Factors Associated With Young Adults Delaying and Forgoing Driving Licenses: Results From Britain

Objective: To identify the reasons that young adults (age 17–29) in Britain delay or forgo driving license acquisition. Methods: Using year 2010 British National Travel Survey microdata, we first analyze self-reported reasons (including their prioritisation) for not holding a full car driving license and then estimate a logistic regression model for license-holding to investigate additional factors, several of which extend from previous studies. This study also employs a novel segmentation approach to analyze the sets of reasons that individual young adults cite for not driving. Results: These results show that, despite the lack of a graduated driving license system at present, many young adults indicate that issues associated with the driving license acquisition process are the main reason they do not hold a full driving license. About 3 in 10 young adults can be interpreted as not viewing driving as a priority, though half of those without a license are either learning to drive or are deterred principally by the cost of learning. We calculate that after their 17th birthday (the age of eligibility for a full driving license) young adults spend a mean of 1.7 years learning to drive. Young adults citing the costs of insurance or car purchase are likely to cite them as secondary rather than the main reason for not driving, whereas those citing physical/health difficulties are very likely to cite this as the main reason they do not drive. Two distinct groups of young people are identified that both indicate that costs deter them from driving—one group that is less well off financially and that indicates that costs alone are the primary deterrent and one that reports that other reasons also apply and is better off. Status as an international migrant was found to be an important factor, net of confounding variables, for identifying that a young adult in Britain does not hold a driving license. Further research is needed to understand the relative saliency of plausible causal mechanisms for this finding. We also report that both personal income and household income are independently positively associated with license-holding but that (intuitively) the relationship of license holding with a young adult's own personal income is the much stronger of the two. Conclusions: On the basis of these findings, it can be concluded that a number of previously underappreciated factors appear to be linked with young British adults not acquiring a driving license.


Introduction
There is growing interest in the decline in young adults' license acquisition rates in a number of industrialized counties. Though it is unclear whether this phenomenon is still ongoing or license acquisition has stabilized, across a wide set of Organisation for Economic Co-operation and Development countries the current rates are well below the historic peaks of a decade or two ago (Delbosc and Currie 2013;Kuhnimhof et al. 2011;Tefft et al. 2013).
This article presents results from Great Britain; Figure 1 shows that the peak rate of license-holding for young adults (age 17-29) was in the early 1990s, after which a period of decline in the late 1990s was followed by a more or less steadystate pattern in the 2000s.
Researchers studying the historic drop in license acquisition rates generally appeal to one of two methods: attempting to draw statistical inferences from analysis of survey data sets where both driver status and hypothesized correlates are observed (e.g., Delbosc and Currie 2014;McDonald and Trowbridge 2009;Shults and Williams 2013;Sivak and Schoettle 2012), or collecting and analyzing data in which unlicensed young people are asked explicitly why they do not have a driving license (Schoettle and Sivak 2014;Tefft et al. 2013). The advantage of the former strategy is that it makes use of direct observations (revealed-preference data) of relevant personal attributes, whereas the latter relies on self-reported motivations. In the latter case, the researcher may tailor the questionnaire instrument to the specific needs of the research question but runs the risk of missing aspects of the issue that they do not design the instrument to take into account. This article draws on both of these complementary methods, using data from the 2010 British National Travel Survey (NTS) (Rofique et al. 2011). As with earlier studies in this line of enquiry, we also rely on cross-sectional data and so cannot assert that the associations we report are causal. The relationships we show are, however, in a number of cases consistent with the international literature and extend from it in several dimensions.
The rest of this article is structured as follows. The following section discusses the recent literature regarding young people deferring or forgoing license acquisition. The next section describes the data employed in this study. Then we present results from analysis of young British adults' stated reasons for not having a driving license, followed by results based on a logistic regression analysis of license holding, including a discussion of the findings and the directions these results suggest for future research.

Background
A number of studies in recent years have investigated why young adults' license-holding rates are lower today than in the past (cf. Delbosc and Currie 2013). The longer-term trend in Britain, as can be seen in Figure 1, had been upwards until a reversal in the 1990s. Kuhnimhof et al. (2011) showed that the phenomenon has taken place in other industrialized countries as well. Using U.S. data, Shults and Williams (2013) show that license-holding among high school seniors declined from 86% to 73% from 1996 to 2010, with most of the decline coming after 2006. Though the period 1996 through 2006 saw nearly every U.S. state implement graduated driving licencing requirements, Shults and Williams' (2013) analysis suggests that economic factors are primarily responsible, particularly the financial dislocation from 2007 onwards. Le Vine and Jones (2012) show, however, that in Britain the decline in licensing has been stronger among young men than young women and that young British men's incomes have been in steady decline since 2001, whereas young women's have been stagnant. This is despite growing per capita (all ages) real incomes prior to the 2007 recession. Delbosc and Currie (2014) reached similar conclusions as Shults and Williams (2013) and Le Vine and Jones (2012) regarding the importance of demographic and economic factors, on the basis of logistic regression of license holding and correlates (as opposed to self-reported reasons for not driving) in Melbourne, Australia.
Earlier work in Britain by Noble (2005) shows that pass rates for practical (on-road) driving tests in Britain steadily declined from the early 1990s to the mid-2000s, which was a smooth downward trend with the exception of a one-time increase of a few percentage points just before the introduction of the theory test in 1997. As of 2011, the pass rate for the (computer-administered multiple-choice) theory test is 61% (Driving Standards Agency [DSA] 2012a), down from 65% in 2007. Passing this test is a prerequisite for reserving a timeslot to take the practical driving test, which can take mean a waiting period of up to 3 months (DSA 2013a). The practical driving test has (as of 2012) a pass rate of 47% (up from 44% in 2007; DSA 2013b). A full driving license is awarded upon passing the practical test.
The present study builds on the body of international literature in several ways. First, it draws on the prioritization structure (main vs. secondary) of self-reported reasons for not driving. Second, it provides a novel segmentation approach to highlight patterns that have not previously been identified in the sets of reasons that individual young adults cite for not driving. (The recent studies in the literature that have made use of such self-reported data have employed unidimensional analyses; Le Vine et al. 2014;Schoettle and Sivak 2014;Tefft et al. 2013.) Third, this study establishes the effects of emerging issues not previously investigated in detail, such as international migration status and whether there are distinct impacts of a young adult's own personal income versus income earned by other household members.

Data
The NTS is a large-scale household survey of personal travel that is nationally representative of the British population (Rofique et al. 2011). Data from the NTS are used to generate Britain's official national statistics relating to personal travel (United Kingdom Statistics Authority 2011). The survey consists of both a household interview and 7-day travel diary completed by all respondents, and the sample is a repeated cross section (i.e., different households each year). It has been undertaken by the British Department for Transport and its predecessors in broadly similar form since the first edition in 1965-1966. Following ad hoc NTSs in 1972-1973, 1975-1976, 1978-1979, and 1985-1986, beginning in 1988 the survey has been undertaken on a continuous basis, in part to ensure maximum consistency in the time series (Freeth 2000). Since 2002 the sample size has been approximately 20,000 people annually (the annual sample between 1988 and 2001 was approximately one-third of this size). The NTS's fieldwork is administered in accordance with a rigorous sampling protocol, with the data subsequently reweighted during postprocessing to account for differential response rates. Full details of the NTS's methodology are available in Rofique et al. (2011). This study makes use of data from the 2010 edition of the NTS; analysis was performed using SPSS software (IBM Corporation, Armonk, NY).
Whether an NTS respondent is coded as holding a full car driving license has consistently been defined by their response to the following questions: "Do you hold a full driving license valid in Great Britain to drive either a car, or a motorcycle, scooter or moped? [IF YES] Is it for a car only, a motorcycle only or for both, or is it for a car with special adaptations?" The interviewer is provided with notes instructing them to include "disqualified drivers and international permits/other licenses valid in the UK" in the set of people coded as holding a full driving license (Rofique et al. 2011, p. 82). The interviewer notes are important to interpreting these data because it is legal to move to Britain from elsewhere in the European Union (EU; but not from non-EU countries) and to never (before the age of 70) exchange one's foreign driving license for a Britishissued license.
Unlike many other developed countries where graduated license acquisition regimes have been implemented in recent decades, since the 1930s there have been only 2 main types of driving licenses in Britain (there are exceptions for disabled drivers; DSA 2012b). A provisional license allows the holder to drive with an experienced driver as a passenger (but not without one in the car), whereas a full license has no such restriction.
Self-reported data such as the NTS are inherently subject to reporting biases and errors; an alternative resource we considered is administrative data on license holding. Historic administrative data for Britain are not available in disaggregate form, however (Toft 2012). Even if it were available, historic administrative license-holding data would be subject to different sorts of biases and errors; for instance, they would not record licenses issued elsewhere in the EU that are held by British residents, as the NTS instrument is designed to do. They would, however, include some records associated with people who have died and people who have permanently moved abroad, which is an undesirable characteristic in a data set for use in answering this study's research questions.
For the purpose of this analysis, young adults are defined as those under age 30. The unweighted sample size of people aged 17 (the minimum age for acquiring both a provisional and a full driving license in Britain) to 29 is 2972; the sample is weighted for analysis purposes to be nationally representative of British adults in this age band.
Following identification in the early 2000s of the drop in the license-holding rate among young adults, the British Department for Transport decided to make use of future editions of the NTS to investigate why this has happened. Noble (2005) reported on pilot testing of a set of questions as part of an omnibus survey, questioning that was later incorporated into the NTS protocol. All adult NTS respondents who do not hold a full car driving license and are not learning to drive are asked why they do not drive. They are presented with a showcard with a list of prespecified reasons; if they indicate more than one reason they are asked to identify one as the main reason that they do not drive. Forty-nine percent of young adult respondents who are not currently learning to drive cited only one reason for not driving; the mean number cited was 2.5. When considering the analyses in the next section, it must be kept in mind that NTS respondents' identified reasons for not driving are self-reported and are therefore inherently subject to a variety of errors and biases, such as social desirability bias (cf. Fisher 1993). Particular caution is called for in interpreting the distinction between stated main and secondary reasons cited for not driving; though it cannot be known for certain, it may be difficult for respondents to reliably distinguish between the main and secondary reasons that they do not drive.
In addition to those described previously in this section, the following variables were included in this study. All variables are based on self-reported data from NTS-responding households, with the exception of the urban/rural code and population density variables, which are calculated based on the address of the respondent's residence. Further details regarding variable definitions can be found in Rofique et al. (2011): • Gender • Age (in single years) • A dummy variable of whether a person is employed or not • A dummy variable of whether a person is employed full time or not • A dummy variable of whether a person holds a degree or higher academic qualification • A dummy variable of whether a person lives with a child in his or her household who is at least 15 years younger than him or her, which we interpret as "his or her child(ren)" • A dummy variable of whether a person lives with an adult in his or her household who is at least 15 years older than him or her, which we interpret as "his or her parent (

Analysis of Stated Reasons for Not Holding a License
Fifty-five percent of young British adults in 2010 reported having a full car driving license; the analysis in this section focuses on the remaining 45%.  Table 1 shows the breakdown of not fully licensed young adults by the main and secondary reasons they report for not driving. Those who indicate that they are learning to drive are also shown; in those cases we observe that status but not any other information because they are not asked to self-report any further reasons for not driving.
We see that those learning to drive are the largest single group of unlicensed young adults (28%). They are followed by those who indicate that they are deterred by the costs of learning to drive (23%), and the implication is that for about half (51%) of unlicensed young British adults factors directly associated with the license acquisition process are the main reasons for not driving. Recent studies of license acquisition have not directly addressed the costs of learning to drive, though in Britain it has been shown that the average person acquiring a driving license has taken 47 hours of paid-tuition driving lessons (Wells et al. 2008).
The next most prevalent main reasons for not driving relate to seeing driving as a relatively low priority. Nine percent report that they are not interested in driving, 8% indicate that family or friends drive them when necessary, 7% report that other forms of transport are available, and another 5% indicate that they are "too busy to learn." Thus, just under 3 in 10 (29%) young adults seem to view driving as a low priority. A combined total of 7% cite costs other than learning to drive as the main reason (3% cite the cost of buying a car, another 3% cite the cost of insurance, and 1% cite other general motoring costs). All other reasons are cited as main reasons for not driving by fewer than 5% of unlicensed young adults.
Of the times that each reason is cited as a reason for not driving (either main or secondary), the rightmost column in Table 1 looks at the percentage of the time that it is cited as the main reason. Analysis of this type of data structure-that investigates the relative explanatory power when multiple reasons for not driving are cited by an individual-was not possible in earlier studies, even those that did allow respondents to report multiple reasons. We see that, though physical/health difficulties were cited by only 4% of unlicensed young adults, over 90% of those mentioning this reason cited it as the main reason they do not drive. In other words, few young adults report that health reasons prevent them from driving, but those who do cite their health are very likely to indicate that that is the main reason that they do not drive. A similar pattern was found for the unspecified "other" category (3% cited it, and of these people 82% said it was the main reason they do not drive). At the other end of the distribution are 3 cost reasons-the cost of buying a car (cited as the main reason only 13% of all times that it is cited), the cost of insurance (also 13%), and other general motoring costs (10%). These reasons tend to be cited as secondary factors rather than the main reasons that young adults do not drive. Table A1 (see online supplement) looks at the correlation pattern among the self-reported reasons for not driving. For ease of interpreting the correlation matrix, all not statistically significant correlations have been redacted.
We see in Table A1 that there are strong positive correlations between the 4 cost reasons-citing any cost is associated with also citing other costs. We see strong negative correlations between physical/health difficulties and the 4 cost categories as well as simply not being interested in driving. People mentioning a reason that was not included in the prespecified list on the showcard (other) were less likely to cite the costs of driving as a deterrent to driving. Therefore, people who cited physical/health difficulties, unspecified other reasons, or being uninterested in driving were less likely than other people to also indicate that costs deter them from driving. Indeed, for those citing either health issues or other reasons, there were no statistically significant positive correlations with any of the other reasons for not driving.
We next look at sociodemographic correlates of stated reasons for not driving; to do so, unlicensed young adults who were not learning to drive at the time of their interview were segmented into 3 mutually exclusive groups on the basis of their self-reported reasons for not driving. Main and secondary reasons for not driving were treated identically. For comparison purposes, the same statistics were also calculated for people who are fully licensed and those who are learning to drive. The 5 segments in this analysis were therefore as follows: • Those young adults who hold a full car driving license • Those who are learning to drive • Those who cite at least one cost reason for not driving but none of the non-cost reasons • Those who cite at least one of the cost reasons for not driving and at least one of the non-cost reasons • Those who cite at least one non-cost reason for not driving but none of the cost reasons Table A2 (see online supplement) shows the characteristics of each of these 5 segments. Members of the cost-only segment (mean age 21) tend to be younger than the cost plus noncost (22 years of age) and non-cost-only segments (23 years of age). The cost plus non-cost segment has a larger share of female members (63%) than the cost-only (51%) or non-costonly (47%) segments and the highest mean level of personal income (£8100). Of young adults who cite costs as a reason for not driving, both personal incomes and incomes earned by other household members are lower for those who cite costs alone (£4900 and £21,100, respectively), relative to the segment of young adults who indicate both cost and noncost reasons for not driving (£8100 and £26,900, respectively).
On the basis of this segmentation analysis we can therefore identify the presence of 2 distinct groups of young people who indicate that costs deter them from driving-one group that is less well-off financially and that report that costs alone are the primary deterrent and one that is somewhat better off and that indicates that other reasons also apply. Mean personal incomes are lower for all of the nondriving segments (including those who indicate that they are learning to drive) than for the fully licensed group of young adults, but the learning to drive segment has the highest mean level of income earned by other household members. The employment rate is higher among young adults with a full car driving license (81%) than any of the not fully licensed segments (values range from 35% to 56%).
Young adults in the non-cost reason(s) segment are the most likely to have born outside of the United Kingdom and also the most likely to reside in London (Britain's largest city).
Though the NTS data do not provide us with all family relationships within a household, we constructed 2 variables to serve as proxies for whether a young adult lives with either his or her own children or his or her parents. These constructed variables are, respectively, whether the oldest child in the household is 15 or more years younger than him or her and whether he or she lives with an adult at least 15 years older than him or her. It was found that members of the learning to drive segment were the least likely to live with their own child(ren) (according to this definition) and are the most likely to be living with their parent(s). The final analysis with the self-reported data about reasons for not driving brings in information on when young adults expect that they will acquire a driving license (see Table 2). Here we see that young people who indicate that the main reason they do not drive is that they are too busy to learn or cite unspecified other reasons are the most likely to expect that they will learn to drive in the next year. Young people citing costs mainly report that they expect to drive in the 5-to 10-year timescale-53%, 63%, and 67% of those citing the costs of learning to drive, insurance, and buying a car, respectively, indicate that they expect to drive within 5 to 10 years. (The "general costs of motoring" category and all others with a sample size of less than 50 are not shown in Table A2.) At the far end of the spectrum, those citing safety concerns/nervousness, a lack of interest in driving, or physical/health problems were the most likely to report that it would be more than 10 years (or never) before they learn to drive, at 37%, 39%, and 67% respectively. Only main reasons with unweighted sample sizes larger than 40 individuals are shown.

Logistic Regression Analysis
In this section we look at demographic and economic correlates of whether a person has a driving license or not, using a logistic regression model. As with any cross-sectional regression model, the results show statistical association, which cannot be interpreted as unambiguous evidence of causality.
The set of variables we tested includes spatial variables as well as personal and household characteristics (variable definitions can be found in the Data section). The estimation results are shown in Table 3.
We find that, despite young men's rate of license holding being higher than young women's, the ceteris paribus gender gap in license holding is not statistically significant (P = .12). As one gets older (within the 17-to 29-year-old age band) there is also an all-else-equal not-otherwise-explained increasing likelihood to hold a license-for every year older the all-else-equal odds of having a license increase by 14%. Educational qualifications are positively linked with license holding, as is being employed, particularly full-time employment. All of these results are consistent with the literature.
We find no significant association with the presence of one's children in the household but a negative association with living with one's parents (see definition of proxy variables in the Data section). This latter effect is opposite to that reported in Delbosc and Currie (2014). We believe that it is more intuitively plausible that the independence of living separate from one's parents is linked positively with having a driving license. Whether these different results are, however, an artefact of differences in analytical approaches or an indication of different processes at work in different social and geographic contexts must remain a matter for speculation until further empirical evidence accumulates.
The spatial associations are generally consistent with earlier work (McDonald and Trowbridge 2009). Living in London is negatively linked with license holding, but we did not find a significant relationship with living in other urban settlements in Britain versus rural areas. Greater residential density of the postcode sector where one lives (there are 10,600 postcode sectors in Britain, about one for every 6000 residents) was negatively associated with license holding, and we also found that proximity to public transport is linked with a greater likelihood of not having a driving license, with the structural implications for one's mobility that that implies. We tested for a statistical association between license holding and migration status and found being born outside of the UK to be strongly negatively linked with holding a driving license. The statistical relationship is that the odds of license holding are more than halved (−55%), net of other confounding relationships such as labor force participation, income level, or place of residence. This analysis does not indicate, however, the causal mechanisms for this finding. Possibilities in need of further enquiry include paperwork difficulties, lifestyle preferences, and the unique spatial distribution of social networks and activity patterns (cf. Chatman 2013). There is also the important question (which cannot be directly investigated with British NTS data) of the degree to which this effect is durable or may decay with the length of time one lives in the receiving country, which may well be context dependent.
We estimated separate statistical relationships between license holding and income that is earned by the young adult versus the sum of income from other household members. We found that, in keeping with intuition, one's own income has a stronger relationship with license holding (roughly 4 times as large) than does residual household income and that both of the associations are positive as we would expect. The implication is that changes in personal income versus income earned by others in the household are likely to be associated with very different changes in a young adult's likelihood of holding a license.

Conclusions
In Britain there has been a drop from the historic peak level of young adults' license holding that was seen in the early 1990s, and it now appears that in the 2000s the rate has been essentially steady-state, without a clear overall time trend.
The results presented here agree in many respects with the recent literature on the main correlates of young adults' license holding.
In Britain there is currently no graduated driving license scheme as in much of the United States, but the driving tests are difficult, and more than a quarter (28%) of unlicensed people under age 30 indicate that they are in the process of learning to drive. As 45% of British adults under age 30 do not have a driving license, this implies that young British adults are spending a mean of 1.7 years learning to drive after they turn 17 (28% * 45% * 13 years of age between age 17 and age 29).
In total, about half (51%) of unlicensed young people report that they are learning to drive or are put off by the cost of learning, about 3 in 10 (29%) seem to indicate that they are disinterested in driving, and the remaining fifth cite a wide set of main reasons for not driving.
This study reports several results that are made possible by applying novel methods and data to the issue of young adults' licensing. We show that the distinction between main reasons for not driving and secondary reasons appears meaningful and that earlier studies that do not take this into account may miss important patterns. For instance, over 90% of those citing their health for not driving report that that is the main reason they do not drive, whereas only 13% of those citing the costs of buying a car indicate the same.
We show with a segmentation analysis that there are distinct subgroups of young nondrivers, even among those who cite the costs of driving as a deterrent to driving. The members of one of the segments we identified, for instance, cite costs as well as other non-cost reasons for not driving and are relatively welloff financially and, by contrast, members of a second segment have lower mean income levels and only cite costs as the reason they do not drive. An important issue for the future research agenda is to more fully establish the relationship between age and motivations for not driving; there is now for the first time a growing cohort of British adults in their late 20s and early 30s who do not hold a full driving license (Le Vine and Jones 2012).
On the basis of multivariate logistic regression analysis, we find that being an international migrant to Britain is negatively associated with holding a driving license, an effect that is net of other confounding effects such as income and place of residence. We suggest a number of plausible reasons why this might be, but further research will be needed to disentangle them. We also found that living with one's parents seems to be associated with a lower propensity to hold a license, a finding requiring confirmation because it is different than previous results on this question. Finally, we report the new finding that different types of income relate to license holding in distinct ways, with one's own income being about 4 times as salient in this respect as the income of others in one's household. This has important implications for future rates of license acquisition by young people in Britain, where income growth in the 2000s has been principally among adults over age 30, with young women's incomes stagnating and young men's falling in real terms.