URBAN PLANNING GROUP

Assumptions about the relationship between urban structure and travel behaviour have found their way in planning concepts and policies. A good example is the New Urbanism movement, which proposes new urban design principles to cope with urban sprawl, the predominance of the car in every day life and the liveability of neighbourhoods [Katz, 1994]. Likewise, several countries have developed policies based on assumed, as opposed to empirically validated, reasoning concerning the relationship between urban form and transportation. For example, the Dutch government has developed a location policy for businesses and institutions [Ministry of Housing, Spatial Planning and the Environment, 1990], which regulates the type of location where particular businesses and institutions should be located. This location policy is based on the central assumption that more people will use public transport to go to work if it is made available close to their place of work.
    The nature of the relationship between travel behaviour and urban structure has also been subject of academic debate. Some scholars believe in the effect of urban planning on travel behaviour and state that good planning is bound to be successful [Novem, 1997], others are more cautious and provide empirical evidence that the effect of urban structure on travel patterns is limited at best, unless major boundaries are crossed [Timmermans & van der Waerden, 1997].
    An extensive bibliography on this subject was published in 1997 [Handy, 1997]. The overall conclusion of this bibliography was that the most popular assumptions and beliefs about the relationship between urban structure and travel behaviour are generally supported by empirical studies, in particular those on the effects of density. Other findings were that the positive effects of mixed land use could not consistently be demonstrated and that decentralisation of jobs tends to result in shorter trips but higher shares of car use. As far as the overall structure of the city is concerned, most studies conclude that polycentric forms result in fewer trips, shorter trips and less energy use than monocentric and dispersed forms.
    The research conducted to date poses us with two problems. First, it focuses mainly on either urban characteristics at a large scale and total vehicle miles travelled, or on small scale neighbourhoods in detailed case studies including only a few neighbourhoods, but these two scales are rarely combined in a single study. Consequently, many previous studies are suspect of basic methodological flaws. Secondly, because most research is conducted in countries other than the Netherlands, the results do not necessarily apply to the Netherlands. The organisation of Dutch cities differs in several respects from cities in the United States, where most of the studies were conducted. Moreover, bicycle use is less pronounced in other countries and therefore hardly reflected in the research results.
    In our research project, we try to provide a solution to these shortcomings by collecting data, specific to Dutch cities, from a multi-level perspective. In an extensive survey, we have collected activity and trip data in several neighbourhoods in 9 Dutch cities across the country. This paper reports the results of a first analysis of these data, collected in the Fall of 1997. In particular, we focus on two kinds of trips, home-to-work trips and daily shopping trips. We expect that the former trips generally go beyond neighbourhood borders, whereas the latter trips stay mainly within the neighbourhood or involve adjacent neighbourhoods. Limiting our analyses to these two kinds of trips may thus give us a quick, albeit perhaps still rough, indication of the effects of urban physical characteristics on travel behaviour at both the level of the neighbourhood and the city. It should be emphasised, however, from the very beginning that ultimately a multi-level model is required to better capture the exact relationship, if any, between urban structure and activity patterns.
    This paper is structured as follows. In the next section the research approach will be discussed. Hypotheses will be formulated, the data collection including the selection of variables will be outlined and the form of analysis will be described. Next, we will describe the results of the analysis, followed by a section in which we interpret these results and draw conclusions. Finally, we will give some recommendations for further research.

As mentioned in the introduction, this paper is based on analysis of data that we collected in the Fall of 1997. In this section we will justify and describe the choices underlying the data collection and the analysis. First, we will formulate hypotheses and assumptions that underlie the current research project. Then, we will discuss the data collection process in more detail.

hypotheses and assumptions

home-to-work trips
We assume that the frequency of home-to-work trips does not depend on urban structure. The number of these trips is mainly determined by the kind of jobs (full-time versus part-time) and the labour participation in that neighbourhood. Mode choice for home-to-work trips, and the distance or travel time is more likely to depend on the physical attributes. The urban form, the location of industrial parks, the distance to the city centre, the nature of the local and city-level road networks and available public transport may potentially all be of influence. Because density has often been identified as a major determinant, we decided to include this variable in our analysis as well.
    Proximity of an industrial/office park of a certain size is assumed to lead to shorter home-to-work trips and hence to larger shares of walking and cycling. A high employment rate in the city is assumed to have similar effects. If there are more jobs in the city itself, fewer people are forced to commute to other cities. The proximity of a railway station is assumed to lead to a larger share of public transport in the modal split, and could possibly lead to longer trips because a good accessibility opens up a larger geographic area to search for a preferred job. Housing and population densities are expected to correlate with higher shares of public transport since they lead to a larger market for the commercial exploitation of a public transport system. The availability of such a system in turn is assumed to generate a higher use of public transport.
    The nature of the relationship between different types of road networks and the travel patterns is more difficult to postulate. For instance, it is generally assumed that tree and loop networks are more advantageous for walking and cycling than, for instance, grid networks, which are sometimes viewed as being unsafe and force people to make detours. However, grid networks also represent easy to understand structures. A certain effect of the type of network is thus assumed, but we expect the size of this effect to be small.

daily shopping trips
Some of the assumptions expressed for home-to-work trips may also hold for daily shopping trips. In particular, we expect this to be the case for transport networks. Since the accessibility of the shopping facilities may be influenced by the type of transport network, the number of shopping trips, their length and the associated transport mode may be correlated to the type of network. In addiction, we assume that the local availability of a shopping centre leads to an increase of shopping trips, to shorter trips and to more non-motorised transport. Finally, we expect that higher population and housing densities may be related to a more compact distribution of shopping facilities and hence will lead to shorter travel distances and therefore to higher shares of non-motorised transport and shorter travel times.

social characteristics
Apart from physical attributes, various social characteristics are also very likely to influence travel behaviour. At the individual level, household type, education level, income, gender and car availability may be determinants of travel behaviour.

data collection
The data used in the present study were collected by administrating an activity diary using a paper and pencil survey. Respondents, randomly selected from the neighbourhoods in question, were asked to report various aspects of their everyday trips, e.g. for work, school and shopping. These aspects included the number of trips, their origins and destinations, the mode choice and the travel time. Respondents were also asked to keep a two-day-activity-diary. However, the diary was not used for this paper. In total, we approached 300 households per neighbourhood with the question if they were willing to cooperate. Of the 594 households who agreed in principle to complete the diary, 344, involving 586 respondents, returned the questionnaire.
    In addition to the survey and diary data, information was collected for each city and neighbourhood about their physical structure. These data were obtained from official sources such as the Dutch Bureau for Statistics [CBS, 1996; CBS, 1998], the municipalities involved and several other sources of spatial information (maps [Falkplan; ANWB, 1996]).

cities and neighbourhoods
To allow a more rigorous test of the relationship between urban structure and activity patterns, the selected spatial entities should vary systematically in terms of the selected physical attributes which are assumed to measure key aspects of urban form and urban structure. Effects that appear at some level of spatial aggregation may cancel out at another level. Hence, to allow for multi-level analysis, we selected 9 Dutch cities using three criteria. First, the cities all had to be of approximately the same size: between 100.000 and 150.000 inhabitants, which is indicative of a medium-sized Dutch city. Secondly, we selected cities that offer a maximum difference in terms of urban form. Finally, cities needed to differ in terms of the type of main transport network. Within these cities neighbourhoods were selected, based mainly on their location vis-à-vis the city centre and the railway station. We refined our choices based on characteristics of the neighbourhoods, such as road network types at the neighbourhood level and the proximity of sub-centres. Table 1 gives a quick overview of the selected cities and neighbourhoods and their main features.

dependent variables
The following dependent variables were chosen for analysis.

• mode choice for home-to-work trips

Although our survey included the option to report up to three variants for home-to-work trips, we focused this analysis on the characteristics of the most frequently made trip. The transport modes were recoded into 3 types: motorised private transport (e.g. car, motorcycle), non-motorised transport (walking and cycling) and public transport (train, bus, underground and tram).
• duration of home-to-work trips

The duration concerned the most frequently made one way trip from home to work. For some of the analyses, trip duration was recoded into 3 categories: less than 15 minutes, 15 to 30 minutes and more than 30 minutes.
• mode choice for daily shopping trips originating from home

In contrast to home-to-work trips, respondents were asked to report their major mode choice for 3 different shopping locations and for different times of the week:  weekdays during daytime, weekdays in the evening and Saturdays. The analysis focused on the most frequently visited shopping location, and on trips made on weekdays during daytime and Saturdays. Mode choice for shopping was recoded into only two categories: motorised private transport and other transport modes (mainly walking and cycling).
• duration of daily shopping trips originating from home

Trip duration was analysed for the selected shopping trips. For some analyses, duration was recoded into two categories: 5 minutes or less and more than 5 minutes.
• frequency of shopping trips originating from home

To analyse the frequency of shopping trips, the total number of shopping trips for daily goods made per month was calculated.

social attributes
In any empirical analysis, it is difficult to conclude whether a significant relationship between urban structure and travel patterns is caused by aspects of urban form of the neighbourhood of by difference in the spatial distribution of population characteristics which may also be related to travel patterns. Therefore, several social attributes were also used as independent variables in the analysis. At the level of the individual respondent, household type, education level, income, availability of motorised transport and gender were selected. Respondents were classified in three household types: single household, couples and households with child(ren). Education level was classified as follows: elementary education or less, secondary (vocational) education and  higher (vocational) education. Personal net income was used as a measure of income and recoded into 5 classes. Finally, average neighbourhood income was used as a socio-economic variable for the whole neighbourhood.

techniques used for analysis
In order to test the relationship between the dependent and independent variables described above, two simple statistical techniques were used. For nominal and ordinal variables crosstabs and Cramer’s V value (a chi-square based measure) were calculated. For ratio variables one-way ANOVA was applied. The results of these analyses are reported in the next section. We will only discuss the results that are significant at the 95% confidence level.

home-to-work trips
The overall modal split for home-to-work trips is 45.5% motorised transport, 40.5% non-motorised transport and 14.0% public transport. For 43.1% of the respondents, it takes 15 minutes or less to go to work, for 28.6% it takes 15 to 30 minutes and 28.3% travels for more than half an hour to go to work. Average travel time is 27 minutes.

mode choice
Mode choice for home-to-work trips is significantly correlated with 8 independent variables, 5 of which are of a physical nature. At the level of the city, the city, its location vis-à-vis to the Randstad Holland and its basic urban form are significant factors. The cities Enschede, The Hague, Almere and Zoetermeer show the highest share of car use (over 50%) whereas Leeuwarden and Eindhoven have shares of less than 40%. Leeuwarden also has a high share of non-motorised transport (over 50%), together with Apeldoorn. Almere is obviously a divergent city, with a high share of motorised transport and a very high share of public transport, 36% against an average of 14%. The cities located outside the Randstad Holland, on average, have higher shares of non-motorised transport than those located within this area (49% against 28%).
    Looking at urban form the crosstabs show that the lobe-cities seem to be most friendly towards alternatives for the car since they have the lowest share of motorised transport (42%). The polycentric city has a high share of public transport (36%) and a very low share of non-motorised transport (11%). The employment rate is also significantly correlated with mode choice. The higher this rate the higher the share of non-motorised transport rising steadily from 12% to 61%.
    At the neighbourhood level, the local road network is significantly correlated with mode choice. The grid structure has the lowest share of motorised transport (33%) and the highest share of non-motorised transport (55%).
    Concerning the social characteristics of the respondents and their neighbourhood, it is not surprising that the availability of motorised transport is significantly correlated with mode choice. Approximately 50% of the respondents who have a car (or motorcycle) actually use it for their trip to work, which is 5% more than on average. Moreover, with an increase in personal income increases the share of motorised transport increases (from 21% to 58%). Finally, men show higher shares of motorised transport (50%), and women higher shares of non-motorised transport (50%). Note that there is only a weak correlation between income and gender in our dataset.

travel time
Travel time for home-to-work trips is correlated with 10 independent variables, 8 of which are of a physical nature. At the city level, the city, its location vis-à-vis the Randstad Holland, its basic urban form, its main road network type and the employment rate are significant factors. The cities Apeldoorn, Arnhem, Enschede and Leeuwarden have relatively high percentages of less than 15 minutes travel time (more than 50%). Almere, on the other hand, has a high share of over 30 minutes (67%). Respondents in cities outside the Randstad Holland have shorter travel times (52% fifteen minutes or less) than inside the Randstad Holland (45% over 30 minutes). As was the case with mode choice, the cities outside the Randstad Holland again seem to perform better. The polycentric city has a large share of over 30 minutes (67%) and a small share of 15 minutes and less (15%). It seems that the polycentric form poses its inhabitants with the longest distances to travel. It must be said however that Almere is the only city with the polycentric form. Cities with a ring structure have high shares of over 30 minutes (38%), contrary to cities with a grid or radial structure, which show high shares of 15 minutes and under (19% and 23%). The effect of employment in the city is also visible: more jobs per 1000 inhabitants is related to shorter travel times. With the lowest employment rate only 15% of respondents live within 15 minutes of their work, for the highest employment rate this is 59%.
    At the neighbourhood level, the local road network, the distance class to the city centre and population density are significantly correlated with travel time. Neighbourhoods with a loop/grid structure have a high share of over 30 minutes travel time (62%). Neighbourhoods with a grid structure, again, perform best and have a high share of 15 minutes and less (57%). Distance class 1 neighbourhoods have a high share of ‘15 minutes or less’ travel time (65%), whereas neighbourhoods in classes 3 and 6 have high shares of ‘more than 30 minutes’ (approximately 40%). As far as population density is concerned, the category of ‘40 to 60 inhabitants per hectare’ has a high share of ‘over 30 minutes’ travel time (37%), while the category of ‘60-80 inhabitants per hectare’ shows a low share of over 30 minutes travel time (15%).
    Social characteristics that are correlated with travel time for home-to-work trips include personal income of the respondents and their gender. Higher incomes have a higher share of longer travel times (up to 47%). Since higher income also correlates with a higher share of motorised transport, it seems to be that people with higher incomes live further from their jobs. Women have a higher share of shorter travel times than men (47% against 37%).
    When considering the difference in average travel time for different categories of the independent variables by using a one-way ANOVA, the results are roughly the same as for the Cramer’s V. However, ANOVA identifies a significant difference in average travel times between the neighbourhood (F=2.9; F-prob.=0.0001), which was not found by Cramer’s V.

note!
The analysis of the data showed that the city of Almere and its neighbourhoods are quite different from the other cities. This caused us to run the same analyses again but excluding the data from Almere. The results were quite striking. For mode choice, the only significant correlation left concerns the local road network. For travel time, the only significant spatial factor left is the location of the neighbourhood vis-à-vis the Randstad Holland. This leads us to the tentative conclusion that the examined relationships are either not significant or weak at best.

daily shopping trips
For daily shopping, each analysis was conducted twice, once for trips made on weekdays during daytime and once for trips made on Saturdays.
    The results indicate that the modal split for weekdays and Saturdays differs slightly: on weekdays 29.7% of respondents use motorised transport whereas on Saturdays this figure is 36.4%. Non-motorised transport amounts to 70.3% on weekdays and 61.8% on Saturdays. On weekdays, the travel time for shopping is slightly higher than on Saturdays: 7.6 minutes against 7.0 minutes. This corresponds with the fact that on weekdays 61.8% of the respondents take 5 minutes or less to go shopping, and 38.2% more that 5 minutes, whereas on Saturdays these percentages are respectively 70.2%  and 29.8%. This is most likely related to the fact that on Saturdays the use of motorised transport is slightly higher than on weekdays.

mode choice
The influence of physical attributes defined at the city level on mode choice for daily shopping trips is even more limited than the effects found for home-to-work trips. Only the distance class to the city centre is a significant factor for both times of the week. Especially class 1 neighbourhoods show very high shares of non-motorised transport (for weekdays even 100%). For Saturdays, class 1 and 2 neighbourhoods have relatively high shares of non-motorised transport (over 80%), neighbourhoods in the classes 3 and 6 have high shares of motorised transport (over 50%). For shopping on Saturdays, the location of the city vis-à-vis the Randstad Holland, the urban form of the city and the urbanisation class of the neighbourhood are also significantly correlated with mode choice. Outside the Randstad Holland the share of non-motorised transport is higher (69% against 56%). The polycentric city has a high share of motorised transport (63%), whereas the lobe city has a low share (26%). The highest and lowest urbanisation classes (1 and 4, class 5 is not included in this analysis) have low shares of motorised transport (27% and 35%), whereas the middle classes (2 and 3) have high shares (40% and 48%).
    At the neighbourhood level, the main road network of the neighbourhood and its local road network as well as the availability of shopping facilities are relevant factors. For both the main road network and the local road network, the grid structure has the lowest share of motorised transport (weekdays 13% and 12%; Saturdays 33% and 20%). Neighbourhoods with a loop structure for a main road network have the highest shares of motorised transport (weekdays 58%; Saturdays 65%). The local availability of shopping facilities has a significant influence. When a city centre or city district shopping centre is available in the neighbourhood, the share of non-motorised transport is very high (up to 100%). When there are only a few shops available, the share of motorised transport is high, especially on Saturdays (56%). The same holds for shopping on Saturdays for the large neighbourhood shopping centre (also 56% motorised transport). For shopping on weekdays, the housing density of the neighbourhood also proves to be a relevant factor. High density areas have higher shares of non-motorised transport, and low density areas have higher shares of motorised transport (up to 100%).
    Three social characteristics are significantly correlated with mode choice for daily shopping: car availability, average neighbourhood income and personal income. On weekdays, 35% of people with a car available use it for shopping, on Saturdays this figure rises to 44%. For neighbourhood income, it is not easy to detect a direction in the relationship. Results are quite dispersed. Personal income is only a significant factor for shopping on weekdays. Higher personal income coincides with a higher use of motorised transport. However, the highest income category is an exception to this rule.

travel time
Travel time for daily shopping is only significantly correlated with a few independent variables. For shopping on Saturdays, there are no significant correlations at all. For shopping on weekdays, three physical attributes defined at the city level and two at the neighbourhood level have a significant correlation. There are no significant correlations with any of the social characteristics. The cities of Eindhoven and Leeuwarden have relatively short travel times (more than 75% take 5 minutes of less), while Arnhem has relatively long travel times (62% takes more than 5 minutes). High population densities tend to correlate with shorter travel times (up to 74% five minutes or under). As far as urbanisation class is concerned, classes 1 and 3 have short travel times (69% and 74% five minutes and under), and 2 and 4 have long travel times (45% and 55% over 5 minutes).
    At the neighbourhood level, the main road network for the neighbourhood and its local road network are both significantly correlated with travel time. For the main road network, the axial structure is associated with longer travel times (53% over 5 minutes), while the tangential structure correlates with shorter travel times (74% five minutes and under). For the local road network, the grid structure (again) shows higher shares of short travel times (73% five minutes and under), while the loop/tree structure correlates with longer travel times (52% over five minutes).

number of shopping trips
The total number of shopping trips per month is also hardly correlated with any of the independent variables. Only the local road network of the neighbourhood and the gender of the respondent are significant factors. Especially the tree structure and the loop/grid structure are different with 14 trips per month against 21 trips per month.  Finally, and not surprisingly, women make more shopping trips than men.

note!
Given the effect the two neighbourhoods in Almere had on the results for home-to-work trips, we decided to rerun the analyses for daily shopping trips without the data from Almere. As expected, the effects were not as drastic as in case of the home-to-work trips. For mode choice and travel time during weekdays all correlations were still significant. For mode choice on Saturdays the correlations with urban form, main neighbourhood road network and the urbanisation class were no longer significant. The correlation of local road network with number of shopping trips was no longer significant. These results lead us to the general conclusion that the correlations for weekdays are stronger than for Saturdays.

In this paper, we examined the existence of a relationship between urban structure and activity patterns. Our preliminary analysis suggests that this relationship is weak at best, although we did find some evidence of a significant relationship. It should be realised that some relations tested are bound to be significant for statistical reasons only. Moreover, many of the significant relationships we found between aspects of urban structure and aspects of activity patterns for trips to work, and some for shopping trips, disappeared when one extreme city was dropped from the analysis, indicating that this city was an outlier. In this section, we will reflect on the assumptions and hypotheses posed at the start of this paper and we will draw some tentative conclusions based on the results presented in the previous section, including the data from all cities.

home-to-work trips
Most striking in our analyses of home-to-work trip characteristics is the absence of a correlation with the location of industrial estates. Although frequencies indicate that the existence of a large industrial/office park nearby coincides with shorter travel times and higher shares of walking and cycling, this correlation is not significant. On the other hand, the distance to the city centre seems to be of some importance. The employment rate in the city proved to be a relevant factor, just as postulated. The results support policies that advocate a self-containment of cities with respect to employment. However, it does not seem to be necessary to aim at an even distribution of these jobs over the city. The proximity of a railway station did not have the expected effect, not on mode choice nor on travel times.
    Considering urban form and road network types, we can conclude that both are weakly related to home-to-work trips. The main road network type for the city as a whole is only correlated with travel time. The grid and radial structured cities show relatively short travel times, while the people in ring structured cities have to travel relatively long to work. At the neighbourhood level, the main road network type is not relevant, but the local road network type is. Overall, the grid structure scores the best. It relates to the highest shares of walking/cycling and the lowest shares of car use, and it offers the shortest travel times. On the whole, the grid structure seems to perform best on both the neighbourhood and city level with respect to desired mode choice and short travel times.
    The correlations with housing density and population density, which are often assumed in the literature, are either absent or weak. In other words, based on these analyses, our research does not support the often postulated positive effects of high density on mode choice and travel time, at least not for home-to-work trips.
    Finally, we compared cities/neighbourhoods inside and outside the Randstad Holland, since Dutch spatial policy focuses for a large part on concentrating new developments in this area. Our research shows that cities outside the Randstad area perform better on both mode choice and travel time. However, this result is most likely spurious, because cities outside the Randstad have a significantly higher average employment rate than cities in the Randstad (521 jobs per 1,000 inhabitants against 354 jobs per 1,000 inhabitants).
    We also looked at a number of social characteristics of neighbourhoods and respondents as a reference for the results on the physical attributes. Household type, education level and average neighbourhood income are not significantly correlated with characteristics of home-to-work trips. Yet, personal income and gender are. Finally, it is quite obvious that availability of motorised transport is correlated with mode choice, yet it is not with travel time.

trips for daily shopping
What is most striking in the results for daily shopping trips is that mode choice and travel time on weekdays are much more ‘sensitive’ to the independent variables than they are on Saturdays. Possible explanations for this are that people are on a different time budget and that the company chosen for this type of activity is probably different on Saturdays, which can influence their mobility choices. There are hardly any correlations between number of shopping trips and the independent variables.
    In our hypotheses we assumed that the availability of shopping facilities in the neighbourhood would influence the number of shopping trips, their length and possibly mode choice. We found that availability is only correlated with mode choice. It seems that the choice of transport mode for daily shopping is driven by a slight preference for non-motorised transport modes, within particular distances.
    Considering urban form and road network types, we can see that both are significantly correlated with shopping trips. Urban form is related to mode choice but not to travel time. The main road network type at the city level is not related to shopping trips. Both the main road network type and the local road network type are related to mode choice and travel time. Again the grid structured neighbourhoods perform best on both variables. The number of trips is also correlated with local road network type.
    We expected that high densities would be related to shorter travel times and higher shares of non-motorised transport. This is only the case for shopping trips made on weekdays.
    The social characteristics examined in relation to shopping trips show no correlation with travel time, neither on weekdays nor on Saturdays, but some do correlate with mode choice. Availability of motorised transport and average neighbourhood income are relevant on both weekdays and Saturdays. For weekdays, personal income is also a factor. Gender is significantly correlated with number of shopping trips.

overall conclusions
In the introduction, we explained why we chose two types of trips for the first analyses of the data set. We postulated that home-to-work trips most often go beyond neighbourhood or even city limits and are more likely to be influenced by physical attributes at a larger scale, that of the city as a whole. In contrast, trips for daily shopping usually stay within the neighbourhood or go to an adjacent neighbourhood, and are assumed to be influenced more by neighbourhood characteristics. The results support these assumptions. Where home-to-work trips are mainly correlated with urban form and road network, employment rates and location vis-à-vis the Randstad, shopping trips are more influenced by main neighbourhood road network and local road network, local availability of shopping facilities and densities of housing and population, although overall these relations appear to be weak.
    Physical attributes that often have a significant correlation with the dependent variables examined, include urban form, road network type at several scales and location vis-à-vis the Randstad Holland. For both the work-related trips and shopping trips the lobe-city proves to give the best results for modal split. As explained in the section on home-to-work trips, the polycentric city takes an exceptional position in this research. Given these results we conclude that the relationship between urban structure and travel patterns in general is weak, unless certain boundaries are crossed, as indicated by the results obtained for Almere.
    The grid structure, on several scales, is often the best option considering both modal split and travel times. This concept, popular in the car oriented developments of the sixties, was abandoned in the seventies and eighties as being pro-car and anti-walking/cycling. However, in the last couple of years the concept seems to be gaining popularity again as being a clear and easy network in neighbourhoods, especially for the non-motorised modes like walking and cycling. Based on our preliminary results, we can support this concept for new developments. We must, however, note that in our data all neighbourhoods with a grid structured local road network are located within 3 kilometres of the city centre, which might have influenced the results.
    Finally, densities and income are often mentioned as factors highly influencing choices in transportation questions. Income appears to have some influence but not more than other factors concerned. Our research tentatively shows that the influence of density is very limited for home-to-work trips and of some significance for shopping trips. It is therefore not evident that a high density is a good concept for controlling motorised mobility.

It should be realised that although the present results show a first snapshot of the data, the analysis was very simple and may even be too simplistic to allow more definitive conclusions about the relationship between urban structure and activity patterns. Some of the independent variables are mutually related, implying that the bivariate relationships we found may be spurious. On the other hand, other causal relations may have gone undetected in the present analysis. More advanced multi-level analyses are thus required to elaborate the assumed relationships. Such multi-variate analyses are required to analyse alternative causal structures and gain a better understanding of the relationship between urban structure and activity patterns. We hope to report on such analyses in the near future.

ANWB [1996], ANWB stratenatlas (ANWB street maps). ANWB,  Den Haag.

CBS [1996], De Landelijke Wijk- en Buurtindeling 1994: kerncijfers (National District and Neighbourhood Data 1994: base figures). Centraal Bureau voor Statistiek, Voorburg/Heerlen.

CBS [1998], Statline: statistical database on the Netherlands. CBS website.

Handy, S. [1997], How land use patterns affect travel patterns: a bibliography. Council of Planning Librarians Bibliography no. 279.

Katz, P. [1994] The New Urbanism: toward an architecture of community. McGraw-Hill, Inc., New York.

Ministry of Housing, Spatial Planning and the Environment, Ministry of Transportation and Waterworks, Ministry of Economic Affairs [1990], Werkdocument Geleiding van de Mobiliteit door een Lokatiebeleid voor Bedrijven en Voorzieningen (Working Document Mobility Conduction using Location Policy for Businesses and Institutions). The Hague.

Novem [1997], Energiebesparing in verkeer en vervoer door ruimtelijke ordening (Energy saving in traffic and transportation through spatial planning). Novem, Delft.

Timmermans, H. and Waerden, P. van der [1997], The Structure of Travel Patterns: an international comparison. Paper prepared for the IATBR Meetings at Austin, Texas, September 1997.