Conclusion

6.1 Summary of findings

This study treated the COVID-19 pandemic as a forced experimentation on London Underground, analysed how people’s travel and commuting demand was disrupted by the incident, comparing with the trajectory before the pandemic, and how long did the effect last after the pandemic. Additionally, this study also examined if the stations exhibited patterns spatially, where spatial clustering occurred, and how the pattern changed over time from 2021 to 2023.

The results suggested that, temporally, 2021 was the year that experienced most significant drop in underground demand. But since 2022, the ridership recovered quickly, especially in commuting related trips, which in some cases had exceeded the ridership in 2019 already, and the impact of COVID was barely detectable. According to Larcom et al. (2017), a short 2-day strike took place in London Underground caused long lasting effect in people’s travel behaviour. The results of this study seem to show otherwise, despite the fact that COVID pandemic disrupted London’s public transport for much longer time. However, Larcom et al studied the travellers’ behaviour. While people might choose different Underground lines when facing forced disruption, the outcome of such behaviour will not be reflected in the total ridership counts, since the total demand within the Underground system might remain the same. Since this study analysed TfL’s annual Entry and Exit counts data, it provided no indication of people’s choices, which might explain why the effect of COVID-19 in this study appeared to be short lasted. In contrast, Murcio et al. (2023) did not focus on behaviour science, but their OD dataset of work and home effectively reflected travellers’ purpose, and their results suggested that by 2022, public transport commuting demand in London was still lower than the pre-pandemic level, which indicated a long-lasting effect of disruption on public transport.

Meanwhile, this study did find out that mixed purpose underground demand, calculated with annualised usage data, never reached the 2019 baseline by 2022. However, the result of commuting demand represented by weekday ridership bounced to 150% of pre-pandemic level, remained unexplained. Multiple studies showed reduction in commuting in London and other cites after the pandemic, and adaptation of remote-working might cause reduction in commuting (Bansal et al., 2022; Jenelius and Cebecauer, 2020; Balbontin et al., 2022). This study tried to distinguish commuting demand by selecting Monday to Thursday within the week, the categorisation remained at fairly low-grained level. Even though weekday ridership recovery behaved different from annualised recovery, there was no direct indicator to prove the phenomena was a result of changing commuting demand, since multiple factors can contribute in the demand change of public transport. Finer datasets like ridership counts in 15 minutes intervals (Sharma et al., 2023) can help identifying commuting trips and peak commuting time periods more precisely, and more complex methodology can be used. This study analysed low grain data, thus simple calculation of recovery rate was designed. Nonetheless, compared with the relatively stable pre-pandemic underground ridership, so far we can only confirm 2021 is significantly disrupted temporally.

Spatial findings showed similar pattern to temporal conclusion. Results suggested that spatial clustering pattern was both statistically and visually observed, but became less significant over time. The location of clusters showing low recovery aligned with existed studies. Batty et al. (2020), Trasberg and Cheshire (2021), and Sharma et al. (2023) all detected reduction in public transport demand and social activities in central London areas. The spatial method used in this study controlled clustering by defining number of neighbouring stations. By keeping fixed neighbours for each station, this method prevented the impact of areal station density when forming cluster. Still, since the closest k stations were considered neighbours, in high station density area, the clustering represented overall smaller region, comparing with low station density suburbs. It indicated that central London was an area with lower Underground recovery rate for both mixed purpose and commuting demand. Meanwhile, locations of high recovery cluster could not be clearly identified in this study, while the high recovery location in other studies varied with mode of travel, definition of cluster, time period of dataset, and other factors. It seemed only central London remained as a low public transport recovery hollow, especially for commuting demands, among different datasets and methodologies.

One common approach to further address the spatial pattern is aggregating the public transport ridership index to spatial units, so the local socio-economic factors can be included when analysing recovery trend. Continuous datasets like mobile phone signal or footfall data can be more easily aggregated to spatial units (Forouhar et al., 2025; Trasberg and Cheshire, 2023). The underground station dataset in this study is a sparse, discrete and small group of fixed infrastructure nodes. With the study subjects limited to underground stations, it is more coherent to test spatial clustering on station level, while forcing an areal regression by assigning stations to census units might create distorted outcome.

In conclusion, this study identified London Underground ridership trend before and after the pandemic, distinguished commuting pattern, which showed only significant decrease occurred in 2021. From 2022 demand bounced back fast, exceeded the pre-pandemic ridership in 2023, although the precision of commuting ridership counts remain unverified due to the dataset’s nature. Spatially, significant clustering was observed in 2021, where central London appeared to be the area with lowest recovery profile. High recovery areas varied under different situation, although it was likely that stations of London Underground of District Line and Metropolitan Line had relatively high recovery. Such pattern diminished over time, and by 2023 only a small range of clusters remained detectable, mainly low recovery stations in central London.

6.2 Limitations

First, the earliest annual Entry and Exit data published by TfL dated back to 2017, while the most recent one was the counts of 2023. With more datasets from years before the COVID pandemic and more recent datasets, the recovery rate and pre-pandemic ridership trend might be identified more precisely. Moreover, since 2019 was chosen as the baseline, new stations and lines opened after 2019 were excluded from this study, while with more stations and newer lines the outcomes might be interpreted more comprehensively.

Second, fare zones were used in this study to display the overall central-peripheral distribution of station, but fare zone was administrative rather than functional categorisation. Moreover, stations on the boundaries of multiple zones were grouped to the more central zone, which likely caused bias in analysis.

Third, in the spatial analysis, the k-nearest value remained crucial in identifying neighbours in Local Moran’s I and Getis-Ord Gi* experiments, but the robustness of k value used in Gi* was not tested, and overall the choice of k value could not be validated.

6.3 Future work

This study can be the base of future research with different approaches. First, in the future, when more data is available, this study can be repeated to continue studying the impact of COVID disruption on Underground system. Different types of data like OD flows and mobile signal data can be introduced to the study as well, and calculating Underground recovery rates with datasets of different profile and granularity might provide more insightful details for this topic.

Second, introducing socio-economic factors to this study might help explaining the recovery trend discovered in this study. If better method of aggregating discrete, limited point data to spatial units is developed, or if socio-economic factors can be represented by Underground stations, the study will get richer context, and causal attribution of recovery trend can be interpreted.

Third, since London Underground is only part of the TfL rail service system, in the future, more mode of travel can be introduced. An analysis including underground, overground and other rail travel mode might demonstrate a full picture of recovery pattern in London’s rail system. Moreover, non-rail travel modes like shared bicycle scheme and bus can also be included, since people may choose multiple public transport infrastructures when commuting. Although debated in Susilo and Cats (2014) that travellers using different mode of public transport might have different characteristics, making comparing different mode of public transport user less meaningful, different modes still complement each other in functions, which make it possible that commuters might still use more than one mode of public transport to complete their journey, hence studying multi-mode travel completes the arc of commuting.

6.4 Policy implications

The findings of this study can also be read in relation to planning and operational policies. The persistent under-recovery of central London stations highlights the risk of over-reliance on commuting and tourism trips in the core. As TfL’s funding is closely tied to fare revenues, this uneven recovery raises questions about how the system can remain financially sustainable if central ridership is structurally weaker in the long term. On the other hand, the stronger recovery of peripheral stations points to the resilience of local trips, which may support ongoing policies to encourage more polycentric patterns of activity in London. The results also suggest that new travel rhythms, with altered weekday peaks, may require adjustments to service scheduling and ticketing products. While this dissertation cannot make detailed policy recommendations, it shows how the pandemic revealed vulnerabilities and strengths in different parts of the Underground, which should be considered in debates about long-term resilience of London’s public transport.

In summary, this study used recovery rates and spatial statistics to examine how London Underground demand evolved after the COVID-19 disruption. It found that the sharpest impact was confined to 2021, with demand largely rebounding by 2022, but that central London remained slower to recover than the periphery. By 2023, spatial clustering had weakened, suggesting a new equilibrium rather than a continuing shock. These results extend existing work by applying a fare-zone perspective and station-level clustering methods, and they emphasise the importance of both temporal and spatial lenses in understanding post-pandemic mobility.