and the purposes are mainly shopping and leisure (Sammer et al. 2018).Underreporting of trips is also connected to the representativity of the sample: non-mobile persons are, in general, more often available for interviews.

Representativity
A commonly used quality indicator is the geographic and demographic representativity of the data (Armoogum, Ellison, and Kalter 2018).Socioeconomic representativity is important because characteristics such as age, gender, education, and professional status have been proven to affect travel behaviour (Domencich and McFadden 1975;Ben-Akiva and Lerman 1985).According to Armoogum, Ellison, and Kalter (2018), representativity challenges mainly stem from total non-response, measurement errors, and the sampling frame.

Method
This paper presents a meta-analysis of the current trends in methodology and quality indicators of the NTSs in six European countries, mainly based on a document study.The meta-analysis is inspired by the approach outlined by Glass (1976), i.e. an 'analysis of analyses'.We chose a pragmatic approach (Robson 2002) to data collection because collecting transportation-specific measures of quality for multiple countries and years proved challenging without 'inside access'.

Selection criteria for national travel surveys
The following selection criteria were set when choosing countries: . Data collection is conducted on a national scale .Open access documentation of NTS survey design .Similarities in culture and policy (hence the limitation to Western European countries) Based on these criteria, the following countries were chosen: Norway, Sweden, Denmark, England, France, and Germany.There are two large-scale national travel surveys in Germany: the German Mobility Panel (MOP) and Mobility in Germany (MiD).We have chosen MiD as the case for Germany because MiD is, like the other NTSs, cross-sectional.

Data collection
The document study was split into two phases.First, the public documentation of the methodology of the respective country's NTS was collected from the official web pages of the data collector/project group/government authority.Secondly, information about NTS methodology was collected from peer-reviewed papers and reports which use the NTS data (see Appendix A for a list of materials).
Results affected by COVID-19, mainly from 2020 onwards, are excluded from the analysis.

Quality in national travel surveys
In this paper, we study the methodological choices made and how countries attempt to minimise errors in their NTS.Thus, we describe how the NTSs are conducted to provide an overall picture of the survey design choices.When assessing quality, we discuss the following measures of quality: . Response rates .Two transport-specific measures of quality: immobility and trip frequency .Geographic and socioeconomic representativity This is not an exhaustive list of quality indicators and having issues in one indicator does not necessarily mean that the NTS is of low quality.However, they are important factors when discussing NTS quality.They are also the most accessible in public documentation.
The documentation and reporting systems' detail levels vary between years and countries.This made it challenging to obtain precise data on response rates and transportation-specific measures of quality for certain countries and years.Consequently, own estimations (OE) were carried out (see Appendix B for description).Table 1 shows the information gathered and when OE was necessary.X indicates information was collected for at least one year.If multiple versions of the same estimate due to changing practices were identified (e.g.changing weighting procedures), the most recent estimate was included.A description of how Tables 2-7 were constructed is included in Appendix B.

Limitations and solutions
To our knowledge, none of the countries have made publicly available all the quality indicators used for this analysis.In addition, differences in definitions, reporting practices, and methodologies across countries exist.Therefore, rather than a direct comparison of key figures, the development within countries needs assessing.

Norway (NITS)
The Norwegian transport authorities cooperate in overseeing the NITS, and the first data collection was conducted in 1984/85, with new data collection approximately every four years until 2014 (Hjorthol, Engebretsen, and Uteng 2014).Data collection  is now continuous to follow the development and be more flexible concerning add-on samples (Grue, Landa-Mata, and Flotve 2021; Opinion AS 2021).The respondents can choose between computer-assisted web interviewing (CAWI) or computerassisted telephone interviewing (CATI) (the first invitation is to a CAWI survey).Each respondent is asked to report all trips from the day before the interview and give background information, including access to transport modes and car ownership.The national sample is asked about long trips (above 100 km), but this is optional for the regional add-ons (Vågane, Brechan, and Hjorthol 2011).Reminders and multiple attempts of contact have long been used, although the number varies (Vågane, Brechan, and Hjorthol 2011;Denstadli et al. 2006;Hjorthol, Engebretsen, and Uteng 2014;Stangeby, Haukeland, and Skogli 1999;Vibe 1993;Denstadli and Hjorthol 2002).
Around 2016, three policy changes were made that affected the future of the NITS: (a) a multi-mode solution was implemented (CAWI + CATI), (b) continuous data collection, and (c) the NITS was put out to tender, which led to a change in data collector and quality controller.Between 1984 and 2016, the Institute of Transport Economics (TØI) was responsible for the data collection.The commercial survey company Epinion won the tender in 2016 and was responsible for data collection until April 2020, when they filed for bankruptcy.Opinion, a similar company to Epinion, has been responsible for data collection and quality control since 2020 (Opinion AS 2021).A smartphone/big data solution is currently being evaluated for the Norwegian NTS (Kogenta and Opinion AS 2022).
Except for a small increase between 1998 and 2001, the response rate and representativity have been steadily declining in Norway.Based on the documentation available, it appears that the small increase was due to a larger focus on the recruitment process in 2001.2001 was the first year with an advance letter, and according to the key reports, 'special measures' were made to reach under-reporting age groups (youngest and eldest).Only those older than 80 years were underrepresented in 2001 (Denstadli and Hjorthol 2002;Denstadli et al. 2006).People in their 20s received a written inquiry before contact, their parent's phone number was collected if the sampled person did not have a registered phone number, and the person was interviewed on their cell phone (when possible) if they were otherwise unavailable (Denstadli and Hjorthol 2002).The elderly received a specially adapted information letter and were followed up when necessary.In 2005, they continued with a designated information letter for the elderly, oversampled people in the age groups 20-29 and above 65 years, monitored the distribution of responses, and followed up if necessary, but were unable to get an equally high response rate (Denstadli et al. 2006).
Due to more add-on samples and increasing non-response, the proportion between the national sample and supplementary samples changed dramatically between 2001 and 2019 (Grue, Landa-Mata, and Flotve 2021).In 2018/2019, the national sample constituted approximately five per cent of the data material (ibid.).This challenges the geographical representativity of the NITS sample.Regarding trip frequency, there has been a decline in the average number of trips.According to Grue, Landa-Mata, and Flotve (2021), trip reductions result from a change in methodology, not travel behaviour.They identified this by studying the complexity of the trip chains.

Sweden (NITS)
Trafikanalys is responsible for Sweden's NITS (RVU Sverige).Sweden currently has a multi-mode solution where the respondent reports trips for one day, and data collection is continuous.Sweden started to conduct travel surveys in 1978 and began continuous data collection in 1994.In 2001, the NITS was redesigned, and the questionnaire was shortened.The questionnaire was redesigned again in 2005, 2011, and 2019(Viklund 2019)).The project New solutions for future travel surveys was started in 2016 due to high respondent burden, high costs, and coverage issues (Saxton 2018b).Trafikanalys conducted a stakeholder analysis and pilots using new methods, including a mobile app and mobile network data.According to Trafikanalys (Saxton 2018b), the new methods were promising, but could not replace traditional methods.
Non-response has been a concern in the Swedish NTS (Saxton 2017; Viklund 2019).In the period 2017-2018, Trafikanalys decided to stop data collection completely to work on methodology development.The main reasons for stopping data collection were the low response rate and differing travel behaviour in their data from other data sources (Saxton 2018a).
In 2019 the NITS was redesigned again.The data collection method was changed from using CATI to a combination of CAWI and mail-back PAPI (Viklund 2019).One reason for this was the increasing difficulty of getting in touch with people on the telephone.They redesigned the questionnaire to make it more relevant for the respondent, reduce drop-out during reporting and increase cost-efficiency (Saxton 2018a).[1975][1976][1977][1978][1979][1980][1981][1982][1983][1984][1985][1986], and the first TU was an HTS.In the second generation (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003), there were some variations in data quality and methodology.In 1998, a decline in kilometres travelled and trip frequency was observed (Ortúzar et al. 2011).According to Christensen (2006), the reason for this was an increase in no-trip days among the respondents.Some of it could also be explained by a reduction in mobility among those making trips.In 2004, TU was set on hold due to financing issues (Christiansen and Skougaard The current TU is an NITS where each respondent reports trips for 1 day (the day before the interview).Data collection is continuous and collects information for the entire year and captures seasonal variations.The respondents are sampled from the Centrale Personregister (CPR).There is currently a multi-mode solution (CAWI and CATI) for reporting trips (Center for Transport Analytics 2019).The respondents receive an invitation to participate in a CAWI online (a system named e-boks) or as a paper letter.If they do not report their trips, an interviewer will call them and conduct a telephone interview.Epinion (data collector), seeks contact for a maximum of 3 weeks (Center for Transport Analytics 2022).For the period 2016today, 20 per cent reported using CAWI and the average interview duration was 20 min.80 per cent reported their trips using a CATI and the average response time was 11 min.

England (NHTS)
According to Armoogum et al. (2014)  The interviews are conducted face-to-face and involve two visits (Department for Transport 2020c).On the first visit, the household is recruited to participate, asked background questions, and given the self-administration diary.On the second visit, the interviewer collects the diaries and helps clarify if necessary.All members are interviewed about their trips for seven days (Department for Transport 2020c).Stamps are included in the invitation letter to encourage participation, and each respondent gets a 5-pound gift card if all household members complete the survey (Cornick et al. 2020).
In 2011, a GPS pilot was conducted with an NHTS subsample to reduce respondent burden and costs (Sneade 2011).However, the pilot resulted in a lower response rate and there was a concern that the data obtained were not comparable with traditional diary survey data.Thus, it was decided to continue using traditional methods.Since 2015, the response rate has been steadily declining (NatCen Social Research 2019b).According to the National Centre for Social Research (NatCen) (2019b), this trend can be explained by an increase in refusal rates.The Department for Transport (DfT) implemented measures to stop the declining response rates, including studying the use of incentives and testing advance letter design (NatCen Social Research 2019a), reviewing and piloting alternatives to the traditional paper diary (Rofique, Humphrey, and Killpack 2011;Anderson et al. 2009;Evans et al. 2020;Sneade 2013) and requesting respondent feedback (Department for Transport 2019).
Another issue in the English NTS is that short walks are underrepresented (Department for Transport 2020b).According to the DfT (2020b), the reason for this was the interview structure.As a result, in 2016, they tested collecting information on half of the short walks on day one.Since 2017, information on short walks is collected on day one.Weights were also created to correct for underreporting.The weighting strategy is also supposed to adjust for non-response bias (Cornick et al. 2020).
According to Armoogum et al. (2014), the share of immobiles for the period 2006-2010 in Great Britain was 22 per cent (unweighted).According to Motte-Baumvol, Fen-Chong, and Bonin (2022), 22.7 per cent of individuals in the UK NHTS are immobile, and during the one week of data collection, 52.7 per cent of respondents are immobile for one or more 24-hour periods.Suggested explanations for the high share of immobiles are soft refusal (fatigue) (Armoogum et al. 2014), and individual variability of trips (Motte-Baumvol, Fen-Chong, and Bonin 2022).
Regarding geographical representativity, the NHTS is not designed for analysing smaller geographies than regions (Department for Transport 2020b).Furthermore, the data is weighed to make the sample representative of England, making it problematic to analyse it on a disaggregated level.Still, response rates are much lower in London compared to the rest of the country (Department for Transport 2020b).Thus, the NTS oversamples London to get a big enough sample size so that it is proportional to the population.

France (NITS)
The first NTS in France was conducted in 1966-67, and it had two purposes (Armoogum et al. 2014).The main purpose was to get a description of travel behaviour nationwide.The second purpose was to gather information for planning and development.The French National Bureau of Statistics was responsible for the survey, and the interview involved a minimum of two visits for each household.Since 1976, a detailed methodology for conducting the survey was developed, called the 'CERTU standard', which was later standardised by Cerema (Richard and Rabaud 2018).
The respondents receive an invitation letter before data collection (Armoogum et al. 2014).On the first visit, a face-to-face interview is conducted to recruit and gather background information from the entire household.Then, one member of the household is sampled and asked about their trips in the second interview.Since it only collects trip data from one household member, we treat it as a NITS in this paper (it is usually referred to as an HTS since it samples households).The French NITS is conducted with longer time intervals between data collection than the other countries included in this paper.According to Armoogum et al. (2014), this is not a problem, given that changes in behaviour are slow.
Some changes have been made to the survey design.From 1993 to 1994, the respondents received a self-administered vehicle diary.The sample frame has changed two times since the NHTS inception.From 2007-2008, respondents were given the option to use a GPS device (Armoogum et al. 2014).However, according to Armoogum et al. (2014), a GPS solution will not be a part of future data collections.Unfortunately, there is little open access information about future French NITS plans.

Germany (NHTS)
HTSs have been conducted on behalf of the German Federal Ministry of transport since the mid-1970s (Gruschwitz et al. 2018.MiD is a continuation of KONTIV (1976KONTIV ( , 1982KONTIV ( , and 1989)).MiDs are cross-sectional surveys, and the main purpose is to collect information on everyday mobility for Germany as a whole, but also the 16 federal states.The respondents are randomly selected and assigned a day for reporting trips.Respondents are also asked to provide socio-demographic background information about the household members and ownership and availability of different modes of transport.Regional addon samples were made available for MiD 2017.Thus, local authorities, transport associations, and federal states had the option to enlarge their sample.Since 2017, three sampling frames have been used: postal address, landline, and mobile phone number.Results from MiD are used by the German government in transport planning, policy decisions, and traffic modelling (Hauslbauer, Schade, and Petzoldt 2022).
In MiD, all members of the household are asked to complete the survey, and at least 50 per cent of the household needs to complete the questionnaires for it to be included in the  (Gruschwitz et al. 2018).In the first part, the respondents are interviewed about background information and car ownership.In the second part, the respondents are individually interviewed to obtain more detailed information about each household member, and they are asked about their trips for the reporting day.The respondent can choose between CATI, CAWI, and PAPI.They can even change instruments between the two parts.However, CAWI is always presented first.The purpose of providing multiple options is to ease the respondent's participation and thus increase the response rate.Reminders are also used to improve the response.Gruschwitz et al. (2018) studied the potential survey mode effects in MiD by comparing key numbers of mobility between the different types of survey instruments.They found that respondents report fewer trips when they use survey modes without an interviewer (CAWI, PAPI).

Discussion
In this paper, we set out to address the following research questions: (1) How do countries conduct their NTSs?, (2) What challenges do they experience concerning quality?, and (3) What measures can be envisioned to mediate the challenges?In the following sections, each of them are discussed.

Current situation
The NTSs studied, except Denmark, are mainly funded by governmental authorities.Furthermore, competitive tendering is common in the data collection process.Usually, Major survey design changes are rare, mainly because they would affect the time series.Continuous data collection is the preferred method for NTSs, although France is an outlier here.CATI is increasingly being replaced by CAWI.Except for the Scandinavian countries, the NTSs mainly aim to collect information on travel behaviour at a national level.In general, the cities get oversampled, either due to lower response rates (England), or the incentivization of add-ons (Germany and Norway).If there are local travel surveys to capture the travel behaviour in smaller municipalities (rural vs. urban travel behaviour), this is not necessarily a problem.However, in the case of Norway where most of the previous local TSs are now part of the NTS, this could jeopardize the quality of the NTS.
Currently, there are no formalised standardisation approaches to evaluate data quality beyond the GDPR and national guidelines and regulations.There are private agencies that provide guidance on data handling and processing.However, the use of these agencies is mostly optional.External quality control is rarely used, and the data collector and data controller are often the same.Thus, quality control is dependent on the data collector having the knowledge, resources, and motivation to do a sufficient evaluation.

Challenges
Non-response and representativity issues are present in all countries and ensuring representativity is becoming increasingly difficult.All countries have faced challenges with falling response rates.However, the countries still using face-to-face interviews have fewer problems.Underreporting, especially of short trips, is also a common problem.Immobility rates are increasing in Sweden and Denmark based on the rates suggested by Madre, Axhausen, and Brög (2007), but less so according to rates reported by Stopher et al. (2006).In Norway, a higher number of immobiles was found in the CATI compared with CAWI interviews.There are indications of a problematic high share of immobiles in England.
Another challenge identified is the negative side effects of combining methods.Potential mode effects have been identified in Norway and Germany.Introducing CAWI mainly increases the responses of young people and those with a high level of education.Increasing the share of young respondents will probably increase representativity, but highly educated people are rarely under-represented in surveys.Thus, decreasing the bias of one population segment in the sample could potentially increase the bias of another.
Identifying whether a change in travel behaviour is due to survey design changes or changes in travel behaviour can be challenging and is rarely discussed.There are no international standards on this in travel behaviour research.Usually, these issues are identified by studying if the travel behaviour change is reasonable (i.e. it is evaluated based on best practices in transportation research and/or survey research).However, the results of such a validation are either (a) rarely open access or (b) not done.

Measures
Norway, Sweden, Denmark, and Germany have implemented a multi-mode solution to meet the challenges of non-response.All countries are either considering, planning, or have tested a GPS device or smartphone tracking solution as a supplement to or substitution for traditional methods.Still, none of the countries have taken the full step of implementing the methodology.Implementing GPS devices or smartphones has some challenges with a potential increase in non-response and technical issues.Because tracking automatically registers trips, rather than relying on memory, the trip rate is generally higher compared to traditional data collection.Thus, a transfer would potentially disrupt the time series.
England has gone the pragmatic route by mediating for underreporting of trips, and other countries could perhaps learn from this.However, the English NTS and e.g.Norwegian NTS are quite different in survey design and the solutions for England might not work or be relevant for Norway.For example, moving the interview day for short trips is not an applicable solution for a 1-day cross-sectional NTS.However, the work England has done in collecting information about the respondent's experiences, and their use of incentives (such as a small gift included in the invitation letter) could be useful in the other NTSs in the context of non-response mediation.
Interestingly, the methodology and survey design in Norway and Denmark are almost identical, but the response rate in Denmark is higher.Potential explanations are (a) lower reported average response time in Denmark, (b) heavier use of reminders in Denmark and/or (c) that the Danish CAWI administration tool is more user-friendly.However, neither country has a formula or description of how the response rate is estimated.Thus, the difference could be due to different response rate formulas.

Transparency
Transparency is important and reporting challenges should be rewarded.Thus, comparing countries having multiple documented issues and high transparency to countries with fewer documented issues and low transparency is challenging.The lack of discussion on potential mode effects, sparse documentation on sampling plan and how to deal with underrepresented groups, and omitting response rates from the public documentation is perhaps indicating some problems.Overall, there seems to be a connection between the volume of documentation and transparency.Potential explanations for omitting information are concerns regarding future tenders, funding, or the same institution being both data collector and quality controller.However, it could be that there are no challenges to document.Still, this only highlights the need for clear documentation guidelines.
It is problematic to range from best to worst on transparency because there is room for improvement in all countries studied.For all countries, most of the detailed meta-data were written in their respective language, making information less available internationally.Furthermore, there is a balance between making key information too complicated and describing too little with vague descriptions; the Danish NTS is described in an easy-to-understand way on DTU's web page, but more detailed information, such as response rates, was less accessible.Easy-access information on English trends of immobile respondents is limited.Norway is a good example of how problems arise when the same institution is data collector and quality controller.The 2018/2019 report is the only report written by a different institution than the data collector, and this is reflected in the documents.If the documentation producer is concerned with winning the next tender, they are not incentivized to be transparent about flaws in the data.

Limitations
The limitation of doing a document study is that the official documents are not neutral and reflect the organisation's representations of themselves.The transparency and data quality varies and not all countries report their flaws openly.Furthermore, structural changes (e.g.change of data collector or quality controller) are reflected in the documentation.Interviewing those responsible for the NTS data collection process could have supplemented information on the practice of data collection on a day-to-day basis.

Conclusion
In this paper, we have presented the NTSs of six European countries, discussed their main challenges, and how these have been handled.One aim of the research was to identify solutions for future NTS survey design by studying other countries.Although the survey design varies between countries, the challenges are similar (non-response, underreporting, and representativity challenges).Most of these issues could be mediated with better ways of motivating respondents to participate, avoiding respondents misunderstanding questions and soft refusals.More attention needs to be put on the recruitment phase and questionnaire design.These parts of data collection have partly been neglected by transportation researchers since the mid-2000s, despite the negative trend in data quality.Data quality should get a higher priority because meaningful comparison of travel survey data is not possible in methodologically inconsistent travel surveys.
Transparency and proper documentation are essential to ensure the continuation of the time series and to share experiences across countries.Having a detailed methodology description and a discussion of quality indicators for each year of data collection should be mandatory.Formalised standards on reporting could reduce the transparency issue and ensure high-quality, open-access documentation of methodology.Documentation guidelines could e.g.be included in the tender requirements.
Since CAWI has increased in popularity, especially in multi-mode solutions, future research should evaluate how user-friendly the different versions are (interface, questionnaire, user-friendliness, etc.), and develop correction methods for mode effects.
2015).Due to a combination of poor data quality in 2000-2001 and monetary challenges, experiments with web-based data collection were carried out in 2005 (Christensen 2013).This resulted in the development of a new CAWI, which could either be used by CATI interviewers or by respondents without an interviewer.The web survey was successful, and a tender was launched to reduce the price of the CATI interviews (Christensen 2013).The Danish government decided to continue the NITS as a continuous survey from 2006.The third generation is considered to be from 2006today (Center for Transport Analytics 2022).

Table 3 .
Swedish NTS (2005-2019).Transportvaneundersøgelsen (TU) is the Danish National Travel Survey and is managed byDTU Transport (Christiansen and Skougaard 2015).DTU took over data collection and project management in 2001.TU collects data for multiple purposes (Christensen 2013): transport policies (national and local), national transport modelling, infrastructure investments, and research.The respondents are asked to report trips for one day.TU is financed by 63 partners and is used extensively on a municipal, regional, and local level.DTU divides the TU into generations (Center for Transport Analytics 2022).The first generation was
, the 1965 NHTS in Great Britain is the first documented NHTS.After that, data collection in Great Britain was carried out a few years apart (ad-hoc), and then annually with continuous data collection since 1988(Cornick  et al. 2020).The current NHTS only collects information on residents living in England (Department for Transport 2020b).

Table 7 .
German NTS (2002-2017).from official registers, although France used census data in the latest NTS.The countries are quite consistent in their definitions of a trip (see Appendix C), although weighting practices change over time and the definition varies between countries.The Scandinavian countries (Norway, Sweden, and Denmark) have chosen an NITS solution, and the methodology is similar.The Scandinavian countries and Germany have a multi-mode approach, arguing that this results in reaching a larger part of the population and increased flexibility.The multi-mode solution with CAWI is popular because it can potentially reduce costs while increasing responses among some groups.England and Germany have NHTSs and use personal interviews (CAPI).However, CAWI and CATI are being used more in Germany.France has chosen a combination of household interviews and individual trip reporting.England is the only country that extensively uses incentives as a part of their recruitment strategy.