04 September 2006
STATISTICS ON THE VOLUME OF ROAD TRAFFIC (VEHICLE-KILOMETRES)
C H A P TE R 3 b
Household surveys
UNEDITED DRAFT
for information and comments
COMMENTS, SUGGESTIONS, NATIONAL EXPERIENCES AND COMPARISONS, AS
WELL AS OTHER KINDS OF INPUT TO BE SENT TO:
Mr. Erik E. GRIB
Senior Statistician
Statistics Denmark
Sejrogade 11
DK-2100 COPENHAGEN
Tel: +45-3917 3025
Fax: +45-3917 3037
Email: eeg@dst.dk
1
C H A P TE R 3 b
Statistics Denmark
EEG/
04 September 2006
Household surveys
Household surveys are here meant to be surveys among households, persons or vehicle
owners with the solely purpose or one of the purposes to obtain information about the use of
the vehicles measured in vehicle-kilometres.
1. Sampling methods
The population to be sampled is the aggregate from which the sample is drawn. The
population might be:


The population of residents in the country, or as the scope is to collect data about the
vehicle use the domain of adult persons, as children are usually not vehicle owners
The population of vehicles or selected types of vehicles.
The population is subdivided into sampling units so that every element in the population
belongs to one and only one unit. The units may be persons, families, households, enterprises
etc. The list of sampling units is denoted the sampling frame. The construction of the
sampling frame is one of the major problems of the survey.
Possible sampling frames and their overlaps are shown in the graph.
Graph 1. Sampling frames
Resident natural persons
Resident legal persons
(Companies)
Resident vehicle owners
Resident Company
vehicle owners
Some countries may not have access to an up-to-date population register either because access
is restricted or because the population censuses are carried out with years in between. Such
countries may instead base the survey on lists of households, lists of addresses or lists of
telephone numbers.
Alternatively the territory may be split up into small districts, e.g. postal districts or streets,
and all persons or household in the sampled units surveyed (cluster sampling) or a sub-sample
of the persons or households in the sampled unit may be surveyed (two stage sampling).
2
The survey costs depend on the set up costs and the direct unit costs. In order to reduce survey
costs it might be considered to carry out multi purpose surveys. In such surveys only part of
the questionnaire will deal with the target data, here the vehicle use, and part of the sampled
units will be of no interest for the statistics about vehicle traffic, e.g. the persons or
households without vehicles. Those respondents can be filtered out easily via a filter question.
But the lower the frequency of vehicle ownership of residents or households the bigger will be
the sampling costs to obtain the desired degree of precision.
An existing general panel survey might be used for establishing of sub-sample. In this way the
set up costs can be reduced as the sampling frame is already established and the background
data of the panel will be available for the estimation. A filter question about car ownership is
added to the panel survey and subsequently a sample survey among the vehicle owners in the
panel is carried out to collect the information about the vehicle use, cf. Fact Box 1.
Fact Box 1
In France the SOFRES panel has since 1983 been used as basis for an annual
sample survey among household with car(s). The questionnaire is sent to 10,000
households at the end of each year, and results are available five months later.
Information requested is equipment on vehicles (light commercial vehicles
included since 1994), car purchases, number of kilometres driven by each vehicle
during the whole year, and unit consumptions.
In the French continuous SECODIP panel, 3,300 car owners send, every two
weeks, a questionnaire reporting odometer readings and volume of fuel
purchased, for each purchase made during the period.
In both surveys, detailed information on the type of car is available, and some
information on the driver.
These are examples of multi-purpose surveys for energy and environment
agencies and transport administration, and also for oil companies, car
manufacturers or insurance companies.
Source: TRANS/WP.6/AC.5/2005/7 Estimation of the annual balance of road
traffic and fuel consumption on the French national territory
2. Survey methods
The respondents can be contacted in several ways:



Postal survey
Telephone interview survey
Face-to-face interview survey
Postal survey
In a postal survey the data collecting costs will be about the same for all sampled units. But
the number of useful replies will among others depend on the frequency of vehicle ownership
in the population. Therefore if possible, it is preferable to sample directly the vehicle owners
3
using the national or the regional vehicle registers as the sampling frame or using other list of
vehicle owners e.g. from customer registers of petrol stations , car dealers or garages in order
to reduce the units of no interest to the survey.
Fact Box 2
In the Danish ad hoc survey on the use of light motor vehicles in 1992 3,000
passenger cars (0.2 pc. of the vehicle stock) and 2,500 vans and lorries of less
than 6 tonnes of gross weight (1 pc. of vehicle stock) were sampled from the
vehicles in Central Motor Vehicle Register. Ultimo 1992 questionnaires were
sent by post to the owners/users of the vehicles. Information about the mileage of
the vehicles in the past 12 month in Denmark and abroad was collected. The
response rate was about 75 pc.
Source: Statistiske Efterretninger, serien Samfærdsel og turisme 1993:24 (in
Danish only)
The postal survey method is very useful for simple and shorter questionnaires where the
respondents easily can fill-in the requested information without having to read too long
explanatory notes about how to do and how to interpret the questions. Best is if the questions
are self-explanatory. Reminders should be sent in order to increase the response rate and
prevent biases in statistics because of lower response rates for some segments of the sample.
Fact Box 3
In irregular intervals the German Federal Highway Research Institute BASt
realizes the so-called vehicle mileage survey (last 2002). The survey in 2002
covered approximately 127.000 vehicle owners from the master file at the
Federal Motor Transport Authority (Kraftfahrt-Bundesamt) using a method based
on random sample theory. The overall response rate was approximately 70%. In
this survey information was gathered on vehicle mileage on all German roads and
of vehicles registered in Germany. (FR question : is mileage abroad recorded?)
Source: Volume of road traffic based on data collection of the federal highway
Research Institute (BASt) - an informal document presented in the UNECE
workshop in Copenhagen 1-2 December 2005
If the questionnaire used is very detailed and requires long time to fill-in there is a risk that
the respondents will skip some of the questions, omit part of the questionnaire or totally
abstain from responding. It is therefore a good idea to care about the motivation of the
respondents. This may be in form of a premium for the participation or motivating telephone
calls, e.g. telephone call where the respondents are motivated and asked to complete a trip
diary for a shorter period.
An older Estonian survey based on filled in diaries obtained useful responses from 60 % of
the sampled vehicles although the owners were asked to report for a relatively long period.
4
Fact Box 4
An Estonian diary survey from 19967 examined the use of cars and MCs
belonging to residents of Tallin over a three month period. Odometer readings
were reported for each first day of the month and for each crossing of the city
border (date + km + direction). Further information on the daily use was gathered
(Not used; Used only inside Tallin; Used both inside and outside Tallin; Used
only outside Tallin)
Source: Ilmar Pihlak, Tallin University of Technology
Telephone surveys
In case of telephone surveys, the survey time can be kept short for the units that do not posses
any motor vehicle at all. Here the interview can be restricted to information used for the
estimation process, e.g. size of household, age of household members, etc. plus in multi
purpose surveys a filter question about car ownership.
Fact Box 5
The Canadian Vehicle Survey is partly a partly a postal survey and partly a
telephone survey. The survey has been carried out quarterly since 1999.
The survey provides annual and quarterly estimates of road activity for vehicles
registered in Canada. The estimates are provided by type of vehicle, by vehicle
characteristics (e.g. fuel type, body type, age), by driver characteristics (e.g. age,
gender) and by journey characteristics (e.g. trip purpose, season week day, time
of day, road type according to speed limit).
In the ten Provincial Governments registered owners of 20,000 sampled vehicles
were telephoned, interviewed concerning background data and asked to complete
a seven-day trip log. If respondents agreed a logbook were mailed to them. If
respondents could not be contacted by phone a trip log plus a short additional
questionnaire was mailed out. To increase the number of responses, respondents
were contacted a second time either by phone or by mail. On the first or second
day of the log attempts were made to phone each vehicle owner to answer any
questions the respondents might have. Later an attempt was made to contact by
phone or by mail everyone who did not return logs.
In the three Territorial Governments registered owners of 11,000 selected
vehicles were mailed postcards and asked to provide two odometer readings, one
at the beginning of the quarter and another at the beginning of the next quarter
plus information about he vehicle status (owned, sold scrapped).
Source: Canadian vehicle Survey 2003, Statistics Canada, Ottawa 2004
During the phone contact it is possible to explain towards the respondent the purpose of the
survey and to motivate him to participate. Further guiding can be supplied about how to
interpret the questions.
5
The length of a telephone interview should not be too long. Else the respondent may refuse to
complete the interview or they may seek to find answers that quickly will bring the interview
to an end. If the costs of establishing the sampling frame are relatively small it may be
preferable to limit the survey to one topic instead of having multi purpose telephone
interviews.
Face-to-face surveys
Face-to-face surveys may be used for the collection of the data. This type of survey is
normally used for longer and more complicated interview surveys where the benefit of the
direct personal contact can outweigh the relatively high survey costs of travelling. Here
cluster sampling may reduce the travelling costs of the interviewers.
Fact Box 6
TheBox
Swiss
Fact
2 Microcensus on Travel Behaviour is a survey conducted every five
years. This survey is primarily intended to measure distances travelled by the
permanent resident population aged 6 and over. Roughly 28,000 households
(about 30,000 individuals) were contacted by phone throughout the year 2000.
Phone interviews began with respondents being asked to provide sociodemographic data as well as information about their passenger cars and
motorcycles (age of vehicle, total distance travelled since vehicle was acquired,
total distance travelled in 2000 in Switzerland and abroad). One or two members
of the household (persons aged 6 and over) were then chosen for the rest of the
phone interview. These respondents were asked to describe very precisely where
they went on a specific day (“travel day”). Chosen at random for each household,
the travel day was generally the day before the phone interview. Respondents
were asked to indicate each trip segment (e.g. person going to work rides a bike
to the train station for the first trip segment, takes the train for the second trip
segment and then walks the remaining distance for the third trip segment),
departure and arrival points, departure and arrival times, means of transport used
(e.g. walked, rode bicycle, drove passenger car, rode in passenger car, etc.) and
the distance travelled for each trip segment. Based on random choices of
households, respondents and travel days and with the help of the Swiss Annual
Population Statistics, it is possible to estimate the volume of passenger car traffic
for the entire Swiss population for the year 2000.
Source: Source: TRANS/WP.6/AC.5/2005/13 Swiss calculation methods
3. Coverage of statistics
All motor vehicles or selected types of motor vehicles
The coverage of the statistics can be all motor vehicles or selected types of motor vehicles,
e.g. passenger cars or commercial goods vehicles. The Canadian Vehicle Survey, cf. Fact Box
5, covers all types of motor vehicles except motorcycles, while the Danish ad hoc surveys in
6
1992 and 1993 covered only passenger cars and light goods vehicles. The EU sample survey
of transport of goods by road only relates to good road vehicles above 6 tonnes of permissible
gross weight, cf. Fact Box 7.
Fact Box 7
EU countries (plus Norway and Liechtenstein) carry out quarterly sample surveys
of transport of goods by road as requested by Council Regulation 98/1172/EC.
In this survey information is gathered about the mileage of a sample of goods
road vehicles above 6 tonnes of gross weight that has been used for transport of
goods in the survey period. Some countries include vehicles above 3.5 tonnes of
gross weight. Journeys can be split up into national and international ones.
Source: Council Regulation (EC) No 1172/98 of 25 May 1998on statistical
returns in respect of the carriage of goods by road. Official Journal of the European
Communities, L163/1.
Passenger mobility surveys of journeys by mode of transport
Passenger daily mobility surveys are carried out in many European countries. These mobility
surveys focus on the movement of resident persons. Information is typically collected about
each journey carried out during a period, e.g. the day before, for a selected person in the
household. Journeys may be broken down into trips between activities (work, school,
shopping, leisure, etc.) and the trips split up into stages according to the mode of transport
used, cf. Fact Box 8.
Estimates of the movement of road vehicles may be deducted from the estimates of movement
of passengers. For car passengers it would however be necessary to have “car” as mode of
transport split up into “car as driver” and “car as co-passenger”. A few countries apply such a
breakdown of car passengers. Co-passengers may also exist for other modes of road transport
(MC, mopeds and bicycles) but here it may be assumed that the number of passengers
inclusive the driver on average is close to one.
Fact Box 8
An up-to-date inventory of national passenger mobility surveys report that among
EU-countries, applicant countries and EFTA countries at least one third of the
countries are running regular surveys; five countries have conducted a survey
recently but not indicated periodicity. Other three countries reported surveys that
were specific to travel by car, bus, coach etc. Most of the surveys were postal
surveys or combined postal and telephone surveys
Sources: D1.1_D2.1 Up-to-date inventory of national surveys on passenger and
car passenger mobility, AGILIS (August 2006) and D1.3_D2.2 Up-to-date
inventory of national surveys on passenger and car passenger mobility, AGILIS
(August 2006)
7
Often the passenger mobility surveys cover only non-business transport of passengers in
details. Business transport, e.g. agents transport may be reported as aggregates. The same may
apply to the movements of taxi drivers, lorry drivers and bus conductors where the
movements are not considered as passenger transport.
Coverage errors
Coverage errors may arise when the survey population does not fully correspond to the
population of interest. This may happen if:







The classification of the vehicles in scope divert from the one used in the survey. In
the EU road goods vehicle survey special goods vehicles (e.g. vehicles of heavy
weight or large dimensions); Lorries not used for transport (e.g. Lorries used for snow
removing); and goods vehicles used off road in closed areas (e.g. building sites,
forests), are not included. The vehicle kilometres will therefore not correspond to the
traffic performance of all vehicles classified as goods road vehicles.
The sampling frame is not up-to-date. If the list of vehicles used is based on the
registered vehicles months before the reference period part of the vehicles of interest
cannot be drawn in the sample. This will especially be the case for new registered
vehicles. As the vehicle-kilometres of younger vehicles tend to be higher than for
older vehicles this may lead to an underestimation.
The sampling frame may include vehicles that are scrapped or exported. This may
cause an overestimation.
The sample is drawn from a register of households or resident persons. In such case
the survey can only cover vehicles belonging to those units. Accordingly information
about vehicles owned by enterprises will usually not be reported - unless the vehicles
are permanently used by one and only one employee and thus can be regarded as
belonging to that person. Is the sample frame instead a vehicle register then
commercial vehicles can and will normally be included.
The sample is drawn from telephone directories. In this case part of the population of
interest cannot be sampled, namely the units without telephone and the persons not
registered in the directory either because the telephone directory it is not fully up-todate or because the telephone number is confidential. This may lead to biases of the
estimates. The increased use of mobile telephone instead of fixed net telephones may
have impact on the access to up-to-date telephone directories as mobile telephone
holders not always are registered in the directories.
The scope is to gain knowledge about the vehicle performance on the domestic
territory and the survey frame is only covering vehicles registered in the country.
The traffic of the surveyed vehicles includes traffic abroad.
To some extent the estimation procedure may be able to correct for biases and over and under
coverage.
Vehicles registered abroad
If the target population is the vehicles performing traffic in the country and the information
requested is the vehicle-kilometres performed on the national territory then both surveys
based on resident households and on vehicles registered in the national vehicle register will
8
not fully meet the target, partly because the vehicles surveyed will perform traffic both
domestic and abroad, partly because the traffic by foreign vehicles will be excluded.
The lack of information about the traffic performed by foreign vehicles often has lead to the
assumption that the volume corresponds to the volume of traffic of national vehicles abroad.
This assumption does not always reflect the realities. The cross bordering traffic may be
influenced by differences in level of income, prices, production and tourism supply between
the countries. There is therefore a need for solid information about traffic by foreign vehicles
on the national territory.
Traffic abroad with national vehicles
In passenger mobility surveys trip data are collected. If origin and destination of the trip is
reported it will be possible to obtain data for cross bordering trips. The distance travelled
abroad on a trip may be calculated on basis of the border crossing place obtain from the
survey or estimated on basis of route planners for the shortest distance between the two
places. For iteration journeys abroad calculation of the volume of traffic is more problematic.
Either one will have to request information for each single trip on the journey, which will
present an enormous burden on the respondent, or one may ask the respondent to report the
total journey distance or the distance driven abroad, which may then be more or less accurate.
Fact Box 9
In the DATELINE survey, a European long distance mobility sample survey,
carried out in the EU 15 Member States in 2001/2002), data were collected on a
household basis for each journey of more than 100 km travelled during the past
calendar month. Supplementary data about the household and the vehicle stock
was collected too. The journeys were described on a trip level with data about
O/D, time of departure and arrival, means of transport (type of road vehicle, own
or rented), major stops underway etc. The data collected would allow for
calculation of distances in the countries visited.
Source: DATELINE Deliverable 5,
http://www.elmis.org/docs/DATELINE_del5.pdf
The long distance mobility surveys like the above mentioned DATELINE survey will not
include journeys shorter than the threshold, e.g. 100 km. Daily mobility surveys where the
focus is on the trips performed on a selected day, typically the day before the interview, will
not include the longer holiday journeys.
But the mobility surveys may include background questions about the overall use of the
vehicles belonging to the household in the past year. This might be information about the
technical data of the vehicles (age, type, size, fuel type), about the mileage run domestically
and abroad, and about the fuel consumption.
9
5. Other types of errors
The following types of errors may also influence the quality of the statistics:




Sampling errors
Recall errors
Measurement errors
Errors due to underreporting
Sampling error
Sampling error is caused by the fact that the statistics are based on a sample of units instead of
all units in the population. The size of the sampling error depends on the sample size,
stratification and estimation procedures.
In general the survey target is to reach information about the vehicle mileage in a calendar
year. The surveyed units may be asked to report:



The mileage run in a calendar year or a 12 month period close to the calendar year
The mileage run in a shorter period
The odometer reading at the start and at the end of a period chosen, e.g. a fortnight
The smaller the reference period the smaller is the real sample size. If 10 out of 100 vehicles
are sampled and the reference period is 1 week the sampling rate is 1 vehicle-week out of
5.200 vehicle-weeks, i.e. 0.2 %, but if the reference period instead was the full year the
sampling rate would be 10 %. But what is gained by extending the reference period may be
lost because of the recall effect.
Recall errors
Only few persons keep records on the daily use of their vehicle. Trip data for a shorter period
may be obtained, cf. the Canadian survey description in Fact Box 4. But the more trips carried
out the bigger the risk of underreporting.
Alternatively the desired information may be reported as an estimate of the volume of traffic
for a longer period, e.g. a calendar year. The respondent may calculate the vehicle-kilometres
on basis of:



The normal use of the vehicle in a week
The average annual use of the vehicle based on the age of the vehicle and the
odometer reading
The visits to a garage for regular services, oil replacement, motor check etc. for each
15.000, 20.000 or 30.000 km driven
The calculated estimate may in all cases divert from the actual use of the vehicle and may
most likely be affected by the respondent’s assumptions of the normal use of the vehicle.
10
New vehicles, sold or scrapped vehicles
A special variant of “recall” error is found in the case where the vehicle has only been in the
possession of the owner for part of the reference period. Here the owner normally cannot
supply information covering the full period. The situation is especially of relevance when the
reference period is long. If a sampled vehicle is bought brand new or if it is bought second
hand during the reference period then the car owner can instead be asked to report either the
expected annual use of the car or the use in his part of the reference period.
Fact Box 10
In the Danish ad hoc surveys in 1992 and 1993 sampled car owners of vehicles
acquired in the reference period were asked to report the mileage driven only in
the period of ownership. The annual mileage for the vehicle was then estimated
on basis of the information hold in the Central Vehicle Register about the date of
acquisition of the vehicle.
Source: See Fact Box 2
Measurement errors
To avoid errors stemming from estimates of the mileage of the vehicle in a given period one
could instead asked the vehicle owner to report the actual odometer reading the morning of
the first day of the reporting period and again at the evening of the last day of the reporting
period either in one questionnaire or in two consecutive questionnaires.
Fact Box 11
Fact Box
In the German BASt survey owners of motor vehicles were asked in six waves
over the year to report the readings of their odometers in the beginning an in the
end of a period.
Source: See Fact Box 3
But here errors may also occur in form of measurement errors, e.g. because of misreading of
the odometer. The respondent may wrongly:




Round the vehicle-kilometres to the nearest thousands, 5 thousands or ten thousands
Include decimal figure if the odometer presents kilometres with decimal
Report the wrong vehicle if more vehicles belong to the household or the enterprise
surveyed
Change procedure between first and second reading
Measurement errors may also occur to other variables.
11
Underreporting
Non-response and underreporting may happen if the burden on the respondent is felt too high.
It may be in form of a lengthy interview or a questionnaire requesting reporting of many data.
To avoid biases stemming from non-response measures can be taken to reduce the level of the
non-response, e.g. accompanying the questionnaires with information about the use of the
statistics, and use of reminders – postal or by telephone.
Underreporting takes place if a respondent only states part of the vehicle journeys. The
underreporting is not easily detected for the single respondent but will be revealed in the final
statistics if comparable data are available from other sources.
Fact Box 12
In the Danish road goods transport survey carried out as part of the EU survey of
transport of goods by road, cf. Fact Box 7, a significant gap between the
estimated vehicle-kilometres from the sample survey data and the calculated
vehicle-kilometres obtained from the odometer readings at the mandatory
periodical technical vehicles has been observed. The reason is presumably the
comprehensive request for detailed journey data in the transport survey. It is
supposed that some enterprises omit part of the journeys made or falsely report
that the sampled vehicle has not been used for national transport in the reporting
period.
Similar observations have been made in other countries.
Source: Declarations of content: Transport of goods by road by Danish vehicles
in national traffic.
http://www.dst.dk/HomeUK/Guide/documentation/Varedeklarationer/emnegrupp
e/emne.aspx?sysrid=992
6. Data collected
The data collected in households surveys fall into three types:

Vehicle data
o Type of vehicle
 Passenger car
 Bus or coach
 Light goods vehicle, e.g. van or small lorry
 Lorry > 6 tonnes gross weight
 Road tractor
 Articulated vehicle
 Road train
 Moped
12


 Bicycle
o Weight of vehicle
 Kerb weight
 Gross weight
 Load capacity
o Axles
o Body of vehicle
o Age of vehicle (or year of first registration)
o Type of fuel
 Diesel
 Gasoline
o Odometer reading
o Vehicle-km in a period
Driver characteristics
o Gender
o Age of driver
o Residency of driver
o Socio-economic data
Journey characteristics
o Purpose of journey
 Commuting
 Shopping
 Leisure
 Holiday
 Commercial
o Time of journey
 Season
 Week day
 Hour
o Road type (e.g. according to speed limit)
o Domestic/international
o Length of journey, km
13