Transport Analysis (NSI)

Statisticon (Statistics and Research) (Producer)

Transport Analysis 

Transport Analysis

Rosenlundsgatan 54

118 63 STOCKHOLM

SWEDEN

Road freight data collection consists of three datasets with quarterly periodicity:
1. Vehicle related variables dataset for a sample of road good motor vehicles contain, among others, indication about the type and age of the vehicle, the economic activity of the owner of the vehicle, the total kilometres loaded or empty.
2. Journey related variables dataset which contain, among other indications about weight of goods, regions of loading and unloading, distance driven, and tonne-kilometres performed during a journey
3. Goods related variables dataset which contain among others the type of goods transported and their weight and distance travelled, regions of loading and unloading.

Statistics on carriage of goods by road apply the following statistical classifications:

• The type of goods transported by road was collected, until reference year 2007 (included), according to the 24 groups of goods following the "Standard Goods Classification for Transport Statistics/Revised (NST/R )", . Since 2008 the coding system NST 2007 has been used for classification of goods, see Standard goods classification for transport statistics, 2007.
• The regional coding is done in conformity with European legislation: "NUTS Nomenclature of territorial units for statistics" (see also NUTS).
• Economic activity class of the vehicle owner is an optional variable. Data were collected according to NACE Rev. 1 until the end of 2007. From the beginning of 2008 onwards, the economic activity has been classified according to NACE Rev. 2 (see  "Statistical Classification of Economic Activities in the European Community, Rev. 2 (NACE Rev. 2)").
• Type of dangerous goods is a mandatory variable, but it is reported only for those goods that fall into this category. The classification is based on "European Agreement concerning the International Carriage of Dangerous Goods by Road", Chapter 2.1, as published in Annex E of Council Regulation (EC) No 1172/98. Directive 2008/68/EC makes reference to ADR as regards the transport of dangerous goods by road.

The optional variable type of cargo follows the Classification of Cargo Types of UNECE (United Nations, Economic Commission for Europe - codes for types of cargo, packages and packaging materials, Recommendation 21 adopted by the Working Party on Facilitation of International Trade Procedures, Geneva, March 1986).

The main concepts used in Road freight statistics are the following, more details can be found in the Road freight transport methodology manual : 

A goods road motor vehicle is any single road transport vehicle (lorry), or combination of road vehicles, namely road train (lorry with trailer) or articulated vehicle (road tractor with semi-trailer), designed to carry goods

National transport is Road transport between two places (a place of loading and a place of unloading) located in the same country by a vehicle registered in that country. 

International transport is Road transport between two places (a place of loading and a place of unloading) in two different countries and cabotage by road. It may involve transit through one or more additional country or countries.

Cross trade is International road transport between two different countries performed by a road motor vehicle registered in a third country.

Transit is any loaded or empty road motor vehicle, which enters and leaves a country at different points by whatever means of transport, provided the total journey within the country is by road and that there is no loading or unloading in the country.

Goods carried by road are any goods moved by goods vehicle

Place of loading/unloading is the geographical area for loading/unloading coded according to NUTS3-level

Place of loading/unloading of a goods road vehicle on another mode of transport

• Place of loading (of the goods road transport vehicle on another mode of transport): The place of loading is the first place where the goods road motor vehicle was loaded on to another mode of transport (usually a ship or a rail wagon).
• Place of unloading (of the goods road transport vehicle from another mode of transport): The place of unloading is the last place where the goods road motor vehicle was unloaded from another mode of transport (usually a ship or a rail wagon).

The reporting unit for road freight transport statistics is the goods road transport vehicle.

A selected goods road vehicle report all basic transport operations (consignments) during the reference week. This information is transformed to journeys which constitutes the statistical unit used for estimating population parameters such as total tonnes, tonne-kilometres and vehicle-kilometres. 

The target population is all goods transports on public roads carried out by Swedish trucks / tractors with a load capacity of 3.5 tonnes or more during the reference period. Other exclusion criteras regarding vehicles are:

• Vehicles 30 years and older and vehicles that are not operating.
• Some body type codes for which transport of goods is not possible, such as ambulances, hearses, breakdown lorries, fire engines (military vehicles are not included in the Vehicle Register).
• Vehicles owned by companies that are not registered in the central register of corporations.

  The population size of vehicles is around 60 000 to 70 000 vehicles. Since the population is large a (random) sample of vehicles is drawn each quarter.

The data provided are goods vehicles registered in Sweden.

Quarterly data from year 2000. 

Not applicable.

Data are collected regarding each basic transport operation (BTO) during a selected vehicle. The metric variables are loaded goods in kg and driven distance in km. The variable tonne-km is a derived variable, hence the vehicle owner does not need to register ton-km. Regarding number of journeys this is obtain when BTO:s are converted to journeys.

The statistics presented in tables are in tonnes or thousands of tonnes for loaded goods, km or thousands of km for driving distance, in millions of tonne-km and in thousands of journeys. 

The reference period is quarter (13 weeks), i.e. the data delivered to Eurostat can be used to estimate parameters for a quarter. If four quarters are added together the statistics can be presented regarding a year (52 weeks).

National level:

• It is mandatory to reply to the questionnaire according to law 2001:99 and regulation 2001:100 regarding Official Statistics and also according to Transport Analysis (NSI) regulation (TRAFAFS 2014:2) 

European level: 

• REGULATION (EU) N° 70/2012   

National level :

Transport Analysis (NSI) publish estimates of totals of tonnes, vehicle-kilometres and tonne-kilometres and also divided into several subgroups (domains) on a quarterly basis. Microdata or specialized tables to researchers and other users is provided by Transport analysis after approved secrecy examination and / or disclosure control.

From Eurostat :

Eurostat submits annually semi-aggregated data (data exchange tables, see Commission Regulation (EU) No 202/2010 amending the Commission Regulation (EC) 6/2003) back to the reporting countries so that they can compile the total road freight transport on their national territories, including the operations by national hauliers and also those of all other reporting countries.

These data exchange tables include more detailed breakdowns than the publicly available tables. They also include, for each value, the information on the number of observations that the estimates are based on. In this way, the reporting countries can estimate the reliability of results that they aggregate from the data exchange tables.

National level:

• For production and publication of statistics in Sweden, public authorities are bound to secrecy according to chapter 24, 8 § in the Public and Privacy Act (2009:400)  To protect individuals' or companies' confidential information it is ensured that they cannot be disclosed directly or indirectly in the statistics that are made public.

European level:

• Regulation (EC) No 223/2009 on European statistics (recital 24 and Article 20(4)) of 11 March 2009 (OJ L 87, p. 164), stipulates the need to establish common principles and guidelines ensuring the confidentiality of data used for the production of European statistics and the access to those confidential data with due account for technical developments and the requirements of users in a democratic society.

Results are pulished on Transport Analysis (NSI) web site on the following link

https://www.trafa.se/en/road-traffic/swedish-road-goods-transport/

The statistics are published on Transport Analysis (NSI) web site www.trafa.se. An email to users and subscribers will be sent on the day the statistics are published. The date for quarterly and annual publication is stated in the autumn the year before the reference year.

Quarterly and yearly. Se section 8.1 for more details.

Annual and qurterly publication: Press release regarding annual and quarterly publication is available on Transport Analysis (NSI) web page (only in swedish) https://www.trafa.se/om-oss/press#/

Quarterly and annual publication: Technical documentation (quality declaration- in swedish) in PDF. Annual publication: Summary report (PDF). All publications are available on Transport Analysis (NSI) web page: https://www.trafa.se/en/road-traffic/swedish-road-goods-transport/

Excel tables with published results can be found on https://www.trafa.se/en/road-traffic/swedish-road-goods-transport/

A database developed for customized extraction of results (estimates, not micro data) is available at: https://www.trafa.se/vagtrafik/lastbilstrafik/?cw=1

 Please note that this particular web page is not translated into English.

Various statistical methodological reports.

National characteristics of surveys, conducted in the reporting countries in 2017, were published in Methodologies used in surveys of road freight transport in Member States, EFTA and Candidate Countries. This latter publication also contains data on response rates, vehicle registers' quality, sampling rates and statistical errors in surveys carried out in 2016.

A technical documentation (quality declaration) with more details regarding Sweden can be found on https://www.trafa.se/en/road-traffic/swedish-road-goods-transport/

Please note that this documentation is only in Swedish.

Both manual controls, indata controls and contacts with reporting companies are made.

The quality assurance of the Swedish road freight transport statistics data is based on the following principles:

The statistical error (percentage standard error) of estimates based on the micro-data transmitted to Eurostat shall not be greater than 5% (special conditions apply to smaller countries, see Commission Regulation (EC) 642/2004).

In addition to the data, we inform Eurostat on the national characteristics of the data collection. This information is published as part of the quality documentation.

Eurostat validates the incoming micro-data, record by record, by applying detailed validation checks as described in Chapter 12 of Road freight transport methodology.

Benchmarking studies are carried out to verify the existence of possible bias in the road freight data by comparing the results with other independent data sources when needed

Tools for the assessment of the quality of the aggregated road freight data are for example by comparisons with other registerbased sources.

The aim is to keep the users of the data informed on all these aspects by means of the methodological manuals.

The main users of the road freight statistics are the European Commission, national authorities, research institutions, trade organizations, industrial organization and contractors.

The statistics are used in the Government's work on freight and other transport issues analyses on competition, regional policies, modal shift analyses and environmental analyses.

No formal survey towards users have been carried out. Instead, Transport Analysis have a Council of users. In the Council the most important users of the statistics are represented. The Council meets about once a year. In the meetings the users can express their view regarding the statistics in different matters. 

The published statistics are in line with the needs of most users. Two areas where the user needs are not fulfilled are:

• dangerous goods
• geographical breakdown

Regarding dangerous goods estimates are on a national level divided into national and international transport. Estimates are provided by ADR-classes. Some users would like to see estimates divided into finer subgroups which for the moment cannot be provided, because the sample size is not large enough.

Regarding geographical breakdown the user need is to provide estimates on municipality level, not NUTS3 level. This has until now not been possible, since municipality is not collected.

However, local geographical areas (LAU) has been collected since long ago. But without ability to connect the LAUs to the municipality. During the autumn 2019, due to a development project, we can now match the LAUs to the municipality. This has been done for data from 2012 and forward. This means that the geographical breakdown on municipalitys are possible from now on.

 Still, many users ask for detailed data. If the published data does not include the statistics the users need, specialized tables can be provided by Transport Analysis after approved secrecy examination and / or disclosure control.

The statistics are subject to uncertainty. The main sources of uncertainty are sampling, non-response and measurement. The uncertainty due to sampling is quantified using confidence intervals (95 percent). The non-response is compensated in the estimation procedures through straight expansion within strata. No special method or model is used for potential measurement uncertainty.

In 2014 an evaluation study was conducted regarding non-response. The results of the analysis show that for some variables (load capacity, geographical regions and age of the vehicle) no bias occurs, but for the variable register-based vehicle kilometers there is a bias. The non-response analyzes show that non-responding vehicles have a longer average daily (register-based) vehicle kilometers than the responders. Since there is a correlation between the register-based vehicle kilometers and the actual vehicle kilometers during the measurement week, this implies a risk of bias regarding mileage. This can lead to underestimation of the true levels for the total number of kilometers traveled.

One other aspect that affects the estimates is false reporting of no activity. False reporting means that for a certain vehicle the reply to the questionnaire is that no activity was performed during the measurement week, when in fact activity was performed, i.e. a false reply. In 2014 a new method to account for false reporting of no activity was introduced. This lead to a raise in the estimates with about 30 percent. These higher levels better reflect the actual level of road freight transport and align better with corresponding data from other sources.

The sampling frame with vehicles is stratified into 52 strata. A simple random sample is drawn within each stratum. The total sample size per quarter is around 3000 vehicles which leads to roughly 12 000 sampled vehicles per year.

Uncertainty in the estimates derived from sampling is quantified via confidence intervals. The confidence intervals are dependent on the scale of the variable so to facilitate the comparison the relative uncertainty margin (also called the relative margin of error or sometimes the percentage standard error) is calculated for certain parameters. The precision requirement from Eurostat is that the relative margin of error shall not exceed ±5 % for the six parameters in the table below.

Relative margin of error:

In Sweden the sample size is dimensioned so the precision requirements should be fulfilled. With approximately 3000 vehicles per quarter (12 000 per year) the precision requirements are fulfilled for all parameters in 2021 (rounded to whole numbers). 

 A common classification of sources of non samling errors is frame coverage, measurement, non-response, data processing and model assumptions. Each source is described briefly below

Frame coverage

Frame coverage comprises both over- and under-coverage. Overcoverage, i.e. that a vehicle is included in the sampling frame, even though it does not belong to the target population, may occur, for example, because the vehicle does not carry out road freight transports on public roads but only within a defined area (e.g. industrial area). Another cause of over-coverage depends on the time aspect. The register that forms the base for the sampling frame is retrieved from the Swedish Transport Agency about 1.5 months before the quarterly start. The last weeks in a specific quarter thus occur just over four months after the register was taken. At that time, the vehicle can e.g. have been scrapped, canceled or unregistered.

There were 361 vehicles classified as over-coverage during 2022 in the Swedish survey. This corresponds to 3,1 percent of all 11 741 sampled vehicles. The calculation 100-3.1 = 96.9 percent gives a measure of the register quality in terms of non-over-coverage.

Under-coverage is generally a more serious problem than over-coverage. The under-coverage is methodologically handled in the estimation procedure. In principle, the method assumes that the number of vehicles in the frame (N_h) is replaced by an estimated value for the number of vehicles at the midpoint of the quarter. The sampling frame for e.g. quarter 2 in a certain year took place on February 15 that year, which is the midpoint for quarter 1 the same year. This means that the number of vehicles from the sampling frame for quarter 2 serves as (best) estimate of the number of vehicles for quarter 1. A similar procedure is used for all quarter.

Measurement

The method with questionnaires depends on accurate and truthful answers from the respondents. The respondents also depend on the instructions and definitions being clear and understandable. The information requested is sometimes complicated and even with definitions that appear clear, misunderstandings can arise. It is difficult to know whether all the data collected is accurate and complete. Below are some examples of situations where measurement uncertainty occurs.

Uncertain vehicle kilometer data and missing empty runs may occur. In case of missing vehicle kilometers data, these are imputed by the statistical producer with the help of route planning programs.

For some types of consignments, the respondents have difficulties knowing the weight of goods. This can occur if the transport assignment is to pick up a semi-trailer at one loading terminal and transport it to another loading terminal. The importance of the goods may in some of these cases be unknown to the responder. In these cases, the weight must be estimated. In these cases, the type of goods may also be unknown.

A potential measurement error is the under-reporting of individual consignments during the measurement week. This can happen if a vehicle owner responds to the survey but forgets to include certain consignments, such as empty runs. In some situations, this can be identified at registration. No in-depth studies of the extent of under-reporting of individual consignments have been conducted.

Another important measurement uncertainty is that some responders report no activity during the measurement week when in fact activity was done, i.e. false reporting. This problem is serious but from 2014, a new estimation method, taking false reporting of no activity into account was introduced.

Non-response

Unit non-response affects the statistics in two ways. On one hand the base (number of responders) is reduced, which adds to the uncertainty. However, this is not a major problem, at least for high aggregation level estimates. The other effect, which is more serious, is the risk that results are biased due to non-response. If those who do not respond differ systematically (regarding the survey variables) from those who have answered, there is a risk of bias in the estimates. The method to compensate for non-response, straight expansion within strata, is based on the fact that the non-response occurres randomly within strata.

The response rate in the Swedish road freight survey was 57,3 percent 2022. Non-response is increasing in surveys and Transport analysis is investigating possibilitites to extracting traffic data from vehicle Fleet management systems in order to aviod non-response and to reduce the response burden. 

In 2014 an evaluation study was conducted regarding non-response. The results of the analysis show that for some variables (load capacity, geographical regions and age of the vehicle) no bias occurs, but for the variable register-based vehicle kilometers there is a bias. The non-response analyzes show that non-responding vehicles have a longer average daily (register-based) vehicle kilometers than the responders. Since there is a correlation between the register-based vehicle kilometers and the actual vehicle kilometers during the measurement week, this implies a risk of bias regarding mileage. This can lead to underestimation of the true levels for the total number of kilometers traveled.

Item non-response is very limited in the survey. Regarding the mandatory variables, no item non-response exists in the databases delivered to Eurostat. Some (limited) imputations are however needed to obtain that. Regarding the voluntary variable degree of loading (volumetric situation), which is delivered to Eurostat, there exists some 30-40 percent item non-response.

Data editing and coding

Data editing and coding is also a source of uncertainty. This source of uncertainty is regarded as marginal. In support of this claim it can be mentioned that when data is submitted to Eurostat extensive validation checks are carried out. If incorrect values (errors) are detected they need to be corrected. However, errors in data submitted to Eurostat have been very uncommon.

Model assumptions

The most important model assumption in the survey concerns how compensation for the non-response and the method used to account for false reporting of no activity. The estimation method is called straight expansion within strata and means that the respondents within a stratum are considered as if they were the sample. Alternatively put, this implies an assumption that the non-response occurs randomly within strata, i.e. there is no systematic tendency who choose to answer or not to answer. Reasons for not responding may be refusal, forgetfulness, do not have time, the form, etc. An evaluation study regarding the non-response was conducted in 2014. Above, regarding non-response, this was described and what conclusions were drawn from the analysis.

The method to account for false reporting of no activity is based on a separate survey: The No Activity Survey. The frame for the No Activity Survey is the same as for the Road Freight Survey. The sample size is 500 vehicles every quarter The sample in the No Activity Survey is negatively coordinated with the ordinary survey, i.e. a vehicle sampled in the ordinary Road Freight Survey can not be selected in the No Activity Survey. Only one question is posed in this survey: Did you use the vehicle for goods transportation last week? Based on the answer from this survey and the answers from the ordinary road freight survey a correction factor (inflation factor) is calculated. This inflation factor is used in the estimation process. Normally the estimates are inflated some 30-40 percent compared to not using the inflation factor.

This assumption is difficult to verify. Rather, when comparing the level for the estimates from the road freight survey to data sources it has been found that the underestimation at national level is in the order of 30–40 percent. By using the inflation factor to account for false reporting of no activity, the estimates are raised 30-40 percent which should reflect the true levels better.

Quarterly data and publication: The publication of quarterly report, tables with estimates and documentation is normally done at Transport Analysis web site almost three months after reference quarter. The A1, A2 and A3 datasets are normally delivered to Eurostat 2,5 months after reference quarter.

Annual publication: The publication of annual report, tables with estimates and documentation is normally done at Transport Analysis web site almost five months after reference year. No additional annual data is delivered to Eurostat (only quarterly data).

The dates for quarterly and annual publication are stated on Transport analysis web page well in advance. The date for actual publication has, so far, always been the planned dates.

Regarding delivery of data to Eurostat, the Regulation (EU) 70/2012 stipulate that the A1, A2 and A3 datasets should be delivered within five months after reference quarter. Sweden normally delivers the datasets 2,5 months after reference quarter.

Sweden conducts the road freight survey according to Regulation (EU) No. 70/2012. That means that the definitions are the same with those presented in the manual. As stated on Eurostat Road freight transport measurement (road_go) Reference Metadata in Euro SDMX Metadata Structure (https://ec.europa.eu/eurostat/cache/metadata/en/road_go_esms.htm) there are some national characteristics of the survey that should be kept in mind when comparing the estimates. Here we give comments regarding Sweden in those matters:

1. Weight. Gross weight of goods is collected (i.e. not gross-gross weight), which means that containers, swap bodies and pallets are excluded. However, for journeys with no goods but a container, the information to the respondents is to register the weight of the container and use NST-code 16.
2. Multi-stop journeys. Multi-stop journeys are coded by consignments. The respondent records each basic transport operation in the questionnaire. These records are then recalculated to journey level.
3. Classification of containers. Swap-bodies are reported as "large freight containers".

Sweden has provided data since 2000. The years 2000-2013 are fully comparables. In 2014 a break in the time series occurred. False reporting of no activity hade for some time been a problem in Sweden. False reporting means that for a certain vehicle the reply to the questionnaire is that no activity was performed during the measurement week, when in fact activity was performed, i.e. a false reply. In 2014, a new method to account for false reporting of no activity was introduced. This lead to a raise in the estimates with about 30 percent. The years 2014 to present are fully comparable.

From 2015, total loaded freight weight is requested for distribution and collection rounds in the form. In previous years, the average loaded person was demanded. This has meant that the total evel of loaded goods during a quarter has increased with an estimated 6–8 percent.

As pointed out on Eurostat Road freight transport measurement (road_go) Reference Metadata in Euro SDMX Metadata Structure (https://ec.europa.eu/eurostat/cache/metadata/en/road_go_esms.htm) there are some principal differences between road freight transport and other transport modes. In addition to what is stated in that text, one might add that comparability cross domains is facilitated if the same nomenclature is used for different variables. For example, those transport modes that classifies geographic information according to NUTS3-levels (which is done in road freight transport) are coherent, and thus comparable, in that matter.

The road freight transport survey uses the classification system stated in section 3.2 above. Other transport modes that also use this classification systems are coherent and can be compared in those matters.

Sweden has no other comments to internal coherence than what is stated on Eurostat Road freight transport measurement (road_go) Reference Metadata in Euro SDMX Metadata Structure (https://ec.europa.eu/eurostat/cache/metadata/en/road_go_esms.htm)

In Sweden the burden for the responders to reply to the questionnaire is estimated to approximately 4 600 hours in 2022. Please note that this is a very rough estimate. Assuming a certain cost per hour the corresponding total cost can be estimated. The estimate for 2022 is roughly 4.4 millions SEK which (approximately) corresponds to 414 000 EUR.

The median time to fulfill the questionnaire 2022 is 30 minutes for all respondents (among those who answered the question about time). For those with journey data the median time to fulfill the questionnaire is 45 minutes and for those without journey data the median time is 5 minutes.

Important major efforts that are being made and have been implemented are

• Annual revision of the form

• Reduced sample size in 2013

• Introduction of web survey in 2015

• A new web interface was introduced in 2022. In the web interface, there is an Excel template that has the same appearance and columns as the dispatch journal that informants must fill in. Thus, an informant can fill in dispatches in the Excel template and then paste this into the web survey.

• When sending out the survey, there is now a possibility that a respondent can receive an invitation to the survey via e-mail, instead of a postal survey. This was introduced in 2022.

Planned future actions are
• Retrieve data (answers) via the companies' IT systems instead of answering via questionnaires
• Make general improvements to the web survey to make it easier to answer

Sweden obey the policy stated on Eurostat Road freight transport measurement (road_go) Reference Metadata in Euro SDMX Metadata Structure (https://ec.europa.eu/eurostat/cache/metadata/en/road_go_esms.htm).

It can also be stated that the policy in Sweden is that the datasets (A1, A2 and A3) should be 100% error-free (regarding the validation procedure at Eurostat).

If errors occurs in the validation process (regarding datasets A1, A2 or A3), the practice in Sweden is to immediately correct the errors and re-deliver all datasets. However, errors are very unusual.

Construction of sampling frame for the survey

Name of register:

Vehicle Register (VR), Commercial Traffic Register (CTR), Central register of corporations (FDB) and the vehicle-kilometre database (VKD). These registers are used to construct the sampling frame.

Name of organisation who maintains the register:

Swedish Transport Agency (the registers VR and CTR) and Statistics Sweden (the registers FDB and VKD).

Frequency of update:

• VR, CTR and FDB: daily
• Vehicle-kilometre data (VKD): yearly

Frequency of construction of frame:

Once a quarter

Statistical unit in sampling frame:

Tractive vehicle

Arrangements for accessing the register:

Transport Analysis is the responsible authority for the survey and they have since year 2009 commissioned the company Statisticon AB to produce the survey. Statisticon AB currently has the option for this production until reference year 2023. The VR and CTR is delivered from the Swedish Transport Agency to the producer at specified dates. For quarter Q the register data is received 1.5 months in advance. The specific dates each year are:

• First quarter: November 15 (previous year)
• Second quarter: February 15
• Third quarter: May 15
• Fourth quarter: August 15

The sampling frame is constructed based on the Vehicle Register data where the object is vehicle. In the process various steps are taken, including omitting vehicles not belonging to the target population (e.g. load capacity should be 3.5 tonnnes or more). One other step includes merging data from the central register of corporations (FDB) and only keeping those vehicles belonging to formally registered companies. Yet another step includes merging register based data on driving distance (kilometres travelled) from previous year for each vehicle based on information from the vehicle- kilometre database (VKD). 

Information obtained from the register:

Most of the information regarding a vehicle is obtained from the Vehicle Register. Important variables are: vehicle registration number, organisation number of the enterprise/owner of the vehicle, name and address, year of first registration, vehicle in use/not in use, type of transport (hire/reward or own account), number of axles for the vehicle, vehicle body code plus the information is used in the stratification as follows:

• County codes where the vehicle is registered are used to get the NUTS 2-level codes
• Maximum permissible laden weight and the service weight of the vehicle. The difference between those two concepts gives the vehicle load capacity, which is used in the stratification
• From the Commercial Traffic Register we obtain the number of permits for international traffic for an enterprise. This variable is used to form a stratification variable whether the vehicle has a permit for international traffic or not
• The variable kilometres travelled previous year is also used in the stratification
• Age for the vehicle (used in the stratification)

Types of vehicles excluded:

• Vehicles with load capacity under 3.5 tonnes are excluded.
• Vehicles that are not operating
• Vehicles 30 years and older.
• Some body type codes for which transport of goods is not possible, such as ambulances, hearses, breakdown lorries, fire engines (military vehicles are not included in the Vehicle Register).
• Vehicles owned by companies that are not registered in the central register of corporations.

Continuous, 52 weeks of the year (53 weeks when applicable).

Sampling methodology

Statistical unit:

Tractive vehicle 

Time unit for measurement week:

1 week

Time unit of quarter included in the survey:

All (13 weeks)

Stratification:

The lorries in the sampling frame are stratified in 52 strata. The stratification is done with respect to several variables. The first variable is whether the lorry owner has a permit for international traffic or not. If the owner (i.e. the company) has a permit for international traffic the lorry or the lorries are categorized to international stratum, otherwise to national stratum. There are 35 national and 17 international strata. The next stratification variable is where the lorry is registered. The eight NUTS2-regions in Sweden categorized into 5 geographic areas according to the following:

1. SE11 and SE12
2. SE21 (without the island Gotland) and SE23
3. SE22
4. SE31, SE32 och SE33
5. Gotland

Due to special circumstances regarding the island Gotland it is kept as a region of its own.

For the 35 national strata the stratification is done according to the following principles:

• The lorries are geographically divided according to the geographic regions above
• In every geographic region, except Gotland, the lorries are divided into two groups according to their age; lorries 0–5 years old and lorries older than 5 years old
• Within each age group the lorries are divided into six subgroups according to total kilometres travelled per year (register variable) and load capacity (register variable) according to:
  • Lorries with load capacity up to 13 tonnes (regardless of yearly distance travelled)
  • Lorries with yearly distance travelled 100 000 km or less and load capacity 13 - 16 tonnes
  • Lorries with yearly distance travelled 100 000 km or less and load capacity over 13 tonnes
  • Lorries with yearly distance travelled more than 100 000 km and load capacity over 13 tonnes
• Lorries registered in Gotland are divided into two groups according to load capacity; 13 tonnes or less and over 13 tonnes

The 35 national strata

For the 17 international strata the stratification is done according to the following principles. In a first step road tractors that fulfil the following criteria are placed in a separate stratum (200000):

• Lorry owned by company with 16 or more permits for international traffic.
• The lorry is younger than 11 years old
• The lorry is registered in NUTS2 region SE11 or SE12 or the counties Blekinge or Västra Götaland and has a yearly total distance travelled over 100 000 km and a load capacity over 10 tonnes
• The lorry is registered in the county Skåne or Halland and has a load capacity over 10 tonnes

In a second step the additional 16 strata are created by dividing the lorries into four geographic regions according to:

• NUTS2 region SE11, SE12 och SE21
• NUTS2 region SE22
• NUTS2 region SE23
• NUTS2 region SE31,SE32 och SE33

For each geographic region, road tractors are divided into a separate stratum regardless of yearly distance travelled or load capacity. The remaining lorries are divided into three strata according to yearly distance travelled and load capacity according to:

• Lorries with yearly distance travelled 100 000 km or less and load capacity 10 tonnes or less
• Lorries with yearly distance travelled 100 000 km or less and load capacity 10 tonnes or more
• Lorries with yearly distance travelled more than 100 000 km

 The 17 international strata:

Allocation of the sample:

The initial quarterly total sample size is 2800 vehicles. These are allocated according to

• 1800 to national strata
• 150 to stratum 200000
• 850 to international strata

The allocation among strata within the national and international strata is Neyman allocation according to the variables tonnes based on data from the eight most recent available quarterly data. If a stratum, according to the allocation principle, obtain fewer than 18 vehicles, the sample size is raised to 18 vehicles. This means that the actual quarterly sample size is slightly larger than 2800, normally slightly more than 100 extra vehicles which gives a quarterly sample size of slightly more than 2900 vehicles.

Sampling procedure:

The vehicles are randomly selected within strata according to a simple random sampling procedure. In a second step the each vehicle is assigned a randomly selected week during the quarter. The week is also selected accoring to a simple random sampling procedure. 

Data collection and weighting

Recording of journey data sent to Eurostat:

Single stop: The respondent is allowed to record the main type of goods if there are several types of goods. Otherwise the respondent will record mixed goods for such a journey.

Multi stop: Multi-stop journeys are coded by consignments. The respondent records each basic transport operation in the questionnaire. These records are then recalculated to journey level by the producer. The method used can be described as follows:

 The tonne-kilometres per consignment (each basic transport operation) is calculated by multiplying tonnes with travelled distance (km).

• The tonne-kilometres for each congsignment is added to a total tonnes-kilometres for the journey
• The kilometres driven for the journey is registrated by identifying the distance between loading point and unloading point (normally not the same as adding the km driven for each basic operation)

Then the tonnes for the journey is caclulaed according to

A / B = C

A = Tonne-kilometres for the journey

B = Kilometres driven on the journey

C = Average tonnes for the journey

Tonne-kilometres will be the same regardless of which file, A2 or A3, that is used for their calculation. Exact figures in kilos are used in the calculations.

 The type of goods for the total journey is calculated as the main type of goods (in respect of kilos). 

Regarding journey type 2 (multi-stop-journeys), we use the principle that if a trailer was used for the first consignment of the journey a trailer was used for the entire journey.

Collection/delivery: In the Swedish survey we allow the respondents to decide if the journey can be seen upon as a collection (c) or distribution (d) round or a combined collection/distribution round (c/d). If the journey consists of five or more stops the respondent is allowed to give information on the journey as a whole. The respondent is asked to indicate the c/d-round with a “D” for distribution or “U” for collection (the U refers to the swedish wording) or “DU” for combined distribution/collection rounds in the questionnaire. The respondent is also asked to registrer the number of stops (or approximate number if there are hundreds of stops). If the journey is considered as a c, d or c/d-round the respondent is asked to indicate the total weight for the round as a whole, the total kilometres driven during the round and the main commodity group. In the instructions to our respondents it is stated that the round is considered to start at the first loading point and finished at the last unloading point. This means that the possible empty leg must be recorded as a separate journey before and/or after the round. The information from the Swedish survey in the A2 file and the A3 file is the same regarding type 3 journeys. The tonne-kilometres are calculated according to the principles in the manual vol 1 2016 in section 6.5. More specifically, for delivery rounds (only) the type 1-principle is used. For collection rounds (only) the type 2-principle is used. For combined delivery and collection rounds the type 5-principle is used.

Special notes on some variables:

The respondents are asked to fill in the UN-number instead of the ADR-number for hazardous goods. The UN-number is then converted into ADR-number.

Other variables: Regarding trailers we allow the respondent to record the most common trailer or combination of trailers used during the week for measurement.

Calculation of weighting factors:

The weighting factor is based on the methodology called straight expansion within strata. However, since 2014 the weighting factor is multiplied with an inflation factor that accounts for false reporting of no activity. False reporting means that for a certain vehicle the reply is that no activity was performed during the measurement week, when in fact activity was performed, i.e. a false reply. Inflating the weighting factors leads to a raise in the estimates with about 30 percent. These higher levels are better estimates of the true levels of the parameters, e.g. total km driven. The weighting factor is given by

h = is the index for stratum, h=1,2,…,H (and H=52)

N_h = the number of vehicles in stratum h

m_h = the number of responding vehicles in stratum h. A vehicle is regarded as responding if it belongs to category B1:5 (vehicles responding with journey data) or B1:6 (vehicles responding without journey)

g = is the index for stratum in the No Activity Survey (help survey) g=1,2,…,G (and G=11).

w_g = an inflation factor that account for false reporting of no activity

Remark regarding N_h: For a certain quarter the number of vehicles in a stratum is taken from the following quarter (to account for under-coverage, se 13.3 above). Example: for quarter 1 the numbers of vehicles in a stratum is taken from the frame from quarter 2. The rationale behind this is that the frame for Q2 originates from February 15, i.e. the midpoint of Q1 in time. The number of vehicles at the midpoint of Q1 is a better source for the population size than the number of vehicles in the frame for Q1 which is originated from November 15 the previous year. This method agrees with the suggested method in the reference manual vol 1 2016 chapter 7.2.2.

Remark regarding w_g: A parallel help survey called the No Activity Survey (NAS) is performed together with the ordinary Road Freight Survey (RFS). The target populations are the same and the same frame is used in both surveys. The sample size in the NAS is 500 vehicles each quarter and in the Road Freight Survey (RFS) about 3 000 each quarter. The stratification in the NAS is based on company characteristics rather than vehicle characteristics which are used in the RFS. If a vehicle is selected in the RFS it is non-eligible in the NAS for one year. The reason for the NAS is that there are (strong) indications that the amount of no activity is too large in the RFS Survey. If a vehicle falsely reports no activity, when in fact journeys were performed, the estimates of e.g. total km driven will be underestimated. In the NAS, performed by telephone, only one question is posed: “Did you use the vehicle for goods transportations last week”. Since only one question is posed we believe that an accurate answer is obtained. Based on each survey, NAS and RFS, the proportion of vehicles with activity and no activity can be estimated. The ratio between the two estimates of proportion of vehicles with activity forms an inflation factor w_g. If, for example, the proportion of vehicles with activity in the NAS is 0.85 and 0.70 in the RFS (in a certain stratum), then the inflation factor is w_g = 0.85/0.7 = 1.21, i.e. a raise of 21 %. Since the sample size in the NAS is fairly small the inflation factors can vary between quarters. A stabilizing procedure is used. For a certain quarter Q a weighted average of the inflation factors from year t, t-1 and t-2 is calculated. One (weighted) inflation factor w_g is calculated for each stratum g and is multiplied with the straight expansion weight 13×(N_h/m_h). Since the stratification is different in both surveys the weighting factors will not be constant within a stratum h. Hence no weighting factors can be presented in the supplementary table B1.

Estimation of maximum permissible laden weight:

The variable maximum permissible laden weight regarding the vehicle is register based information. The maximum permissible laden weight for the trailer or semi-trailer is collected through the questionnaire. If no trailer or semi-trailer is used the maximum permissible laden weight registered in the A2 dataset (variable A1.4) is thus only based on register information. If a trailer or semi-trailer is used, the maximum permissible laden weight for the entire vehicle configuration is calculated as the sum of the vehicle and trailer/semi-trailer maximum permissible laden weight.

Optional variables covered:

• Vehicle empty kilometres
• Type of cargo
• Vehicle operator’s NACE category of activity (register information)
• Axle configuration (register information regarding vehicle, collected information regarding trailer or semi-trailer)
• Degree of loading of the vehicle

Additional variables compared to the legal requirements:

A1. Vehicle-related variables:

• possibility of using vehicles for combined transport: No, not delivered in dataset A1
• vehicle operator’s NACE Rev. 2 at class level (four-digit level); Yes, delivered in dataset A1 - (register information)
• vehicle empty kilometres : Yes, delivered in dataset A1

A2. Journey-related variables:

• axle configuration according to the nomenclature defined in the Regulation on road transport statistics: Yes, delivered in dataset A2
• place of loading, if any, of the road transport vehicle on another means of transport: Yes, delivered in dataset A2
• place of unloading, if any, of the road transport vehicle from another means of transport: Yes, delivered in dataset A2
• situation ‘fully loaded’ (procedure 2) or ‘not fully loaded’ (procedure 1) of the goods road transport vehicle during the journey in question, in terms of maximum volume of space used during the journey (procedure 0 = by convention for unladen journeys): Yes, delivered in dataset A2

A3. Goods-related variables:

• type of freight (Cargo types) as defined in the Regulation: Yes, delivered in dataset A3

Procedure for reminders:

First reminder: sent out by post after one week and 3 days after the due date.

Second reminder: sent by post after another week.

Third reminder: performed by telephone after another week. The telephone reminder process continues for two weeks.

A normal figure for the response rate is around 60 percent. Considering the potential effect the non-response might have on the estimates, we judge the response rate to be satisfactory for the purpose of the survey.

Some main figures regarding frame, sample and data collection

National level :                   

The data is controlled at several stages and in numerous ways at the national level (in Sweden). Briefly, the data in the paper questionnaire and in the web questionnaire are manually controlled by staff that register the data as well as automatically controlled. If necessary the vehicle owner is contacted for clarification and/or correction. At the end of a quarter, the data is controlled programmatically regarding allowed codes and amounts as well the variables logical consistency. The vehicle owners can be contacted at this stage to for clarification and/or correction. After this step, the data is processed in several steps before a final database is obtained. Among other things, the weighting factor is calculated and some imputation is performed. Each step in the procedure is controlled and monitored. The final data is then delivered to Eurostat.

From Eurostat :

Since 1999, micro-data from the reporting countries have to be submitted according to Commission Regulation 2163/2001. The data are then checked and validated by EUROSTAT (verification of many different codes used (NUTS 3, numeric or alphabetic variables) correctness of linked questionnaires in the different dataset, etc…). Detected errors are then reported back to the data sender with the request for correction, this is an iterative process until at least 99.5% of all data records are validated and loaded in the database.

 In section 18.4 the data editing or data processing was breifly described. In 18.3 the calculation of weighting factor was described. 

Road freight data is not seasonally adjusted.

However, one aspect regarding adjustment can be mentioned. If a year contains 53 weeks the last quarter contains 14 weeks. This is taken into consideration in the estimation procedure in Sweden. At first glance it might seem straight forward to replace the expansion factor 13 in the weighting factor (se section 18.3) with the factor 14. But that is not how it is handled in Sweden.

First, we establish what parameter we are interested in. Let t denote the true parameter for a total during a quarter containing 14 weeks, e.g. total tonnes loaded during 14 weeks. We introduce a “normal quarter” adjusted parameter that is defined as t_n=(13/14)×t. The parameter t_n can be  interpreted as the down-adjusted total of e.g. tonnes loaded as if the quarter would have consisted of 13 weeks. t_n is (obviously) always smaller than t. The rational for the formulation of the parameter t_n is that comparing the parameter t the 4^th quarter a certain year (that consists of 14 weeks) to the parameter t the 4^th quarter previous year (that consists of 13 weeks) is misleading. A larger value in e.g. tonnes loaded for current year compared to previous year is not due to more activity in the sector, but rather due to an extra week in the quarter. Hence, the parameter t_n is more relevant for comparisons.

How do we estimate the parameter t_n? The weighting factor when estimating parameter t if the quarter containes 14 weeks is 14×(N_h/m_h)w_g  (this is due to the fact that the selection of 1 measurement week should be expaneded to 14 weeks). Then, estimating the “normal quarter” adjusted parameter t_n the weighting factor is (13/14)× 14×(N_h/m_h)w_g=13×(N_h/m_h)w_g.

The conclusion is that the “normal quarter” adjusted parameter t_n is estimated by using the “ordinary” weighting factor with factor 13. A more thorough technical description can be found in the Swedish technical documentation.

Sweden has no other comments than the ones given on Eurostat Road freight transport measurement (road_go) Reference Metadata in Euro SDMX Metadata Structure (https://ec.europa.eu/eurostat/cache/metadata/en/road_go_esms.htm)