Previous studies that projected vehicle stock and ownership levels have used Gompertz, log-linear, and logistic functions to project vehicle ownership. Among the models listed, the Gompertz function is deemed more appropriate and better fits the historical trends than other models (Wu, Zhao and Ou 2014).

Typically, the Gompertz function has per capita GDP as a key independent variable. The country’s per capita GDP directly affects the change in vehicle ownership, depicted by an ‘S’- shaped curve or a sigmoidal curve (Arora et al. 2011). It assumes that growth will start slowly at low-income levels, then accelerate as income rises, and finally slow down as saturation approaches (Dargay et al. 2007; Wu et al. 2014). Equation 1 below provides the Gompertz function. A more detailed explanation of the function and its terms is given in Section 3.5.

Vt is the vehicle ownership at year t; is the saturation level; is the speed of adjustment; α and β are shape parameters; G denotes per-capita income. The saturation level (γ) is the maximum level of vehicle ownership (vehicles per 1,000 people) for a country, and varied methods are used to determine it. The speed of adjustment (θ) is the rate at which vehicle ownership approaches saturation (0<θ<1)

The studies involving projections for the road transport sector in India analyse vehicle stock growth by segment, using the projected values to estimate future oil demand, emissions, and energy consumption (Arora, Vyas and Johnson 2011; Singh, Mishra and Banerjee 2020). A previous CEEW study (Soman et al. 2020) developed the vehicle stock across all passenger vehicle segments at national level. However, none of these studies have attempted to project the vehicle stock at a regional level. This study aims to address this gap by providing regional-level projections of vehicle stock, allowing for more granular understanding of trends and variations in different regions.

Different studies have used different saturation levels. The level used in the Gompertz function for the UK and other industrialised economies lies between 0.4 and 0.7 (vehicle ownership per 1,000 people) (Button, Ngoe and Hine 1993). The study by Huo et al. (2007) analysed the data from 18 countries and deduced that the saturation rate for the US as a representative of North America was 0.8. Similarly, Europe’s saturation level was 0.6, and Japan’s, as a representative of Asian economies, was 0.55.

For India, the study by Singh, Mishra and Banerjee (2020) examined the growth of twowheelers (2W) and cars individually under two distinct scenarios: conservative and aggressive. They have set the saturation levels at 0.25 and 0.35 for 2W and 0.15 and 0.25 for cars, respectively, in each scenario. The CEEW study (Soman et al. 2020) divided the vehicle stock into 2W and combined all other vehicle categories (as the combined saturation level for the latter is still much lower than that of 2W). They have assumed the saturation level to be 0.25 for 2W and 0.15 for all other vehicles combined. The Indian studies assumed saturation levels based on trends observed in other countries like China and Japan.

From the literature review, we found that all prior studies used pre-determined saturation levels that the countries could eventually reach. De Silva et al. (2022) hypothesised that countries could have varying saturation levels, and that it would not be a universal constant. The saturation level (γ) is influenced by many variables such as infrastructure, household size, population density, and share of public transport. Considering India’s regional diversity in culture, income levels, infrastructure, etc., the use of γ as a near-constant value across districts and states may not be appropriate. Thus, using geography-specific γ will provide a clearer picture of vehicle ownership estimates, allowing for more complex policy variable interactions.

In this context, our current study estimates vehicle ownership at district level in India for 2050, building upon the CEEW 2020 study and its approach. It delves into more detail, examining the heterogeneity in vehicle ownership patterns across individual states and districts which allows us to discern regional disparities and trends. We use district-specific saturation rate (γ) and speed of adjustment (θ) (see Section 3.5) to estimate the district-level vehicle stock and its growth, aiding sub-national level policymaking.

The vehicle stock projection and subsequent new vehicle registration estimates are based on the aforementioned data and assumptions. This section discusses the results of vehicle stock projection and ownership levels by segment at the regional and national level. We then go on to compare our national stock estimates with existing transportation stock forecast models.

National-level results

We projected the vehicle ownership by segment and district from 2024 to 2050 using Equation 4. We then calculated the national-level estimates by aggregating at the district and state level.

We estimate that the total vehicle ownership (inclusive of all segments mentioned in Table 1) will reach 309 per 1000 people in 2050, compared to 163 per 1000 people in 2023, as observed in Figure 3. The slight change in trajectory observed between 2019 and 2022 is caused by the COVID-19 pandemic, when actual new registrations were at an all-time low, significantly impacting the stock and, subsequently, the ownership levels. In comparison, China had approximately 310 vehicles per 1,000 people in 2023, while the European Union countries had an average of nearly 600 vehicles per 1000 people (Xinhua News Agency 2024, Eurostat 2024). Interestingly, the GDP per capita of China (at USD 12,970 in 2024) is around five times higher than that of India (at USD 2,700), while that of the EU (at USD 46,640) is 17 times higher (International Monetary Fund 2024). However, the vehicle ownership in China and the EU are only 2 and 3.7 times higher respectively, as these regions are more urbanised and have much bigger public transport systems.

By 2050, India’s GDP per capita (in constant prices) is projected to reach USD 10,500, a level comparable to China’s in 2024. Correspondingly, vehicle ownership in India is expected to reach a similar level to China’s current ownership level.

Figure 4 illustrates the relationship between vehicle ownership and GDP per capita across various countries, categorised by income group based on the World Bank classification (The World Bank n.d.). Among the listed nations, New Zealand records the highest vehicle ownership, with 934 vehicles per 1,000 people and a GDP per capita of approximately USD 47,000. We attribute this high ownership level to the country’s low population density and dispersed housing patterns, making vehicles essential for mobility. The USA is next, with 850 vehicles per 1,000 people and a GDP per capita of approximately USD 78,000. A relative lack of public transport and car-centric urban planning contributes to the high rate of vehicle ownership in that country. Australia, the European Union (EU-28), and Japan also rank among the top five, exhibiting high vehicle ownership due to similar urban development patterns and economic conditions.

In contrast, China, Russia, and South Africa, categorised as middle-income countries, have ownership rates ranging between 200 and 350 vehicles per 1,000 people. However, Brazil stands out within this group with close to 460 vehicles per 1000 people due to favourable government policies and incentives promoting vehicle ownership.

By 2050, two-wheelers are expected to account for approximately 71 per cent of the total vehicle ownership, translating to 219 2W per 1,000 people out of a total of 309 vehicle per 1,000 people. In comparison, 2W ownership in 2023 stood at around 127 per 1,000 people. While the absolute number of 2W increases by 73 per cent, their share in the overall vehicle mix declines from over 78 per cent in 2023 to 71 per cent in 2050.

Our analysis indicates that private cars (4W and SUVs) drive the overall vehicular growth, growing from 24 cars per 1,000 people in 2023 to 57 cars per 1,000 people in 2050. Even by 2050, car ownership in India is projected to be significantly lower than the 2017 levels of China, with 105 cars per 1,000 people, and Colombia, with 65.5 cars per 1,000 people (Helgi Library n.d.).

Figure 6 illustrates ownership levels of other vehicle categories (except for 2W and 4W). Other growth drivers include passenger three-wheelers (3W-P) and light goods vehicles (LGV). The number of passenger three-wheelers is projected to increase from 3.4 per 1,000 people in 2023 to 8.2 per 1,000 in 2050. Similarly, the number of LGVs is expected to rise from 4.4 per 1,000 people in 2023 to 11 per 1,000 in 2050. The number of buses per 1,000 people is low at 0.8 in 2023 but it is projected to increase modestly to 2.3 per 1,000 people in 2050. The other vehicle categories, including taxi, HGV, MGV, and 3W-G, are at the lower spectrum of the S-shaped curve, which means they might take longer to saturate.

The overall stock was estimated to be at 226 million by 2023 and is projected to increase to 494 million by 2050, thus growing by nearly 2.2 times between 2023 and 2050. We estimated the total 2W on-road to be 350 million by 2050, doubling from 175 million in 2023, and forming 70 per cent of the total stock even in 2050.

Unless there is a significant upward mobility in per capita incomes, substantial increases in road infrastructure and lower taxes on 4W (currently 28 per cent for 2W and going as high as 50 per cent for SUV), we do not expect 4W to gain much share. The 4W stock will grow from 30.5 million units in 2023 (13 per cent share) to 84.9 million in 2050 (17 per cent share). Similarly, the SUV stock grows by 2.7x between 2023 and 2050, buses by 3.3x, taxis by 2.9x, LGV by 1.9x, HGV by 1.8x, and 3W-P by 1.8x.

State and district-level results

Our analysis estimates that, by 2050, Uttar Pradesh will be the top state in terms of overall vehicle stock, followed by Maharashtra, Bihar, Madhya Pradesh, and Gujarat. Uttar Pradesh and Bihar, collectively home to approximately 27 per cent of India’s population (in 2024), are also among the poorer states in the country. With large population bases and low income levels in these states at present, we anticipate substantial growth potential in terms of per capita incomes. These two states combined will account for 25 per cent of India’s total vehicle stock in 2050. Uttar Pradesh’s stock, in particular, will grow from 29 million vehicles (in 2023) to 93 million vehicles by 2050, a threefold increase.

Among the top ten states by absolute vehicle stock, Gujarat is projected to have the highest vehicle ownership in 2050, with 351 per 1,000 people. Uttar Pradesh, Karnataka, Tamil Nadu, and Andhra Pradesh follow, with ownership levels of 336, 332, 328, and 305 vehicles per 1,000 people, respectively. Notably, while Tamil Nadu’s ownership increases by 86 vehicles per 1,000 people between 2023 and 2050, Uttar Pradesh experiences a higher increase of 214 vehicles per 1,000 people—2.5 times that of Tamil Nadu. Andhra Pradesh, Karnataka, and Maharashtra each see an increase of approximately 120 vehicles per 1,000 people, while Gujarat records a rise of 139 vehicles per 1,000 people. Meanwhile, Bihar and Madhya Pradesh witness a steep rise of 156 vehicles per 1,000 people, and 158 vehicles per 1,000 people, respectively. This trend suggests that states with lower current income levels are expected to experience more rapid growth in vehicle ownership, whereas relatively wealthier states are projected to see moderate growth in vehicle ownership.

Figure 10 shows the total vehicle stock of the country on a simplified district map, splitting 2W and all other vehicles. Between 2023 and 2050, the growth in vehicle stock can be primarily attributed to the states of Maharashtra, Gujarat, Delhi, Uttar Pradesh and Bihar. Interestingly, the growth in non-2W stock is lesser in states like Uttar Pradesh, Bihar, and West Bengal compared to the 2W growth; this is likely due to the lower GDP per capita levels in these states, indicated in a preference for cheaper modes of road transport. Overall, the southern states do not show a marked increase in stock as observed in northern states, apart from some urban centres like Mumbai and Bengaluru. Southern states are experiencing population peaking in the coming decade and rising income levels; the vehicle stock thus stabilises in the near future.

Figure 11 shows the district-level distribution of vehicle stock seen in Figure 8 at three aggregated levels: private cars (4W and SUV), public transport (3W-P, Taxi and Bus), and goods vehicles (3W-G, LGV, MGV, and HGV). Contrasting the 2W stock in Figure 10 with the private car stock in Figure 11 shows similar patterns of adoption in high-income regions; lower-income regions see lower adoption of cars than 2Ws. Public transport growth is also observed mainly in highly dense urban agglomerations like Delhi, Mumbai, Thane, Bengaluru, Ahmedabad, etc. The trends observed in new registrations over the past ten years indicate that there will be a limited role for public transport in tier-3 towns and rural areas. The projections indicate that there will be an increase in demand for goods transportation across the country, somewhat in contrast to the trends seen in private and public transport. The prevalence of goods vehicles is observed in districts that have major ports, such as in Kachchh, Surat, Mumbai, Chennai, etc. Notably, the state of Nagaland shows a notable increase in the number of goods vehicles; this is due to the large number of goods vehicles that are getting registered in the state due to tax and other regulatory benefits.

Figure 12 shows the top ten districts by vehicle stock in 2050. These 10 districts alone (out of 640 districts as per the 2011 census) account for over 10 per cent of the total national stock in 2050. The National Capital Territory of Delhi (represented as a single district) has the highest number of active vehicles, growing from 4.3 million in 2023 to 10 million in 2050. Bengaluru district in the state of Karnataka will have the second highest stock, growing from 4.0 million in 2023 to 6.1 million in 2050. The state of Maharashtra has three districts in the top ten— Thane, Pune, and Mumbai—totalling 15 million vehicles in 2050 (up from 7.2 million in 2023). The stock growth rates for almost all the top ten districts fall substantially over the years; the compound annual growth rate (average for the top ten) of 4.3 per cent from 2020–2030 reduces to 3.3 per cent from 2030–2040 and then drops to 1.7 per cent from 2040–2050. This signifies that the stock growth approaches saturation around the year 2050.

The states with the highest and lowest 2W ownership levels are shown in Figure 13. Puducherry and Goa, both coastal regions with high tourist footfall, also have the highest penetration of two-wheelers. Puducherry has over 400 2W per 1,000 people already by 2025, much higher than the national average of 150 (in 2023) or even 230 (in 2050). The states of Telangana, Punjab, and Tamil Nadu complete the top five. Apart from Uttar Pradesh, the other states start saturating in ownership due to plateauing/falling population levels towards 2050. The northeastern states of Manipur, Arunachal Pradesh, Meghalaya, Sikkim, and Nagaland have the lowest penetration of 2W; the challenging weather and terrain in these regions are the likely reasons.

Figure 14 shows the states with the highest and lowest penetration of private four-wheelers. Chandigarh already has a much higher vehicle ownership level, potentially due to its wide roads, ample parking, and higher income levels. Himachal Pradesh is among the top ten states in terms of per capita income and has lower levels of public transportation owing to the mountainous terrain and narrow roads, thus leading to higher penetration levels of fourwheelers. Flow of income from other regions into a given state or district is difficult to track; for example, an RBI bulletin from 2018 shows that the state of Kerala received the highest amount of inward remittances from abroad, owing to a large expat population especially in the Middle East (RBI 2018). Therefore, even though Kerala ranked 15th in GDP per capita in 2024, it still recorded the fourth highest 4W ownership level. Haryana also has high income levels (rural prosperity is high) leading to high ownership levels. Goa is among the top-three states in terms of GDP per capita and has a large tourism economy, which contribute to the higher prevalence of four-wheelers.

According to the National Family Health Survey (NFHS-5), Chandigarh, Goa, and Haryana have the highest proportions of their population in the top wealth quintile, at 79 per cent, 61 per cent, and 48 per cent, respectively (International Institute for Population Sciences and ICF 2021). This wealth concentration is associated with higher car ownership in these states. The bottom-five states are characterised by lower GDP per capita levels and/or higher income inequality between districts. Bihar has the lowest GDP per capita among all states, with approximately 43 per cent of its population in the lowest wealth quintile, resulting in lower car ownership. Even by 2050, Bihar is projected to be among the lowest in terms of GDP per capita, keeping car ownership below the national average. Similarly, West Bengal ranks among the bottom-ten states in GDP per capita and has the second-highest wealth inequality in the country, with nearly 32 per cent of its population in the lowest wealth quintile, contributing to lower car ownership.

Sensitivity analysis

We conducted a sensitivity analysis to identify the effect of saturation limits on stock projections. The saturation limit was varied in increments of 25 or 50 to observe its effect on the stock. As we had mentioned earlier, the saturation and speed were provided as a range for the solver module in the sum of least squares approach to better fit the historical data. For the sensitivity analysis, the upper bound of the range was incremented by 25 or 50 units depending on the vehicle segment, and the results are illustrated in Figure 15.

In the 2W category, we expanded the upper limit of the saturation level in increments of 50 units, establishing four scenarios: 250 (baseline), 300, 350, and 400. Taking 250 per 1,000 people as our baseline scenario, we observed variations in the projected vehicle stock for 2050. Specifically, compared to the baseline, the 300-unit scenario resulted in a 4 per cent higher stock, the 350-unit scenario in an 8 per cent higher stock, and the 400-unit scenario in a 12 per cent higher stock in 2050.

For the 4W segment, we increased/decreased the upper bound by 50, establishing five scenarios: 50, 100, 150 (baseline), 200, and 250. By 2050, stock increased by 1 per cent in the 200- and 250-unit scenarios, while it declined by 6 per cent and 14 per cent in the 100- and 50- unit scenarios, respectively. Similarly, for SUVs, the stock decreased by 9 per cent and 3 per cent in the 50- and 100-unit scenarios, while it increased by 1 per cent and 2 per cent in the 200- and 250-unit scenarios. Across other vehicle segments, the change in saturation range revealed minimal impact (between 1 and 2 per cent) on the overall stock as these segments had achieved optimal parameters.

The baseline was set at 250 per 1,000 people for 2W, and 150 per 1000 people for other vehicle categories, aligning with the saturation values used in the Soman, et al. (2020) study. The baseline values were selected from this study, as the shape parameters of the Gompertz equation were also derived from it.

Comparison with existing models

We compared our vehicle stock estimates against the existing projections. Eight studies were chosen, including an estimate from a CEEW study conducted in 2020, as listed in Table 4.

Table 4 Studies used for comparison and validation of projections

Our estimates are close to the previous forecast by CEEW (Soman et al. 2020). The differences can be attributed to using the same values of saturation and speed of adjustment in the mentioned study and the disruptive impact of COVID-19 on vehicular growth. Our estimates are 48 per cent lower than NITI-RMI’s (NITI Aayog and Rocky Mountain Institute) 2017 estimate. It is unclear whether the NITI-RMI study uses Gompertz to predict the future vehicle stock, but they have assumed a national GDP growth with a CAGR of 6.7 per cent to arrive at 598 million vehicles as of 2030.

The comparison of overall stock estimates between this study and TERI for the year 2050 reveals a significant variation (The Energy and Resources Institute 2024). Our analysis projects the overall vehicle population to be 494 million in 2050 while TERI’s estimate is higher at 847 million, reflecting a variation of approximately 42 per cent. This variation can largely be attributed again to differences in methodology, particularly the use of singular saturation levels for the entire country in TERI’s study. The TERI study uses saturation levels of 300 per 1,000 people for 2W, 50 for 3W-P, 200 for 4W and SUV, and 25 for MHGVs. It applies these values at a national level. In contrast, our study derives the saturation level dynamically, based on historical patterns observed at district level. Additionally, a key distinction is that TERI uses a top-down approach to aggregate stock at the national level, whereas our study adopts a bottom-up approach to aggregate stock at the district level.

The 2W stock matches closer with the IIT-B estimates, with the variation largely constant in absolute terms through the future years (Singh, Mishra and Banerjee 2020). For example, the IIT-B study projected higher 2W stock by 34 per cent in 2015, and the difference narrowed to 32 per cent by 2030 and 20 per cent by 2050. The significant difference could be the base year and the difference in saturation values and speed of adjustment. The IIT-B study’s projection year starts in 2012, while in our study, the projection begins in 2024. Similarly, the IIT-B study uses a fixed saturation limit and speed of adjustment, while the two parameters in our study are chosen based on the best fit of the historical data.

In contrast, for the car segment (4W and SUV combined), the difference between our study and IIT-B estimates diverges from 2015. The IIT-B study projected car stock 51 per cent higher in 2015, rising to 218 per cent by 2030 and showing a marginal decline to 190 per cent by 2050. This is again due to the underlying assumptions in the model and the differences in the saturation and speed of adjustment values.

Beyond absolute vehicle stock, it is also important to examine other key metrics, such as total passenger demand and freight demand. The GDP per capita increase reflects not just on the number of vehicles, but also on the distance driven by the population. With growing affluence, leisure trips increase in distance and duration. Similarly, with increased industrialisation and consumption, the total tonnage of freight and distances goods are moved will increase. In 2023, the total passenger demand was approximately 7,370 billion kilometres, while the total freight demand was approximately 2,095 billion tonne-kilometres. By 2030, passenger demand is projected to rise to 10,700 billion kilometres, with freight demand reaching 3,000 billion tonne-kilometres.

Our passenger demand estimate match much closer with IRADe and TERI’s estimates. The difference between our estimates and those of the CEEW-GCAM model can be attributed to variations in modelling approaches, including their exclusion of taxis and other underlying methodological variations (CSTEP, CEEW, IRADe, PNNL, and TERI 2019).