Revival of positive nostalgic music during the first Covid-19 lockdown in the UK: evidence from Spotify streaming data

The study investigates how the first UK Covid-19 lockdown (beginning 26 March 2020) altered music listening behavior on Spotify, focusing on nostalgia (operationalized as listening to songs older than five years) and positivity (song valence). Motivated by evidence that music serves as an emotion regulation strategy and that negative emotions can trigger nostalgia, the paper asks whether lockdown led users to listen more to older (nostalgic) and more positive music. It aims to provide observational, behavior-based evidence using streaming data, complementing prior self-reported survey findings, to understand how widespread societal shocks affect cultural consumption and emotion regulation behaviors.

Prior research shows music impacts mental health and is used for emotion regulation generally and during the pandemic. Surveys during Covid-19 reported increased listening to nostalgic music and to happy or childhood-related songs (Ferreri et al., 2021; Fink et al., 2021; Granot et al., 2021). Theoretical links suggest negative emotions predict nostalgia (Wildschut et al., 2006), nostalgia can have positive effects (Van Tilburg et al., 2013; Cheung et al., 2017), and music can evoke nostalgia (Barrett et al., 2010). Pandemic-era work found music-evoked nostalgia functioned as an emotion regulation strategy (Gibbs & Egermann, 2021). Literature on nostalgia consumption and marketing indicates difficult times can heighten nostalgia-seeking. The study situates itself within this body by using large-scale streaming data to test for behavioral shifts toward nostalgic and positive music, and to disentangle nostalgia-seeking from a general positivity bias.

Data: Spotify UK daily Top 200 chart, including song titles, daily play counts, release dates, and Spotify audio features, from 1 Jan to 31 Jul 2020 (covering the first wave). Approximately 4 trillion plays across more than 4,000 unique charted songs are analyzed. Lockdown dates come from official announcements, and UK Covid-19 incidence rates (per million) from the EU Open Data Portal. The UK is chosen for its high streaming volume and predominance of English-language songs. The same January–July window in 2019 is used for robustness checks. Operationalization: Old (nostalgic) music is defined as songs released more than 5 years (1825 days) prior to the play date; otherwise, recent. A binary indicator flags old vs. recent. Positivity is proxied by Spotify valence; songs are categorized as positive (valence > 0.667), negative (valence < 0.333), or neither (thresholds approximate the 75th and 25th percentiles). Main analysis: A weighted logistic regression (difference-in-differences design) models the probability that a charted play is an old song (Y=1). Each song–day observation is weighted by its daily play count to reflect popularity. Standard errors are clustered by day. The key regressors include: a lockdown indicator (1 on/after 26 March 2020), time (t) centered at lockdown day, t^2, interactions Lockdown×t and Lockdown×t^2 to allow post-lockdown trend changes, the log of the 7-day moving average of confirmed Covid-19 cases per million, and controls: number of newly released songs (last 30 days), log total daily plays, the average oldness level on the same calendar day in 2019, and weekday fixed effects. This tests for both level and trend changes associated with lockdown. Robustness checks: (1) Breakpoint analysis: repeat the regression assuming 21 hypothetical break dates (−10 to +10 days around 26 March) to see where the slope change peaks (structural break). (2) 2019 placebo: run analogous models over Jan–Jul 2019 with hypothetical break dates to test whether any similar mid-March trend shift occurs absent a lockdown. Audio features comparison: Compare old vs. recent charting songs on seven Spotify audio features (acousticness, danceability, energy, liveness, loudness, tempo, valence) using unequal-variance t-tests. Time-series analysis: Construct four aggregate series: (a) log difference in total plays between positive and negative songs within recent songs; (b) within old songs; (c) log difference in total plays between old and recent songs within positive songs; (d) within negative songs. After filtering out total daily plays and weekday effects and ensuring stationarity (ADF tests), estimate autoregressive models (lag order selected by BIC; typically AR(1), AR(2) for one case). Regress each series on a lockdown indicator (to capture intercept shift), pre- and post-lockdown time trends (to capture slope changes), and lags of the dependent variable. This assesses whether lockdown altered preferences for positivity and/or nostalgia within these strata.

- Lockdown and old-music listening: The logistic regression indicates a significant upward change in the trend of choosing old songs after lockdown (positive Lockdown×t). The data show an immediate shift in slope around the lockdown date, with no strong evidence of a large instantaneous jump, implying a gradual lockdown effect. - Breakpoint robustness: The estimated change in slope peaks at t≈−1 day relative to the actual lockdown (26 March 2020), consistent with a structural break caused by lockdown announcements/implementation. - 2019 placebo: Repeating the analysis for Jan–Jul 2019 yields no comparable hump-shaped slope change around the same calendar period; in 2019 the trend tends to shift toward recent songs, supporting that the 2020 effect is specific to lockdown. - Revived songs: 28 old songs (e.g., Mr Blue Sky, Dancing in the Moonlight, Don’t Stop Me Now, Dreams) increased their Top-200 chart presence from April–July relative to Jan–March 2020, typically peaking in May. - Audio features: Old charting songs differ significantly from recent ones on most features and are notably more positive on average (valence mean old 0.5829 vs. recent 0.4657; p<0.001). They are also more energetic and lively, less loud, and less danceable on average. - Time-series preference shifts (Table 4): • Within positive songs (Old−Recent): significant intercept increase at lockdown (µ2≈0.145, p<0.10) and a persistent positive post-lockdown slope (µ4≈0.006, p<0.001) → growing preference for old over recent among positive songs. • Within negative songs (Old−Recent): significant intercept increase (µ2≈0.171, p<0.05) but no significant post-lockdown slope change → a one-off shift toward old among negative songs without continued growth. • Within recent songs (Positive−Negative): significant intercept increase favoring positive (µ2≈0.101, p<0.001) followed by a significant negative post-lockdown slope (µ4≈−0.002, p<0.001) → initial positivity boost among recent songs that wanes over time. • Within old songs (Positive−Negative): no significant intercept change (µ2 not significant) but a significant positive post-lockdown slope (µ4≈0.004, p<0.001) → a gradual, sustained rise in positivity among old songs. - Overall, listening to old music increased after lockdown in both positive and negative song subsets, but the rise was more persistent among positive old songs. Positive recent music saw an initial uplift that reversed over time. Covid-19 incidence levels were also positively associated with old-music listening in the cross-sectional logistic model.

Findings indicate that the UK’s first lockdown likely prompted a behavioral shift toward nostalgic music, consistent with music serving as an emotion regulation strategy under distress. Nostalgic songs—familiar and linked to personal memories—may offer certainty, comfort, and a psychological escape during uncertainty. The analysis also shows a positivity bias during lockdown (greater preference for positive over negative songs), with nostalgia and positivity reinforcing each other: positive old music exhibited the most persistent growth. Nonetheless, increased old-music listening in both positive and negative subsets suggests nostalgia-seeking was not solely a by-product of searching for positive songs. The results contribute to understanding how large-scale societal stressors influence cultural consumption and underscore the potential of nostalgic music in supporting well-being, while recognizing individual differences in mood regulation and music preferences.

The study provides evidence that the first UK Covid-19 lockdown triggered an increase in listening to songs released at least five years earlier on Spotify, consistent with a surge in nostalgia-seeking. This pattern is not observed in the same period in 2019 and aligns with a structural break around the lockdown date. Analyses further show that nostalgia-seeking and a preference for positive music both rose, and mutually reinforced each other: positive old music displayed a more persistent increase than positive recent music. At the same time, the nostalgia effect appears partly independent of positivity, as old-music listening also increased for negative songs. Future research could integrate individual-level demographic and psychological data to directly link emotional states and nostalgia choices, investigate mechanisms by which nostalgic music supports emotion regulation, and explore cross-country or platform differences and longer-run dynamics beyond the first wave.

- Observational streaming data lack individual-level demographics and psychological measures; motives and emotional states cannot be directly inferred. - The analysis is limited to Top 200 chart data (truncation below rank 200), potentially omitting long-tail listening behavior. - Nostalgia is operationalized via an absolute age threshold (>5 years), which may not reflect individual-specific nostalgic periods. - Lockdown relaxation was gradual and not explicitly dated in the model; effects are captured via trends rather than discrete endpoints. - Positivity classification uses arbitrary valence cutoffs, albeit close to quartiles; Spotify valence is an algorithmic proxy, not a direct measure of affect. - Potential external channels (e.g., media exposure, playlists, algorithmic promotions) that could influence old song discovery are not observed. - Generalizability is limited to UK Spotify usage and the early 2020 period.