Abstract
In the digital economy era, data have become a core production factor and a key driver of industrial transformation, with the electric vehicle market being one of the most affected sectors. This paper uses a distinctive car-human big-data platform that integrates automotive data with mobile signaling data. The analysis is based on a 2024 sample of battery electric vehicle (BEV) and plug-in hybrid electric vehicle (PHEV) users in a major Chinese municipality. I construct a dummy variable linear regression model to test how commuting distance affects the certified CLTC electric range chosen by electric vehicle users. The empirical results demonstrate that commuting distance has a statistically significant influence on electric vehicle range selection. Compared with short-distance commuters, users with medium commuting distance opt for vehicles with roughly 17.8 km more range, while long-distance commuters choose vehicles with about 39.9 km more range. This effect is almost entirely concentrated among BEV users, indicating that BEV users are more range-sensitive than PHEV users. In addition, among BEV buyers, the price segment and vehicle class also moderate the relationship: long-distance commuters who purchased low-price cars or A00/A0/A-class cars show a significantly stronger tendency to buy vehicles with longer range. These findings provide theoretical guidance for market segmentation and product positioning in the electric vehicle industry. The study also acknowledges limitations related to data precision and omitted variables, which indicate directions for future research.
Keywords: electric vehicle; new energy vehicle; commuting distance; range demand; user profiling; dummy variable linear regression
1. Introduction
In the context of the booming digital economy, data elements are profoundly changing global industry and consumer behavior. The Chinese government has successively issued the Fourteenth Five-Year Plan for Digital Economy Development and the Opinions on Constructing a Data Basic System to Better Play the Role of Data Elements, aiming to establish a complete governance system for data property rights, circulation, transactions, and security. These policies create institutional guarantees for releasing the value of data. The automobile industry, as a typical data-intensive sector, possesses rich information about user behavior, vehicle attributes, traffic conditions, and mobility patterns. Such data can be used to monitor dynamic market demand and to optimize product design in a real-time, evidence-based way. This data-driven optimization is especially prominent in the new energy vehicle industry, where electric vehicle consumers are increasingly heterogeneous in their acceptance of technology, driving style, and range requirements.
As the market penetration of new energy vehicles continues to grow in China, driving range has become one of the most important parameters in consumers’ purchase decisions. For traditional fuel vehicles, commuters mainly care about fuel costs, road conditions, and travel time. In contrast, for owners of battery electric vehicles and plug-in hybrid electric vehicles, the match between the daily commuting distance and the electric range of the vehicle is a direct determinant of usage convenience and overall utility. A BEV with a short real-world range may not satisfy the user’s daily round-trip need, especially in cold weather or with high-speed commuting. Consequently, mismatches between commuting distance and electric vehicle range can create severe range anxiety, reduce post-purchase satisfaction, and even generate negative word of mouth.
Despite the broad acknowledgment that range is a decisive factor for electric vehicle adoption, existing research has not yet established a precise quantitative link between commuting distance and the range that consumers actually select. Many industry surveys simply report stated preferences, while few studies use revealed purchase data to measure the trade-off between range and commuting behavior. Moreover, the heterogeneities between BEV and PHEV owners are seldom examined through a rigorous regression design. For instance, a young single person may care more about appearance, smart functions, and driving pleasure, while a middle-aged user with children may emphasize range, safety, and interior space. Without a systematic analysis of user profiles, manufacturers cannot tailor their product portfolios and marketing strategies to different commuter clusters.
This paper responds to that gap. Based on the car-human big-data platform jointly developed by an automotive data company and a mobile network operator, I obtain a large sample of 2024 BEV and PHEV users from a Chinese municipality. The dataset contains 11,219 observations, including 4,859 BEV records and 6,360 PHEV records, with 13 variables covering range, commute distance, vehicle class, price segment, age, gender, family structure, and living arrangements. I then construct a richer user-profile system that combines demographic attributes, household attributes, and mobility features. This profile system supports the examination of several important subgroup differences, such as the intensity of family needs, whether the drivers are elderly, whether elderly parents live in the same city, and whether pregnant women are present in the household.
To quantify the impact of commuting distance on range choice, I use a dummy variable linear regression model. The model includes dummy variables for medium and long commuting distances, with short commuting distance as the omitted reference category. This design makes the coefficients directly interpretable as the average range difference between commuter groups. To ensure robustness, I estimate both a baseline model without control variables and a full model with a broad set of control variables. The coefficients of the two specifications are very close to each other, showing that the key conclusions are not driven by observed demographic confounding. Subsequently, I extend the model with interaction terms between the commuting-distance dummies and the main grouping variables, including powertrain type, price segment, and vehicle class. These interaction terms enable me to measure how the sensitivity of range demand to commuting distance varies across market segments.
The central result is straightforward. Commuting distance significantly affects the range choice of electric vehicle users. Medium commuters, with distances between 10 and 25 km, choose vehicles with about 17.8 km more range than short commuters. Long commuters, with distances above 25 km, choose vehicles with about 39.9 km more range. When I separate the sample by powertrain, the effect is almost exclusively driven by BEV users. Among medium-distance commuters, BEV users choose about 12.3 km more range than PHEV users relative to the corresponding short-distance contrast; among long-distance commuters, the relative difference reaches about 30.0 km. For PHEV users, commuting distance has no material influence on the electric range choice, which is intuitive because PHEVs can be driven as hybrids or purely with gasoline, reducing the range anxiety faced by drivers.
I further concentrate on BEV users and explore whether price segment and vehicle class moderate the commute-range relationship. The results show that cheap electric vehicle users are more responsive to commuting distance. Long-distance commuters who bought vehicles under 200,000 RMB choose about 28.9 km more range than short-distance commuters in the same price group. In contrast, for expensive electric vehicle users above 200,000 RMB, commuting distance does not change their range choice in a statistically significant way. Similarly, when I compare A00/A0/A-class vehicles (small and medium vehicles) with B/C/D-class vehicles, only the former show a significant commuting-distance effect. This implies that product positioning modifies how strongly commuting needs are converted into range demand. The findings give automakers a useful toolkit for anticipating the range preferences of different consumer clusters and for recommending suitable vehicles to prospects.
The remainder of the paper is organized as follows. Section 2 presents the data source, variable definitions, and descriptive framework. Section 3 introduces the econometric model and identification strategy. Section 4 reports the empirical results, including the baseline estimates and the heterogeneity analyses. Section 5 discusses the implications and limitations. Section 6 concludes the paper with policy and managerial recommendations.
2. Data and Variables
2.1. Data source and sample
The data used in this paper come from a collaborative “car-human big data platform” established by an automotive data company and a major mobile network operator. The platform integrates detailed vehicle registration data, user location data, and vehicle specification data. Due to privacy restrictions, the two parties do not directly exchange raw files. Instead, they use a proprietary privacy-enhancing technology that allows computation on encrypted or federated data, ensuring that data remain “available but invisible” within each party’s domain. This feature guarantees that all analyses respect individual privacy while still preserving the richness of big data.
The analytical sample covers users of battery electric vehicles (BEVs) and plug-in hybrid electric vehicles (PHEVs) in 2024. The sample is drawn from a large municipality in China so that the commuting distance definition can be consistently constructed from residential and workplace locations. After merging the range database with the user-profile database, the total number of valid records is 11,219, of which 4,859 are BEV users and 6,360 are PHEV users. The final regression samples after removing observations with missing values in the control variables are slightly smaller, as indicated in the tables. All vehicle range values are measured as the certified CLTC electric range in kilometers; this standard metric is widely used in the Chinese new energy vehicle market and is printed on the vehicle catalogs and energy-consumption labels. For plug-in hybrids, the relevant range metric is the pure electric range under CLTC conditions, because that is the range parameter that consumers see when assessing the electric-driving capability of a PHEV.
2.2. Variable definitions
The main variables used in the empirical analysis are summarized in Table 1. The dependent variable is the certified electric range in kilometers. The core explanatory variables are two commuting-distance dummies: Commute_medium equals 1 if the commuting distance is between 10 and 25 km, and 0 otherwise; Commute_high equals 1 if the commuting distance is more than 25 km, and 0 otherwise. The reference category is thus a short commuting distance of less than 10 km. Such a coding scheme avoids assuming linearity in commuting distance and allows me to estimate a flexible relationship.
| Variable | Type | Meaning | Sample values / categories |
|---|---|---|---|
| Range | Numeric | Certified CLTC electric range of the vehicle (km) | 54.6, 93.8, 350.0, etc. |
| Commuting distance | Categorical | Workday straight-line distance between residence and workplace | 0–2 km, 2–5 km, 5–10 km, 10–25 km, >25 km |
| BEV | Binary | 1 if the vehicle is a battery electric vehicle; 0 if it is a plug-in hybrid electric vehicle | 0, 1 |
| Vehicle class | Categorical | Classified based on overall length and wheelbase | A00, A0, A, B, C, D |
| Price segment | Categorical | Official factory guide price segment of the vehicle | ≤50k RMB, 50–80k, 80–100k, 100–150k, 150–200k, >200k |
| Age | Categorical | Actual age based on the year of birth | 20–25, 25–30, 30–40, etc. |
| Gender | Binary | 1 if male, 0 if female | 0, 1 |
| Has child | Binary | Whether there is a child in the household | 0, 1 |
| Has pregnant woman | Binary | Whether there is a pregnant woman in the household | 0, 1 |
| Elderly co-residence | Binary | Whether the household contains an elderly member aged 60 or older | 0, 1 |
| Elderly same city | Binary | Whether the user’s elderly parents live in the same city as the user | 0, 1 |
| User is elderly | Binary | Whether the user is older than 60 | 0, 1 |
| Has spouse | Binary | Whether the user has a spouse | 0, 1 |
In Table 1, I list the three explanatory variables of interest and nine additional individual/household variables. The demographic and household covariates are included in the regression to reduce omitted-variable bias. Since the commuting distance of a user may correlate with residential location, work location, life stage, and family structure, holding these covariates constant is essential for isolating the causal effect of commuting behavior on range demand. Although all control variables are measured categorically or binary, they capture the key dimensions of the multidimensional user profile that is recommended for electric vehicle market segmentation.
2.3. Construction of the effective commuting-distance and range variables
The original commuting-distance variable in the dataset is recorded as an interval: 0–2 km, 2–5 km, 5–10 km, 10–25 km, or longer than 25 km. I therefore aggregate the first three intervals into the reference category of short commutes (<10 km), the 10–25 km interval as the medium-commuting dummy, and the last interval (over 25 km) as the long-commuting dummy. This aggregation is consistent with the empirical convention in transport research: medium-distance commuting often represents a threshold where range anxiety begins to appear in electric vehicle usage, while extremely long commuting distances impose a binding constraint on the daily driving cycle.
Although the official range is expressed in continuous kilometers, it is actually determined at the product design stage before the buyer chooses a car. Therefore, when we examine a used or newly purchased vehicle’s certified range, we are observing the outcome of a consumer choice within the available product menu in the corresponding price segment and vehicle class. A user with longer commuting needs will generally scan the set of models within his or her budget and will select one whose range is sufficiently high. This behavioral mechanism is consistent with a discrete-choice approach, but a linear regression with dummy variables offers a convenient and interpretable summary of the average relation between commuting distance and range demand.
3. Econometric Model and Identification Strategy
3.1. Baseline dummy variable linear regression model
To estimate the effect of commuting distance on electric vehicle range choice, I specify the following dummy variable linear regression model:
$$
\text{Range}_i = \beta_0 + \beta_1 \text{CommuteMedium}_i + \beta_2 \text{CommuteHigh}_i + \alpha \mathbf{X}_i + \varepsilon_i ,
$$
where \( i \) indexes an individual user. The dependent variable \( \text{Range}_i \) is the certified electric range of the vehicle owned by user \( i \). \(\text{CommuteMedium}_i\) is a dummy variable that takes the value 1 if the user’s commuting distance belongs to the 10–25 km interval, and 0 otherwise. \(\text{CommuteHigh}_i\) is a dummy that takes the value 1 if the commuting distance exceeds 25 km, and 0 otherwise. \(\mathbf{X}_i\) is the vector of observable control variables including gender, the presence of a child, the presence of a pregnant woman, the co-residence of an elderly member, the same-city residence of elderly parents, whether the user is elderly, and whether the user has a spouse. \(\varepsilon_i\) is an error term reflecting unobserved random shocks.
The omitted reference group is represented by the constant term \(\beta_0\), which corresponds to users with a commuting distance below 10 km. Therefore, \(\beta_1\) measures the average difference in range choice between medium-distance commuters and short-distance commuters, conditional on controls. \(\beta_2\) measures the average difference between long-distance commuters and short-distance commuters, conditional on controls. I estimate both a model without controls (by setting \(\alpha=0\)) and a model with the full set of controls. Comparing these two specifications checks whether the relation between commuting distance and range demand is confounded by demographic or household variables. All parameters are estimated by ordinary least squares (OLS). Because the explanatory variables are dummy variables, the regression is simply a tool for computing conditional group means and group differences.
3.2. Heterogeneity by powertrain type
BEV and PHEV users are structurally different. A PHEV has an internal combustion engine that extends its total driving range, mitigating the user’s range anxiety. The user of a PHEV can rely on gasoline when the battery is depleted. Therefore, even a long commuting distance does not force a PHEV buyer to purchase a vehicle with a high pure electric range. In contrast, a BEV buyer faces a hard constraint: if the battery range is not enough for the daily round trip, the car must be recharged during the day, which causes inconvenience. Thus I expect the effect of commuting distance to be stronger for BEV users. To test this hypothesis, I include interaction terms between the commuting-distance dummies and a BEV indicator:
$$
\begin{aligned}
\text{Range}_i =& \beta_0 + \beta_1(\text{CommuteMedium}_i \times \text{BEV}_i) + \beta_2(\text{CommuteHigh}_i \times \text{BEV}_i) \\
&+ \beta_3 \text{CommuteMedium}_i + \beta_4 \text{CommuteHigh}_i + \beta_5 \text{BEV}_i + \alpha \mathbf{X}_i + \varepsilon_i .
\end{aligned}
$$
In this model, the coefficient \(\beta_3\) captures the effect of a medium commuting distance relative to a short commuting distance for PHEV users. The sum \(\beta_1 + \beta_3\) captures the corresponding effect for BEV users. Similarly, \(\beta_4\) gives the long-distance effect for PHEV users, while \(\beta_2 + \beta_4\) gives the long-distance effect for BEV users. Finally, \(\beta_5\) measures the average gap in range selection between BEV users and PHEV users within the short-distance reference group. This is a standard difference-in-difference-type interaction specification.
3.3. Heterogeneity by price segment and vehicle class
Consumers with different budgets choose from very different product menus. A high price usually provides access to larger batteries and longer-range technology. However, for a high-income consumer, the marginal value of an extra kilometer of range may be relatively small because the consumer already has a large budget and does not need to economize on range. By contrast, a budget-constrained electric vehicle buyer may regard range as the most critical attribute because it is hard to upgrade the range after purchase.
I define a low-price dummy \( \text{LowPrice}_i \) that equals 1 if the vehicle price is below 200,000 RMB, and 0 otherwise. I then estimate:
$$
\begin{aligned}
\text{Range}_i =& \gamma_0 + \gamma_1(\text{CommuteMedium}_i \times \text{LowPrice}_i) + \gamma_2(\text{CommuteHigh}_i \times \text{LowPrice}_i) \\
&+ \gamma_3 \text{CommuteMedium}_i + \gamma_4 \text{CommuteHigh}_i + \gamma_5 \text{LowPrice}_i + \alpha \mathbf{X}_i + \varepsilon_i .
\end{aligned}
$$
In this specification, the main terms \(\gamma_3\) and \(\gamma_4\) represent the commuting effects for high-price users, while \(\gamma_1+\gamma_3\) and \(\gamma_2+\gamma_4\) identify the effects for low-price users. The same logic is applied to vehicle class. I define \( \text{Mini}_i = 1 \) when the vehicle is an A00, A0, or A-class model (the compact and economically oriented classes), and 0 otherwise. The interaction model becomes:
$$
\begin{aligned}
\text{Range}_i =& \delta_0 + \delta_1(\text{CommuteMedium}_i \times \text{Mini}_i) + \delta_2(\text{CommuteHigh}_i \times \text{Mini}_i) \\
&+ \delta_3 \text{CommuteMedium}_i + \delta_4 \text{CommuteHigh}_i + \delta_5 \text{Mini}_i + \alpha \mathbf{X}_i + \varepsilon_i .
\end{aligned}
$$
I call the parameters \(\gamma_1, \gamma_2\) (and analogously \(\delta_1, \delta_2\)) the “price-commuting-range elasticity coefficients” and “class-commuting-range elasticity coefficients,” respectively. They describe how the range sensitivity to commuting distance differs between market segments.
4. Empirical Results
4.1. Baseline results for the full sample
Table 2 reports the baseline estimates of the effect of commuting distance on electric vehicle range selection. Column (1) displays the model without any control variables, while Column (2) adds the full vector of individual and household control variables. The two columns yield highly similar coefficients for the key commuting-distance dummies, suggesting that the results are robust to the inclusion of observable characteristics.
| Variable | (1) No Controls | (2) With Controls |
|---|---|---|
| CommuteMedium (\(\beta_1\)) | 18.983*** (0.000) |
17.823*** (0.001) |
| CommuteHigh (\(\beta_2\)) | 40.064*** (0.000) |
39.944*** (0.000) |
| Constant (\(\beta_0\)) | 250.607*** (0.000) |
243.276*** (0.000) |
| Control variables | No | Yes |
| Observations | 10,677 | 10,624 |
Note: The dependent variable is the certified electric range in kilometers. The omitted commuting category is a distance below 10 km. Coefficients report OLS estimates with p-values in parentheses. * p<0.1, ** p<0.05, *** p<0.01.
Focusing on Column (2), the coefficient \( \widehat{\beta}_1 = 17.823 \) indicates that a user with a commuting distance of 10–25 km chooses, on average, an electric vehicle with 17.8 km more range than a similar user with a commuting distance shorter than 10 km. Likewise, a user commuting more than 25 km chooses a vehicle with \( \widehat{\beta}_2 = 39.944 \) km more range than a short-distance commuter. Both effects are statistically significant at conventional levels. The magnitude of the long-distance effect is roughly twice the size of the medium-distance effect, which is consistent with a monotonic but nonlinear relationship between commuting intensity and range demand: as the commuting trip becomes longer, users rapidly upgrade their range requirement. The stable coefficients across specifications suggest that commuting distance is not a pure proxy for income, age, or household composition; it carries its own informational content about the user’s mobility need.
4.2. Do BEV users respond more strongly than PHEV users?
The full-sample result pools BEV and PHEV users. However, the behavioral mechanism behind the range demand is very different between the two technologies. PHEV users can use the internal combustion engine as a backup, so their daily commute should not impose a strict constraint on the pure electric range. I therefore estimate the interaction model in Section 3.2. The regression output is presented in Table 3.
| Variable | (1) No Controls | (2) With Controls |
|---|---|---|
| CommuteMedium × BEV (\(\beta_1\)) | 12.506** (0.010) |
12.311** (0.011) |
| CommuteHigh × BEV (\(\beta_2\)) | 30.405*** (0.000) |
30.032*** (0.000) |
| CommuteMedium (\(\beta_3\)) | -0.572 (0.721) |
-1.451 (0.371) |
| CommuteHigh (\(\beta_4\)) | 1.598 (0.554) |
0.801 (0.766) |
| BEV (\(\beta_5\)) | 338.857*** (0.000) |
339.780*** (0.000) |
| Constant | 102.498*** (0.000) |
243.276*** (0.000) |
| Control variables | No | Yes |
| Observations | 10,677 | 10,624 |
Note: The omitted commuting category is <10 km. Controls include gender, child, pregnancy, elderly co-residence, elderly same-city, user elderly, and spouse status.
Column (2) is the preferred model. The interaction coefficient with the medium-commuting dummy is 12.311 and is statistically significant, which implies that when an electric vehicle user commutes between 10 and 25 km instead of less than 10 km, the BEV-PHEV range gap widens by about 12.3 km. The corresponding interaction for long commutes is 30.032 km and is highly significant. This indicates that the difference in range choice between BEV and PHEV users becomes much larger for longer commuting distances.
The main effect of CommuteMedium (\( \beta_3 \)) is -1.451 and not significant. The main effect of CommuteHigh (\( \beta_4 \)) is 0.801 and not significant. Thus, for PHEV users, commuting distance does not have any measurable influence on their choice of electric range. This finding is intuitive: because PHEVs are not limited by the battery-only range, their owners do not condition their purchase on the daily commuting distance. In contrast, for BEV users, the marginal effect is:
$$
\text{Medium effect for BEV} = \widehat{\beta}_1 + \widehat{\beta}_3 = 12.311 – 1.451 = 10.860 \text{ km},
$$
$$
\text{Long effect for BEV} = \widehat{\beta}_2 + \widehat{\beta}_4 = 30.032 + 0.801 = 30.833 \text{ km}.
$$
These derived effects are summarized in Table 4 for clarity. In the short-commuting benchmark group, BEV users already select vehicles with an electric range about 339.8 km longer than PHEV users. As commuting distance grows, this range premium further increases because BEV users face a binding daily electric-driving requirement.
| Group | Medium vs. short commute (km) | Long vs. short commute (km) |
|---|---|---|
| PHEV users | -1.451 | 0.801 |
| BEV users | 10.860 | 30.833 |
Note: Computed from Table 3, column (2), using the formula \(\beta_3+\beta_1(\times BEV)\) or \(\beta_4+\beta_2(\times BEV)\).
The contrast between BEV and PHEV users has strong implications for electric vehicle market analysis. On the one hand, if an automaker sells PHEVs, the commuting distance should not be used as a primary tool to predict electric range preferences. On the other hand, for BEV pure-electric markets, commuting distance is a strong predictor of range demand. Consumer profiling platforms should therefore collect commuting-related variables, especially for BEV prospects.
4.3. Heterogeneity among BEV users by price segment
Given that BEV users are the main responsive group, the next step is to explore which BEV buyers produce the significant commuting effect. The first dimension that I examine is the price segment. I define a low-price BEV as one whose official guide price is below 200,000 RMB. The estimation results are shown in Table 5, and only observations for BEV users are included.
| Variable | (1) No Controls | (2) With Controls |
|---|---|---|
| CommuteMedium × LowPrice (\(\gamma_1\)) | 5.809 (0.363) |
5.502 (0.392) |
| CommuteHigh × LowPrice (\(\gamma_2\)) | 27.234*** (0.006) |
28.867*** (0.004) |
| CommuteMedium (\(\gamma_3\)) | 1.109 (0.819) |
0.790 (0.872) |
| CommuteHigh (\(\gamma_4\)) | 0.628 (0.929) |
-1.368 (0.847) |
| LowPrice (\(\gamma_5\)) | -187.712*** (0.000) |
-186.955*** (0.000) |
| Constant | 577.486*** (0.000) |
573.775*** (0.000) |
| Control variables | No | Yes |
| Observations | 4,775 | 4,752 |
Note: LowPrice equals one if the vehicle official price is below 200,000 RMB.
Column (2) of Table 5 shows that the interaction term between the long-commuting dummy and the low-price dummy is large and statistically significant: \( \widehat{\gamma}_2 = 28.867 \). The interaction term for the medium commute is positive but smaller and insignificant (5.502 km). The main effects \( \gamma_3 \) and \( \gamma_4 \) are both close to zero and insignificant. Therefore, among the high-price BEV segment (price ≥ 200,000 RMB), commuting distance does not generate any significant difference in the selected electric range. In contrast, for low-price BEV users, the implied effect of a long commute is:
$$
\text{Low-price long effect} = \widehat{\gamma}_2 + \widehat{\gamma}_4 = 28.867 – 1.368 = 27.499 \text{ km}.
$$
The medium-commuting effect for low-price users is \( \widehat{\gamma}_1+\widehat{\gamma}_3 = 5.502+0.790 = 6.292 \) km, although this value is not statistically distinguishable from zero. The derived effects by price segment are presented in Table 6.
| Price segment | Medium vs. short commute (km) | Long vs. short commute (km) |
|---|---|---|
| High price (≥ 200,000 RMB) | 0.790 | -1.368 |
| Low price (< 200,000 RMB) | 6.292 | 27.499 |
The behavioral interpretation is straightforward. Users who can afford a high-priced electric vehicle are likely to be less budget-constrained. They can buy long-range models regardless of their commuting needs, and many high-priced electric vehicles already have a relatively generous range. Furthermore, high-income users may have other charging options or a second vehicle, reducing the need to tailor range precisely to the daily commute. By contrast, budget-constrained buyers cannot afford unnecessarily large batteries unless long-distance commuting actually forces them to buy a more expensive long-range trim. Thus, long commuting becomes a much stronger driver of range demand in the low-price segment.
4.4. Heterogeneity among BEV users by vehicle class
The second segment analysis considers the vehicle class. Chinese electric vehicle classes are based on body length and wheelbase, with A00, A0, and A classes representing compact, inexpensive, and often daily-commuting-oriented models. B, C, and D classes represent larger and more luxurious vehicles. I define the dummy variable \( \text{Mini} \) as 1 for the A00/A0/A classes and 0 otherwise. Table 7 presents the regression results for BEV users.
| Variable | (1) No Controls | (2) With Controls |
|---|---|---|
| CommuteMedium × Mini (\(\delta_1\)) | 9.757 (0.118) |
9.757 (0.120) |
| CommuteHigh × Mini (\(\delta_2\)) | 19.915** (0.026) |
21.248** (0.018) |
| CommuteMedium (\(\delta_3\)) | -0.457 (0.924) |
-1.303 (0.787) |
| CommuteHigh (\(\delta_4\)) | 5.201 (0.401) |
3.217 (0.606) |
| Mini (\(\delta_5\)) | -195.200*** (0.000) |
-194.785*** (0.000) |
| Constant | 580.144*** (0.000) |
575.632*** (0.000) |
| Control variables | No | Yes |
| Observations | 4,775 | 4,752 |
Note: Mini equals one if the vehicle class is A00, A0, or A.
In Column (2), the coefficient of interest \( \widehat{\delta}_2 \) is 21.248 and statistically significant at the 5% level. The main long-distance coefficient \( \widehat{\delta}_4 \) equals 3.217 and is insignificant, so larger-vehicle users do not alter their range behavior based on commuting distance. For Mini-class users, the implied long-distance effect is:
$$
\text{Compact-vehicle long effect} = \widehat{\delta}_2 + \widehat{\delta}_4 = 21.248 + 3.217 = 24.465 \text{ km}.
$$
For a medium commuting distance, the implied effect is \( \widehat{\delta}_1+\widehat{\delta}_3 = 9.757 – 1.303 = 8.454 \) km. Although the individual coefficient of the medium interaction is not significant, the combined coefficient is positive and consistent in direction with the long-distance effect. Table 8 summarizes the marginal effects by vehicle class.
| Vehicle class | Medium vs. short commute (km) | Long vs. short commute (km) |
|---|---|---|
| B/C/D classes (large) | -1.303 | 3.217 |
| A00/A0/A classes (compact) | 8.454 | 24.465 |
These results confirm that the price-based and class-based heterogeneity are closely related. Compact electric vehicle classes are usually cheaper and are designed for urban commuting and short-distance applications. When such vehicles are used by long-distance commuters, the range deficiency is particularly salient; therefore, users in this segment actively choose the longest-range variants available in the compact class. Conversely, buyers of B, C, or D-class vehicles are probably purchasing for a larger portfolio of travel purposes, and their range choices are already generous due to brand positioning and technological packages, making commuting distance less important.
5. Discussion
The empirical evidence shows that the market segmentation of electric vehicles should take commuting distance into account. For managers in the electric vehicle industry, the estimated coefficients can be used to forecast the likely range demand of potential customers. For example, if a BEV consumer profile indicates a daily commuting distance above 25 km and a purchase budget below 200,000 RMB, the model predicts that the customer needs a vehicle range approximately 27.5 km longer than the range of a short-distance commuter in the same budget segment. Such guidance can help dealers recommend suitable models and help automakers decide where to expand the battery capacity of entry-level vehicles.
One important implication is that the design of entry-level electric vehicles should not ignore long-commuting users. Many price-sensitive consumers charge at home but do not have easy access to workplace charging. For them, a 25-km one-way commute might translate into a 50-km round trip, which is a nontrivial fraction of the usable battery capacity of a low-cost BEV. Consequently, automakers targeting the low-price and compact segment should offer at least one long-range trim or a larger battery option. This advice is particularly timely because battery costs have been declining, making long-range entry-level electric vehicles more affordable.
Another important implication concerns the information collected during consumer profiling and online recommendation. Most digital marketing platforms currently collect demographic and income data, but they do not systematically collect the consumer’s daily commuting distance. My results show that for BEV buyers, commuting distance is not merely a useful variable; it is as important as price and vehicle class. Therefore, in market research surveys and lead-generation forms, automakers should include a simple question about the respondent’s usual one-way driving distance to work. Based on the responses, the system can segment traffic into short, medium, and long commuters and recommend vehicles with appropriate ranges in the corresponding price and class categories.
The analysis also has implications for policy-makers who design electric vehicle subsidies or range-related incentives. In China, the central and local governments often adjust subsidies by driving range thresholds. My results suggest that a threshold-based subsidy may be effective in steering long-distance commuters toward longer-range electric vehicles. However, because PHEV buyers are not sensitive to commuting distance in their electric-range choices, subsidy programs that target longer electric range on PHEVs may have limited behavioral effects. Instead, the policy should distinguish between the pure electric driving requirements of BEV users and the backup capability of PHEV users.
Naturally, this study has several limitations. First, commuting distance is measured only in intervals rather than as a precise continuous variable. In future work, I could use continuous navigation distances based on route recommendations to improve prediction accuracy. Second, the commuting distance used in this paper is the straight-line distance from home to work, while actual driving routes may be longer due to road network geometry. This measurement error is likely to attenuate the estimated coefficients and understate the true sensitivity. Third, the data come from only one Chinese municipality. Since weather, topography, charging infrastructure, and traffic congestion vary across regions, the magnitude of the commuting-distance effect may differ in other cities. Fourth, I do not observe many long-run factors such as access to home charging, workplace charging availability, household income in continuous form, or the presence of a second vehicle. In a more complete empirical framework, these factors would further sharpen the causal interpretation. Finally, the dependent variable is the manufacturer-certified CLTC range, not the real-world range experienced by drivers. In practice, real-world range depends on ambient temperature, driving speed, heating or cooling loads, and battery degradation. Users respond to the advertised range but ultimately care about usable range. Future research can combine vehicle telematic data with certified range to estimate the gap between official and real-world performance and to derive a more accurate “effective range” measure of consumer demand.
Despite such limitations, the closeness of the coefficients in the models with and without controls, and the high statistical significance of the interaction terms, provides confidence that the main findings are not artifacts of confounded data. The inclusion of detailed demographic and household covariates already corrects for a large set of observable confounders. The monotonic structure between commuting distance and range demand—with the long-distance effect being roughly twice the medium-distance effect—also corresponds to a rational response to the binding daily travel constraint.
6. Conclusion
This paper set out to understand whether commuting distance can predict the electric range chosen by new energy vehicle users, and whether the relationship is homogeneous across powertrain types, price segments, and vehicle classes. Using a rich sample of 11,219 BEV and PHEV owners from a 2024 car-human big-data platform, I estimate a series of dummy variable linear regression models. The results produce three main conclusions.
First, commuting distance has a significant and economically meaningful correlation with the certified electric range of the purchased vehicle. Medium-distance commuters choose vehicles with about 17.8 km more range than short-distance commuters, while long-distance commuters choose about 39.9 km more range. This finding corroborates the intuition that consumers treat the daily commute as a central anchor in their range requirement when they decide to buy an electric vehicle.
Second, this overall correlation is entirely driven by BEV users. For PHEV users, the commutating-distance coefficients are statistically insignificant, while the corresponding interaction coefficients for BEV users are positive and strongly significant. Long-commuting BEV users choose roughly 30.8 km more range than short-commuting BEV users. Therefore, the distinction between “pure electric” and “hybrid electric” ownership is fundamental for understanding range-sensitive consumer behavior.
Third, among BEV users, the low-price segment and the compact vehicle class exhibit the largest response to commuting distance. Long-distance commuters who choose a cheap BEV need to maximize range within a restrictive budget, so their range choice is very responsive to commuting pressure. The same pattern appears for A00/A0/A-class BEV buyers. For expensive, large BEVs, range is already plentiful and the user has more flexibility; hence the effect of commuting distance vanishes.
For the electric vehicle industry, the practical message is clear: when designing entry-level BEVs, automakers should not treat range as a purely upmarket feature. Many mid-range and compact consumers rely on their only electric vehicle for long daily commutes, and they are willing to pay for a larger battery if the vehicle remains within their budget. Online auto-configurators and dealer recommendation systems should use commuting-distance dummies as a key input, especially for BEV consumers with lower price preferences. By translating commuting behavior into a quantitative range recommendation, automotive companies can better satisfy heterogeneous users, reduce range anxiety, and support the continued diffusion of electric vehicles.
Future research should enrich the model with more precise measures of commute distance, direct information about charging infrastructure and charging behavior, and longitudinal data tracking range choices over time. Incorporating regional climate and road-traffic information would also make the predictions more applicable to the entire Chinese electric vehicle market. From that perspective, this paper is only the beginning of a data-driven exploration of the link between daily travel and electric vehicle range demand.
References
[1] The Chinese State Council. Fourteenth Five-Year Plan for Digital Economy Development, 2021.
[2] The Chinese State Council and the Central Committee of CPC. Data Twenty Articles on Constructing Data Basic Systems, 2022.
[3] Anonymous. Analysis of differences between real-world driving range and certified range of battery electric vehicles. Journal of Automotive Technology, 2020.
[4] Anonymous. Discrete choice experiments on heterogeneous willingness to pay for new energy vehicle attributes. Sustainability, 2024.
[5] Anonymous. Predicting new energy vehicle sales using linear regression. Proceedings of the International Conference on Energy and Environment, 2019.
[6] Anonymous. User segmentation and consumer preferences in the European electric vehicle market. Transport Research Arena, 2020.
[7] Anonymous. Sales forecasting for light trucks based on multivariate linear regression. Automotive Industry Research, 2015.