The impact of bicycle theft on ridership behavior

Abstract

Cities worldwide are promoting bicycling as a sustainable mode of transportation. However, bicycle theft remains a significant deterrent for potential riders, and also influences the behaviors of existing cyclists. Understanding the impact of theft on bicycling behaviors provides a foundation for developing strategies to address the negative impacts of bicycle theft. Our goal is to characterize if and how bicycle theft changes individual bicycling behavior. We gathered responses from 1821 individuals in a survey focused on bicycle theft in North America. We employed bivariate analysis and binary logistic regression models to explore the relationships between demographic factors, bicycle attributes, and pre-theft behavior to explain post-theft bicycling behavior. The results show that 45% of survey respondents reduced or ceased bicycling post-theft, while 6% increased their bicycling. Additionally, 40% transitioned from bicycling to unsustainable modes of transportation for their post-theft trips. Also, 69% of people eventually replaced their stolen bicycles, of which 46% selected models of equal/higher value. Pre-theft bicycling activity emerged as the most influential factor on ridership behavior after a bicycle theft, with occasional riders experiencing the most negative impact, compared to frequent riders, who remained committed to bicycling. Recovery of the stolen bicycles, e-bicycle usage, number of bicycles owned, and income levels were also predictors of future bicycling patterns. The insights from this research can inform targeted interventions for populations most at risk to reduce the negative impact of bicycle theft, such as secure parking for new and low-income bicyclists.

Keywords:

• Bicycle theft
• bicycling
• logistic regression
• sustainable transportation

Introduction

In pursuit of sustainable and environmentally conscious transportation solutions, cities worldwide are supporting bicycling as a mode of transportation. Research has highlighted that concern about bicycle theft serves as a barrier to ridership (Bachand-Marleau et al., 2011; Li et al., 2019; Sidebottom et al., 2009), both deterring potential cyclists (Ledsham et al., 2023; Mburu & Helbich, 2016; Poulos et al., 2012; Winters et al., 2011) and creating barriers for existing riders (Kang et al., 2019; Ledsham et al., 2023). Bicyclists express concerns over the safety of their bicycles when parked for extended periods in public areas like workplaces or public transport stations (Aldred & Jungnickel, 2013). Contributing to this concern is the prevailing perception that bicycle thefts are not adequately addressed by law enforcement agencies (Johnson et al., 2008; Van Lierop et al., 2015) combined with lack of secure and locking facilities in urban regions and transit hubs (Appleyard, 2012; Arbis et al., 2016; Buck & Nurse, 2021). Consequently, concerns over theft may deter bicycling for transportation, especially when it requires leaving bicycles at train or bus stations (Bonham & Koth, 2010) where secure parking is limited (Appleyard, 2012; Arbis et al., 2016; Buck & Nurse, 2021). Furthermore, while e-bicycles are a promising solution for promoting bicycle usage (Jessiman et al., 2023), they are more susceptible to theft due to their perceived higher value, thus amplifying concerns among e-bicycle cyclists (Edge et al., 2018).

Studies highlight the critical role of urban design and crime prevention in mitigating bicycle theft and encouraging cycling. Implementing secure bike storage, increasing surveillance (Wittebrood & Nieuwbeerta, 2000), along with fostering police engagement (Crowe et al., 2013), and employing effective community policing (Zhang et al., 2007) have successfully lowered theft rates. Additionally, targeted security measures in transit hubs, based on routine activities and Crime Pattern theories, have proven beneficial (Levy et al., 2018). Time-targeted policies are also advocated, following findings that specific times and days significantly contribute to theft incidents (Salvanelli, 2019). These studies collectively indicate that thoughtful urban design combined with proactive crime prevention can substantially diminish bicycle theft, thus aiding sustainable transportation initiatives. To effectively address the persistent challenge of bicycle theft, we suggest augmenting the insights from these studies with an understanding of how theft impacts individual cycling behaviors.

Previous research has established a connection between bicycle theft and subsequent bicycling behaviors. A study conducted in Montreal, Canada, reported that 93% of individuals replaced or recovered their bicycles after theft. Among them, 24% reported increased bicycling, 15% cycled less, and 60% reported no change (Van Lierop et al., 2015). Another finding revealed that 25% of individuals who fell victim to bicycle theft reported decreased bicycling activity, while 7% discontinued bicycling altogether (Carl Ellis, 2023). However, these studies did not quantify the predictors affecting changes in post-theft bicycling behaviors, which makes it difficult to shape policy interventions that aim to support people most at risk to reduce bicycling when their bicycle is stolen.

Some studies explore the individual impacts of bicycle theft on post-theft behavior, with a focus primarily on the psychological aspects. Research shows that fear of bicycle theft is a significant barrier to cycling in urban settings (Márquez & Soto, 2021) especially in lower-income households due to the inability to easily replace stolen bicycles or afford high-quality locks (Ledsham et al., 2023). Cyclists perceiving a high risk of theft are less inclined to use or invest in bicycles, and more inclined to choose public bicycle for rail transit access (Ji et al., 2017). The fear of recurrent thefts, especially in neighborhoods with high crime rates and poor guardianship, might deter residents from purchasing or using bicycles, impacting their mobility and lifestyle choices (Zhang et al., 2007). Many are reluctant to invest in high-quality bicycles due to the risk of loss, and the threat of theft also discourages them from using bicycles to travel to certain destinations (Crowe et al., 2013). Conversely, positive perceptions of safety and infrastructure significantly boost cycling participation (Blitz, 2021). In terms of reporting behavior for bicycle theft, older individuals and more serious crimes are more likely to be reported, while public incidents are less frequently reported (Dai & Gao, 2021). However, this research primarily examines only one aspect and does not offer a comprehensive analysis.

Our partnership with BikeIndex.org, a nonprofit bicycle registration group, enables a thorough investigation of post-theft bicycling behaviors. BikeIndex.org is a nonprofit bicycling registration group that has recovered 13,011 bicycles worth a total of over $23 million since 2013 (Bikeindex.org, 2023). Bicycle Index.org, besides aiding in bicycle recovery through registration, has a unique mechanism to engage with individuals who have encountered bicycle theft. Through efforts to recover stolen bicycles, BikeIndex has a database of people who have experienced theft that we were able to contact to answer a survey on the impact of theft on their bicycling.

Our research aims to explore the consequences of theft on individual bicycling behaviors. To meet this goal, we first characterized how theft impacts future bicycling for different groups of people. Second, we characterized how people replace trips that would have been taken by bicycle, immediately after the bicycle theft occurs. Third, we characterized patterns in who replaces their bicycle and what type of replacement bicycle they obtain. Understanding the implications of theft on individual bicycling behaviors, coupled with knowledge of the broader dynamics of bicycle theft highlighted in prior research, can offer a basis for devising comprehensive strategies to tackle the enduring challenge of bicycle theft.

Methods

Data collection

We conducted an online survey in the spring of 2022 that was distributed across North America (United States and Canada) through several mechanisms. By partnering with Bicycle Index, we were able to survey 1821 people who experienced bicycle theft. Bicycle Index sent the survey by email to individuals who had reported their bicycle stolen. We also shared the survey through academic and advocacy email lists, and posted on social media (Instagram, Facebook, and Twitter). To encourage broad participation in the survey, social media posts (Facebook, Instagram, and Messenger) were supported by paid ads (between May 31 to June 7, 2022) in California zip codes [Carson (90,746), Inglewood (90,305), Ladera Heights (90,056), Los Angeles (90,043), Los Angeles (90,008), Los Angeles (90,047), Oakland (94,605), Santa Maria (93,458), 1200 Cheadle Hall and Santa Barbara (+30 mi)]. We chose these zip codes because they were areas with high racial diversity, and we wanted to increase survey participation from people of color. Although our study includes participants from the United States and Canada, we targeted recruitment in California as the research team is based in California and we anticipated a higher participation rate in the state based on the research team’s access to existing networks of bicycle stakeholders.

The survey questions were designed to measure variables from four primary categories: (1) Demographic and socioeconomic information—including participant’s age, gender, country of residence, ethnic origin, household income, and the number of bicycles owned; (2) Bicycle details—features of the stolen bicycle, such as its value, whether it is electric, and its insurance, registration, and recovery status; (3) Pre-theft bicycling behavior—participant’s bicycling behavior before the theft, delving into aspects like bicycling frequency by season and the purposes behind their bicycle usage; (4) Post-theft behavior—investigating changes in bicycling activity, the adoption of alternative transportation modes, the inclination to opt for bicycle replacement, and the value of the selected bicycle. The responses to the questions in the first category function as control variables, while those in the next two categories serve as independent variables. We investigate their influence on post-theft bicycling behavior, which is assessed through the responses to the questions in category 4, acting as the dependent variable in our models.

The demographics of participants was as follows: 61% identified as male, 36% as female, and 1% as nonbinary. Most participants (74%) reported European ethnic ancestry, leading to the exclusion of this variable from our analysis due to limited diversity in ethnic origin. The median annual income for participants fell within the range of $100,000 to $149,999, with an average age of 41 years. Less than 1% of the respondents were “non-binary” or under the age of 18 limiting analysis of these groups.

Data analysis

Pre-theft bicycling activity classification

To include pre-theft bicycling activity patterns as a variable in our analysis, we had to transform survey responses on bicycling frequency into clusters that represent distinct bicycling activity patterns throughout the year. In our survey, participants reported their bicycling frequency by season (Table 1). To create a comprehensive measure of annual riding, we utilized the Agglomerative Clustering method, a well-established technique for identifying patterns within a dataset (Müllner, 2011). Agglomerative Clustering was our preferred choice due to its non-parametric nature, as it does not assume a specific data distribution, and its capacity to bypass the requirement for an initial guess regarding cluster centroids, thereby mitigating sensitivity to initialization challenges. Moreover, typical drawbacks of Agglomerative Clustering, such as performance issues with high-dimensional or large datasets, do not apply here, given the relatively small size of our dataset.

Table 1. Displays the bicycling frequency options provided to participants, who were asked to select one option for each season.

Bicycling frequency options	Value
4 or more days a week	4
1–3 days a week	3
1–3 days a month	2
Less than once a month	1
Never	0

The corresponding values from their selections were used for classification.

The classification of season bicycling frequency resulted in an index that differentiate three types of bicycling routines (Figure 1): Occasional bicycling, Seasonal bicycling, and Frequent bicycling. Occasional bicyclists (12%) demonstrated a negligible bicycling frequency (between never and less than once a month), particularly during the winter. Seasonal bicyclists, consisting of 61% of the participants, showed that during the summer, cyclists ride most frequently (∼1–3 days a week). However, as the weather became colder in the fall, the frequency of bicycling gradually decreased (between less than a month to 1–3 times a month). When the weather became warmer in the spring, there was an increase in ridership. Frequent bicyclists (27%) displayed a consistent bicycling frequency of 4 or more days a week, particularly during the spring season.

Figure 1. Illustrates the classification results of bicycling frequency among participants during different seasons. Each class is denoted by its mean value (depicted as a point) and standard deviation (represented as a vertical line) for each season. The exact values of means and standard deviation are provided in the linked table.

Statistical analysis

In our study, we conducted four Binary Logistic Regression (BLR) analyses paired with bivariate analysis to examine the relationship between the independent variables (i.e. explanatory factors) and the post-theft bicycling behavior of bicycle owners. In our research, we employed bivariate analysis by using Cross Tabulation, which presents the frequencies or counts of observations distributed among the given two variables. While bivariate analysis offers a thorough examination of correlations, it may fall short of revealing potential interactions or side effects among all the variables under scrutiny. To overcome this limitation, we incorporated regression analysis, specifically BLR, to conduct a more comprehensive evaluation of the relationships among our variables.

The choice of BLR was driven by the binary nature of our dependent variables and the characteristics of our independent variables, whether ordinal or dummy, which will be explored further in this section. The BLR yields a coefficient for each predictor, thereby signifying both the direction and strength of its influence on the likelihood of the outcome occurring. The coefficients are further transformed into Odds Ratios (OR), providing a more interpretable measure of the influence. An OR >1 indicates a positive correlation between the predictor variable and the outcome. As the predictor variable increases, the odds of the event also increase, demonstrating a positive association. Conversely, when the OR is <1, it indicates that the predictor variable has a negative impact on the outcome.

Several pre-processing steps were taken to apply our analytical process. First, categorical variables were transformed into ordinal (e.g. <$250 = 0, $250–$499 = 1… $7000 or more = 6) or dummy values (0 or 1). Second, the independence of the predictor variables was evaluated using Spearman’s rank correlation. Third, BLR cannot accommodate missing data, thus any participant responses with incomplete information or those who responded with “do not know/want to respond” regarding the research variables were excluded from the BLR analysis. Nonetheless, these responses were included in the bivariate analysis, given that bivariate analysis mandates completeness only for the two specific variables under consideration. In the following sections, we provide a more detailed explanation of how our analysis was conducted for each dependent variable related to post-theft bicycling behavior.

Post-theft bicycling activity

To quantify the influence of theft on future bicycling behavior, we asked survey participants to specify their post-theft bicycling activities. Then, we categorized their responses into two distinct groups: (1) Negative Impact—this category includes participants who reported a decrease in bicycling activity or had completely ceased bicycling; (2) No Negative Impact—this category comprises participants who stated that their post-theft bicycling activity remained about the same or had increased. In the context of post-theft bicycling activity, we utilized a bivariate analysis not only to establish an in-depth exploration but also to explore more nuanced patterns regarding the smaller groups of those who either completely discontinued their bicycling activity or increased it, which may have been obscured in the BLR analysis.

Alternative transportation mode

To understand how theft impacted transportation mode choice, participants were asked about mode alternative being used after the theft: We categorized the responses into three distinct groups; (1) Not Sustainable Transportation (NST)—participants in this group opted for transportation modes that are not environmentally friendly, contribute to carbon emissions, and rely on nonrenewable resources, such as motorized vehicles; (2) Sustainable Transportation (ST)—participants who chose transportation modes with minimal environmental impacts, such as walking, bicycling, or public transportation; (3) Didn’t Make Those Trips—a smaller subset of participants found themselves in this category, indicating that they had entirely ceased to engage in these specific journeys as a direct consequence of the theft incidents. Given the limited number of participants in Group: “Didn’t Make Those Trips,” the focus of the BLR analysis was directed toward the first two groups. However, the third group is still incorporated in the bivariate analysis, in case any noteworthy insights or patterns emerge concerning this group.

Replacement choices: what bicycles owners opted for after theft

To assess the impact of theft on replacement decisions, participants were asked about their subsequent choices: whether they replaced their bicycles and the value of the new purchase (i.e. if it was a cheaper, or equivalent/more expensive model). We employed two BLR analyses for this purpose. Initially, we analyzed the association between the variables and the bicycle owners’ decisions to replace their stolen bicycles. Subsequently, for those participants who chose to replace, we further examined the determinants influencing the relative value of their new bicycle in comparison to the stolen one.

Results

Descriptive results

The majority of participants experienced changes in their bicycling behavior following a theft incident. Table 2 shows that 51% of the surveyed participants reported a change in their bicycling activity after encountering a theft incident. Of the 51% who experienced change, 15% discontinued their bicycling activities altogether, 30% showed a diminished activity in bicycling, while 6% reported an increased activity post-theft. Following a bicycle theft, 40% of survey participants replaced their bicycle trips with non-sustainable transportation alternatives, while 46% used sustainable alternatives. A smaller subgroup, constituting 13% of the participants, ceased the journeys previously undertaken with their bicycles. The majority (69%) of participants replaced their stolen bicycles, with 46% opting for equal or higher-value models and 23% choosing lower-value replacements.

Table 2. Presents the post-theft behavior variables, the available response options (categorical values), participant counts (along with percentages), and the aggregation of the categorical values into the BLR values.

Variables	Categorical values	Count	Pct.	BLR values
Post activity	I stopped bicycling	263	15%	0 (Negative Impact)
	Less often	511	30%
	About the same	852	49%	1 (No Negative Impact)
	More often	104	6%
Mode alternative	Car, as a driver	611	35%	0 (NST)
	Car, as a passenger	57	3%
	Taxi/ride-hailing service	13	1%
	Motorcycle or scooter	14	1%
	Cycle, personal bicycle	291	17%	1 (ST)
	Walk	282	16%
	Transit	163	9%
	Cycle, public bicycle share	27	2%
	Cycle, rental bicycle	11	1%
	Micro mobility	9	1%
	Didn’t make those trips	221	13%
Is replaced	No	558	31%	0 (No)
	Yes, lower	426	23%	1 (Yes)	0 (Lower)
	Yes, equal, or higher	828	46%		1 (Equal/higher)

Post-theft bicycling activity

Pre-theft bicycling activity and the recovery status of stolen bicycles demonstrated a substantial influence on post-theft bicycling activity (Table 3), with a pronounced decline observed among occasional bicyclists and those who were unable to recover their stolen bicycles. Regarding pre-theft bicycling activity, Table 4 shows that 83% among occasional bicyclists indicated negative shifts in their bicycling activity, with 59% of this subset opting out of bicycling altogether. The negative impact decreased with increased bicycling activity: 46% among seasonal bicyclists and 23% from frequent bicyclists reported reductions. Furthermore, only 4% of frequent bicyclists completely ceased bicycling. Concerning the recovery of stolen bicycles, among participants who were unable to retrieve their stolen bicycles, 31% reported a diminished bicycling activity, while 17% ceased bicycling entirely. Conversely, of those who successfully recovered their bicycles, 22% reduced their bicycling activity, and a mere 6% opted to halt bicycling following recovery.

Table 3. Offers BLR results, sorted by or values, pertaining to the “post activity” categories-negative and no negative impact.

Post activity (n = 1244) 0: Negative impact (“I stopped bicycling” + “Less often”) 1: No Negative impact (“About the same” + “More often”)
Variables	Coeff	[0.025	0.975]	p-Value	OR
Pre-theft activity	0.42	0.19	0.65	0.00	1.52
Was recovered	0.40	0.04	0.77	0.03	1.50
Gender	0.18	−0.08	0.44	0.18	1.19
Country	0.12	−0.20	0.45	0.46	1.13
Age groups	0.09	−0.11	0.28	0.37	1.09
Value approx.	0.05	−0.06	0.15	0.38	1.05
Number of bicycles	0.02	−0.09	0.12	0.71	1.02
Household income	−0.01	−0.08	0.05	0.73	0.99
Was insured	−0.04	−0.32	0.25	0.81	0.97
Was registered	−0.05	−0.31	0.22	0.73	0.95
Purpose	−0.09	−0.25	0.08	0.30	0.92
Was electric	−0.22	−0.46	0.01	0.06	0.80

The table presents coefficients alongside their corresponding confidence intervals, associated p-values, and or. The value “n” denotes the number of responses included in this analytical examination.

Table 4. The findings from the bivariate analysis regarding post-activity and the independent variables.

Variables	Dummy/Ordinal values	Negative impact		No Negative impact
		I stopped bicycling	Less often	About the same	More often
Pre-theft activity	Occasional bicycling	120 (59%)	49 (24%)	32 (16%)	4 (2%)
	Seasonal bicycling	108 (11%)	356 (35%)	477 (47%)	68 (7%)
	Frequent bicycling	16 (4%)	82 (19%)	318 (72%)	27 (6%)
Was recovered	No	248 (17%)	450 (31%)	673 (46%)	87 (6%)
	Yes	15 (6%)	60 (22%)	179 (66%)	17 (6%)
Number of bicycles	Zero	166 (79%)	32 (15%)	9 (4%)	3 (1%)
	One	63 (10%)	235 (37%)	295 (46%)	45 (7%)
	Two	18 (4%)	133 (33%)	217 (54%)	34 (8%)
	Three	6 (3%)	53 (27%)	132 (67%)	7 (4%)
	Four	3 (3%)	20 (19%)	78 (73%)	6 (6%)
	Five or more	0 (0%)	28 (20%)	103 (74%)	8 (6%)
Value approx.	<$250	8 (11%)	20 (27%)	36 (49%)	10 (14%)
	$250–$499	33 (14%)	75 (33%)	105 (46%)	15 (7%)
	$500–$999	115 (23%)	141 (28%)	214 (43%)	26 (5%)
	$1000–$1999	48 (12%)	130 (32%)	203 (50%)	23 (6%)
	$2000–$3999	33 (12%)	85 (30%)	152 (53%)	16 (6%)
	$4000–$6999	20 (14%)	34 (24%)	77 (55%)	10 (7%)
	$7000 or more	3 (3%)	25 (27%)	62 (66%)	4 (4%)
Was registered	No	172 (16%)	318 (30%)	494 (46%)	80 (8%)
	Yes	76 (13%)	176 (28%)	335 (55%)	20 (3%)
Was insured	No	176 (17%)	309 (30%)	483 (47%)	64 (6%)
	Yes	71 (12%)	176 (28%)	327 (54%)	35 (6%)
Was electric	No	225 (15%)	434 (28%)	736 (50%)	90 (6%)
	Yes, with pedal-assist	11 (12%)	23 (26%)	51 (56%)	4 (4%)
	Yes, with pedal-assist and throttle	19 (17%)	40 (35%)	47 (41%)	9 (8%)
Purpose	Recreation or exercise	114 (16%)	202 (28%)	353 (49%)	49 (7%)
	Equal mix	75 (17%)	148 (33%)	203 (45%)	27 (6%)
	Transportation	64 (12%)	157 (28%)	292 (54%)	28 (5%)
Age groups	Young adults (18–24)	30 (32%)	18 (19%)	36 (39%)	9 (10%)
	Adults (25–40)	131 (18%)	232 (31%)	334 (45%)	50 (7%)
	Adults (41–64)	79 (12%)	199 (30%)	355 (53%)	35 (5%)
	Older adults (>64)	12 (9%)	21 (16%)	86 (67%)	9 (7%)
Gender	Woman	115 (19%)	194 (32%)	263 (43%)	34 (6%)
	Man	130 (12%)	295 (28%)	553 (53%)	62 (6%)
Income	<$20,000	17 (21%)	29 (35%)	26 (32%)	10 (12%)
	$20,000–$34,999	29 (26%)	30 (27%)	44 (40%)	7 (6%)
	$35,000–$49,999	21 (23%)	23 (25%)	40 (43%)	8 (9%)
	$50,000–$74,999	37 (17%)	66 (30%)	99 (46%)	15 (7%)
	$75,000–$99,999 per year	39 (17%)	64 (28%)	116 (50%)	11 (5%)
	$100,000–$149,999 per year	42 (16%)	80 (30%)	125 (47%)	19 (7%)
	$150,000–$199,999 per year	14 (8%)	55 (32%)	97 (56%)	7 (4%)
	>$200,000	30 (8%)	106 (30%)	200 (56%)	19 (5%)
Country	CAN	47 (15%)	95 (30%)	160 (50%)	15 (5%)
	USA	216 (15%)	416 (28%)	692 (49%)	89 (6%)

The post-bicycling activity of individuals who own a larger number of bicycles and those whose stolen bicycle had either low or high value may be impacted by theft, even though these correlations may not necessarily follow linear patterns (Table 4), as it is also indicated by their OR, which are close to 1 (Table 3). Participants without additional bicycles experienced the most severe consequences, with 79% discontinuing bicycling altogether, and 15% reducing their bicycling activity. Moreover, 47% of single bicycle owners exhibited a reduction in bicycling activity, with 37% bicycling less and 10% stopping entirely. The decline in future ridership lessened to just 20% for individuals owning five or more bicycles, and none from this group discontinued bicycling. In cases where bicycle value is considered, bicycle with an estimated value of up to $1000 show an increase in negative effects, affecting 38% to 51% of individuals who either reduce their bicycling activity or cease it altogether. However, this trend reverses as the value of the bicycles exceeds the $1000 threshold, decreasing to 30% among those with bicycles valued at more than $7000. Additionally, among owners of bicycles valued under $250, a higher percentage (14%) reported an increase in bicycling activity after theft incidents, in contrast to other value groups where this percentage stood at 6%.

Alternative transportation mode

A higher pre-theft bicycling activity was associated with a greater inclination to continue using ST modes post-theft (Table 5). Table 6 indicates that 64% of frequent cyclists expressed a preference for ST, compared to 39% for seasonal cyclists and 38% for occasional cyclists. Furthermore, only 8% of frequent cyclists chose to discontinue their journeys entirely, in contrast to the 15% cessation rate observed in other cyclist groups.

Table 5. Presents the results obtained from the BLR analysis focused on the categories of “mode alternative”—NST and ST.

Mode alternative (n = 1063) 0: NST 1: ST
Variables	Coeff	[0.025	0.975]	p-Value	OR
Pre-theft activity	0.67	0.44	0.90	0.00	1.96
Was recovered	0.44	0.09	0.79	0.01	1.55
Purpose	0.23	0.07	0.40	0.01	1.26
Was registered	0.18	−0.08	0.45	0.17	1.20
Gender	0.12	−0.15	0.39	0.37	1.13
Number of bicycles	0.03	−0.08	0.13	0.59	1.03
Age groups	−0.04	−0.23	0.15	0.71	0.97
Value approx.	−0.05	−0.16	0.06	0.38	0.95
Was insured	−0.08	−0.37	0.21	0.61	0.93
Country	−0.09	−0.42	0.24	0.60	0.92
Household income	−0.09	−0.16	−0.03	0.01	0.91
Was electric	−0.34	−0.59	−0.10	0.01	0.71

Table 6. The findings from the bivariate analysis regarding mode alternative and the independent variables.

Variables	Dummy/ordinal values	Didn’t make those trips	NST	ST
Pre-theft activity	Occasional bicycling	29 (14%)	96 (48%)	75 (38%)
	Seasonal bicycling	151 (15%)	453 (46%)	391 (39%)
	Frequent bicycling	34 (8%)	124 (28%)	280 (64%)
Purpose	Recreation or exercise	187 (27%)	274 (39%)	237 (34%)
	Equal mix	19 (4%)	202 (45%)	231 (51%)
	Transportation	13 (2%)	215 (40%)	312 (57%)
Income	<$20,000	3 (4%)	25 (30%)	54 (66%)
	$20,000–$34,999	17 (16%)	33 (31%)	57 (53%)
	$35,000–$49,999	13 (14%)	34 (37%)	45 (49%)
	$50,000–$74,999	17 (8%)	92 (43%)	106 (49%)
	$75,000–$99,999 per year	26 (11%)	100 (44%)	102 (45%)
	$100,000–$149,999 per year	32 (12%)	123 (47%)	105 (40%)
	$150,000–$199,999 per year	24 (14%)	72 (42%)	75 (44%)
	>$200,000	60 (17%)	137 (40%)	147 (43%)
Was recovered	No	193 (13%)	605 (42%)	638 (44%)
	Yes	28 (11%)	90 (34%)	144 (55%)
Was electric	No	203 (14%)	575 (39%)	679 (47%)
	Yes, with pedal-assist	5 (6%)	42 (48%)	41 (47%)
	Yes, with pedal-assist and throttle	9 (8%)	66 (57%)	39 (34%)
Was registered	No	156 (15%)	437 (42%)	449 (43%)
	Yes	62 (10%)	228 (38%)	309 (52%)
Was insured	No	116 (11%)	406 (40%)	490 (48%)
	Yes	98 (16%)	254 (42%)	247 (41%)
Value approx.	<$250	12 (16%)	17 (23%)	44 (60%)
	$250–$499	15 (7%)	90 (40%)	120 (53%)
	$500–$999	51 (10%)	203 (41%)	237 (48%)
	$1000–$1999	43 (11%)	167 (42%)	185 (47%)
	$2000–$3999	55 (20%)	118 (42%)	107 (38%)
	$4000–$6999	29 (21%)	58 (43%)	49 (36%)
	$7000 or more	14 (15%)	40 (43%)	39 (42%)
Age groups	Young adults (18–24)	3 (3%)	25 (27%)	64 (70%)
	Adults (25–40)	96 (13%)	306 (42%)	334 (45%)
	Adults (41–64)	103 (16%)	270 (41%)	281 (43%)
	Older adults (>64)	7 (5%)	55 (43%)	66 (52%)
Gender	Woman	81 (14%)	248 (42%)	265 (45%)
	Man	130 (13%)	411 (40%)	482 (47%)
Number of bicycles	Zero	16 (8%)	104 (51%)	85 (41%)
	One	87 (14%)	242 (39%)	299 (48%)
	Two	54 (14%)	167 (42%)	174 (44%)
	Three	26 (13%)	71 (37%)	96 (50%)
	Four	15 (14%)	44 (42%)	47 (44%)
	Five or more	14 (10%)	54 (39%)	70 (51%)
Country	CAN	35 (11%)	124 (39%)	156 (50%)
	USA	186 (13%)	571 (41%)	627 (45%)

Other groups of people who were inclined to maintain their commitment to ST mode included individuals who used their bicycles for transportation purposes, those with an annual income of <$35,000, and those who successfully recovered their stolen bicycles (Tables 5 and 6). Among respondents who exclusively utilized their stolen bicycles for transportation, 57% continued to prioritize ST. This percentage decreased to 51% for those who used their stolen bicycles for dual purposes, serving both recreational and transportation needs, and dropped further to 34% among those who primarily used the bicycle for recreational or exercise purposes. Regarding income, participants with annual incomes below $20,000 demonstrated the highest level of ongoing usage of ST at 66%, whereas this rate decreased to an average of 42% among those with incomes exceeding $100,000. Regarding the recovery status of stolen bicycles, among bicycle owners who successfully recovered their stolen items, 55% remained committed to ST before any potential recovery or reacquisition. In contrast, 44% displayed the same inclination among those who did not eventually recover their bicycles.

E-bicycle owners were most likely to switch travel modes to NST modes (OR = 0.71) of transportation when they did not have access to a bicycle. Among conventional bicycle users, 39% express a preference for NST modes. In contrast, among e-bicycle owners, 48% of those with pedal-assist systems lean toward NST modes, and 57% of those whose e-bikes feature both pedal-assist and throttle mechanisms.

Replacement choices: what bicycles owners opted for after theft

In Tables 7 and 8, we show that there are three groups of people who have a reduced likelihood of replacing stolen bicycles: individuals who managed to recover their stolen bicycle, occasional cyclists, and respondents who reported not owning a bicycle at present. Among those who retrieved their stolen bicycles, 56% opted not to buy new ones, compared to 26% among those unsuccessful in recovery. It is noteworthy that our dataset does not specify whether the replacement occurred before or after the bicycle was recovered. For occasional cyclists, 69% chose not to replace their bicycles, a stark contrast to the 27% of seasonal cyclists and 22% of frequent cyclists who made the same decision. Regarding the number of bicycles owned, a vast majority (92%) of respondents without any current bicycle did not seek a replacement for their lost item. No further patterns were observed across the other number of bicycle groups.

Table 7. Presents the outcomes derived from the BLR analysis that centers on the decision of whether the stolen bicycle was replaced or not.

Is replaced (n = 1247) 0: No 1: Yes
Variables	Coeff	[0.025	0.975]	p-Value	OR
Pre-theft activity	0.71	0.46	0.96	0.00	2.03
Number of bicycles	0.55	0.41	0.69	0.00	1.73
Was insured	0.42	0.10	0.73	0.01	1.52
Purpose	0.13	−0.05	0.31	0.15	1.14
Household income	0.05	−0.02	0.12	0.14	1.05
Was registered	0.03	−0.27	0.32	0.87	1.03
Age groups	−0.01	−0.22	0.20	0.92	0.99
Value approx.	−0.03	−0.14	0.09	0.64	0.97
Country	−0.06	−0.41	0.29	0.74	0.94
Was electric	−0.18	−0.45	0.08	0.18	0.83
Gender	−0.38	−0.67	−0.09	0.01	0.68
Was recovered	−1.91	−2.28	−1.54	0.00	0.15

Table 8. The findings from the bivariate analysis regarding the predictor variables and the decision to replace the stolen bicycle and whether the replacements had a lower or equal/higher value.

Variables	Dummy/ordinal values	No	Yes, lower	Yes, equal, or higher
Was recovered	No	399 (26%)	397 (26%)	739 (48%)
	Yes	158 (56%)	29 (11%)	89 (32%)
Pre-theft activity	Occasional bicycling	141 (69%)	27 (13%)	37 (18%)
	Seasonal bicycling	271 (27%)	249 (25%)	493 (49%)
	Frequent bicycling	96 (22%)	116 (26%)	230 (52%)
Number of bicycles	Zero	196 (93%)	8 (4%)	7 (3%)
	One	145 (23%)	188 (28%)	309 (48%)
	Two	82 (20%)	108 (27%)	214 (53%)
	Three	46 (23%)	53 (27%)	100 (50%)
	Four	22 (20%)	22 (20%)	64 (59%)
	Five or more	28 (20%)	27 (19%)	84 (60%)
Was registered	No	356 (32%)	250 (22%)	515 (46%)
	Yes	183 (28%)	156 (25%)	289 (46%)
Was insured	No	375 (35%)	276 (26%)	428 (40%)
	Yes	152 (24%)	128 (20%)	359 (56%)
Value approx.	<$250	25 (32%)	20 (26%)	32 (42%)
	$250–$499	88 (36%)	68 (28%)	88 (36%)
	$500–$999	165 (32%)	138 (27%)	217 (42%)
	$1000–$1999	131 (31%)	101 (24%)	194 (46%)
	$2000–$3999	90 (30%)	55 (19%)	151 (51%)
	$4000–$6999	36 (25%)	24 (16%)	86 (59%)
	$7000 or more	21 (22%)	18 (19%)	57 (59%)
Age groups	Young adults (18–24)	36 (38%)	37 (39%)	21 (22%)
	Adults (25–40)	255 (34%)	172 (23%)	322 (43%)
	Adults (41–64)	177 (26%)	156 (23%)	338 (50%)
	Older adults (>64)	35 (27%)	22 (17%)	72 (56%)
Gender	Woman	190 (31%)	153 (25%)	266 (44%)
	Man	311 (30%)	233 (22%)	502 (48%)
Income	<$20,000	29 (35%)	37 (45%)	17 (20%)
	$20,000–$34,999	40 (36%)	39 (35%)	31 (28%)
	$35,000–$49,999	34 (37%)	22 (24%)	36 (39%)
	$50,000–$74,999	72 (33%)	61 (28%)	85 (39%)
	$75,000–$99,999 per year	75 (32%)	47 (20%)	109 (47%)
	$100,000–$149,999 per year	86 (32%)	54 (20%)	128 (48%)
	$150,000–$199,999 per year	48 (28%)	37 (21%)	89 (51%)
	>$200,000	81 (23%)	71 (20%)	205 (56%)
Purpose	Recreation or exercise	230 (32%)	131 (18%)	359 (50%)
	Equal mix	144 (32%)	115 (25%)	195 (43%)
	Transportation	146 (27%)	157 (28%)	237 (44%)
Was electric	No	479 (31%)	385 (25%)	689 (44%)
	Yes, with pedal-assist	24 (25%)	13 (14%)	59 (61%)
	Yes, with pedal-assist and throttle	39 (32%)	17 (14%)	64 (53%)
Country	CAN	112 (33%)	78 (23%)	146 (43%)
	USA	446 (30%)	348 (24%)	682 (46%)

Household income (OR = 1.18), purpose of use (OR = 0.82), and stolen bicycles being electric (OR = 1.42) were statistically significant determinants of replacement bicycle value (with p-value smaller than 0.05, see Table 9), with e-bicycle owners and individuals earning over $150k typically opting for models of similar or higher cost (Table 8). Among e-bicycle owners, 61% with pedal-assist systems and 53% with both pedal-assist and throttle mechanisms opted for replacements of similar or higher value, in contrast to 44% of conventional bicycle owners. Conversely, while 25% of conventional bicycle owners chose lower-value replacements, 14% of e-bicycle owners did the same. The analysis revealed a positive correlation between income levels and the choice to replace stolen bicycles with bicycles of equal or higher value. The observed percentages display a range, starting at 20% among individuals with an annual income of $20,000 or less and gradually increasing to 56% among those earning $200,000 or more per year. Conversely, for individuals with incomes below $20,000, 45% chose lower-value bicycle replacements, while among those earning $75,000 or more, only 20% made the decision to opt for lower-value bicycle replacements. Regarding purpose usage, Individuals who predominantly employed their bicycles for recreational or exercise purposes displayed a greater inclination (50%) to opt for equal/higher value replacements, compared to 44% among those who used the bicycles for transportation or a mix of usage.

Table 9. Presents the results obtained from the BLR analysis that focuses on the determination of whether the replacement bicycle should possess a lower value or an equal/higher value.

Is replaced (n = 879) 0: Replaced with cheaper model 1: Replaced with the same or more expensive one
Variables	Coeff	[0.025	0.975]	p-Value	OR
Was recovered	0.39	−0.11	0.90	0.13	1.48
Was electric	0.35	0.02	0.67	0.04	1.42
Was insured	0.25	−0.09	0.58	0.15	1.28
Pre-theft activity	0.20	−0.09	0.49	0.18	1.22
Age groups	0.17	−0.06	0.40	0.15	1.18
Household income	0.17	0.09	0.24	0.00	1.18
Country	0.16	−0.23	0.54	0.43	1.17
Number of bicycles	0.03	−0.10	0.15	0.67	1.03
Value approx.	0.01	−0.11	0.13	0.89	1.01
Gender	−0.02	−0.33	0.29	0.88	0.98
Was registered	−0.16	−0.46	0.15	0.32	0.86
Purpose	−0.20	−0.40	−0.01	0.04	0.82

Discussion

When comparing our research findings on the impact of bicycle theft on post-theft bicycling behaviors to those of other studies, we noted a combination of commonalities and discrepancies. Carl Ellis (2023) reported a 25% reduction in bicycling activity and a 7% complete cessation of bicycling, which contrasts with our figures of 30 and 15%, respectively. Despite the numerical differences, both our study and Carl Ellis (2023 (share a common trend: a larger proportion of individuals exhibit reduced bicycling activity rather than completely ceasing bicycling. Van Lierop et al. (2015) found that 93% of participants either recovered or replaced their bicycles, as opposed to our 78%. Again, while the values differ, the common trend remains: the majority either replaced or recovered their stolen bicycles. Among those who recovered or replaced their bicycles, 60% experienced no change, aligning closely with our figure of 61%. However, their results indicated that 24% increased post-bicycling activity, and 15% decreased it, in contrast to our figures of 7% increased post-bicycling activity and 31% increased post-bicycling activity. The variations may also be attributed to the unique factors considered in our research, which could account for the disparities in responses.

A person’s bicycling activity before theft was consistently found to be an important factor in post-theft bicycling behavior. Occasional bicyclists were most negatively affected by bicycle theft, 62% no longer using sustainable transportation modes, 69% forgoing bicycle replacement, and 59% ceasing bicycling altogether. We found that most of occasional bicyclists possessed bicycles for exercise or recreational purposes with a valuation under $250, underscoring the peripheral role of bicycling in their daily lives. The second class of bicyclists, seasonal bicyclists had more consistent bicycling patterns before and after a theft occurred. A larger fraction of this class (74%) replaced their stolen bicycles and 54% resumed their bicycling routines. However, a significant portion of individuals in this class (46%) also shifted toward NST modes. Frequent bicyclists, on the other hand, made few behavioral changes after theft. The marginal shifts in their bicycling routines (23%), coupled with their tendency to replace stolen bicycles (78%) and continued use of ST (64%), highlight the central role of bicycling in their daily lives. Frequent bicyclists generally possess high-end bicycles, frequently valued at more than $4000, and own multiple bicycles, which provides alternatives bicycling even after a theft and further emphasizing their deep-seated engagement with bicycling. To effectively tackle bicycle theft and related concerns, it is essential to address the barriers and motivations for each class separately (Piatkowski & Marshall, 2015). Additionally, interventions to enhance bicycling activity can draw insights from existing research. For example, studies have suggested that factors, such as eye-level greenness exposure (Wang et al., 2020) and mixed land uses, which reduce the distance between home and workplace (Damant-Sirois & El-Geneidy, 2015; Pucher & Buehler, 2008) have shown the potential to encourage more frequent bicycling.

The retrieval of a stolen bicycle emerges also as one of the pivotal determinants in shaping an owner’s post-theft bicycling behavior. Our results demonstrated that individuals who were unable to recover their stolen bicycles experienced more disruptions in their bicycling behavior; individuals who successfully retrieve their stolen bicycles typically have fewer negative changes in their bicycling activity and a less tendency to invest in a new bicycle compared to those who fail to recover theirs. Additionally, the majority within those who recover their bicycle remain steadfastly aligned with ST. However, it can be bidirectionally interpreted, suggesting that individuals dedicated to ST might be more proactive in their efforts to reclaim their stolen bicycles.

Individuals with higher incomes demonstrate a greater tendency toward replacing their stolen bicycles, often opting for replacement bicycles of the same or higher value. In contrast, individuals with lower income levels demonstrate a less likelihood of bicycle replacement. Among individuals with lower income levels who chose to replace their bicycles, there is a preference for lower-value replacements. Furthermore, the findings also reveal that among individuals with lower income levels, bicycling holds a central role in their daily lives. It is reflected in the higher percentages of consistent and high bicycling activity, as well as a greater proportion of bicycle usage for transportation purposes which align with findings from previous studies (Aldred & Jungnickel, 2014; Anantharaman, 2017). Consequently, the ramifications of theft are more problematic for people with lower incomes. Despite these challenges, a large number within this group persist in selecting ST as their mode of choice, however use of ST may be a result of necessity than by choice.

Among e-bicycle users, we found a decreased use of ST after a bicycle theft, which in some instances may be attributed to physical limitations that can be associated with some e-bicycle use (Melia & Bartle, 2021), However, an interesting dichotomy was evident within the e-bicycle user groups. While a higher percentage of e-bicycle owners are part of frequent bicyclists compared to traditional bicycle users, a notable proportion also shifted toward NST modes after theft. Drawing from this observation, the gains in bicycling from the emergence of e-bikes (Johnson et al., 2023), getting people on bicycles who had not considered conventional bikes, might be jeopardized by bicycle theft.

Demographic factors had minimal impact on post-theft bicycling behavior. Age and country of residence exhibited no correlation with post-theft behavior. In addition, prior research suggested a greater tendency among men to embrace ST (Meesit et al., 2023). However, our findings show no such distinction in their use of ST following a bicycle theft incident.

Our study examining the impact of bicycle theft on ridership behavior offers important insights but faces several limitations. Its reliance on a bicycle registration database and social media for participant recruitment may introduce sampling bias, as these methods might not represent the general population. The concentration of respondents in California could also limit the findings’ applicability to other regions, and the self-reported data may be affected by recall and social desirability biases. Demographically, The lack of diversity among respondents could further restrict the study’s relevance. Future research should address these limitations by expanding the sample size and employing more diverse data collection methods to improve the study’s validity and enhance policy-making insights. Additionally, exploring complex relationships through advanced statistical models like spline regression or tree-based models could offer a deeper understanding of the factors influencing post-theft behavior.

Conclusion

Our findings reveal that 51% of the surveyed population changed their bicycling activity following a theft incident, with 45% of them reporting a decrease in bicycling. Furthermore, 40% of respondents switched from ST for their trips post-theft and 69% eventually replaced their stolen bicycles, often (46%) opting for models of equal or higher value. Our study highlights the significant role of pre-theft bicycling activity in influencing post-theft bicycling behavior. Individuals with high and consistent bicycling activity demonstrated greater resilience against the negative impacts of bicycle theft, while those with lower bicycling activity levels are more susceptible to such impacts. Therefore, a concerted focus on initiatives that promote bicycling activity, fostering a stronger commitment to bicycling, can help mitigate the negative impact of post-theft on bicycling behaviors. The recovery of stolen bicycles also alleviates the negative consequences of theft incidents. Strategies like registering bicycles in systems, such as BikeIndex, reporting theft incidents to the police, and sharing them on social media platforms could enhance the chances of recovery. Furthermore, household income levels underscore the disparities in how theft affects individuals with varying economic backgrounds. In other words, individuals with lower incomes, who also rely more heavily on bicycles as a vital means of transportation, experience a more pronounced negative impact on their behavior. Targeted interventions for populations most at risk to reduce bicycling after a bicycle theft are needed, such as secure parking or temporal replacement bicycle service for low-income bicyclists.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by Institute of Transportation Studies, UC Davis [STRP_UCSB_2023-24].