Trajectories of affective and cognitive well-being at times of COVID-19 containment policies in Italy

Abstract

This paper draws on a subsample (N=851) of respondents to ITA.LI - Italian Lives – a recently established panel study on a probability sample of individuals aged 16+ living in Italy – to track changes in the affective (positive and negative emotions such as energy and sadness) and cognitive (life satisfaction) components of well-being during different COVID-19 policy phases, classified according to the severity of key government responses.

An event-study design is employed, which uses mixed-effects ordered logistic models to investigate the change in SWB scores. Given the nested nature of the data, multilevel modelling is chosen as the most appropriate method of analysis.

The results reveal the levels of affective and cognitive well-being were significantly lower during the lockdown period than before the pandemic outbreak potentially reflecting both the direct effects of the confinement and other potential sources of distress, such as trends in infection rates and related media alarm. Once the lockdown was lifted, there was no evidence of an immediate and general improvement in well-being. In the following policy phase, with the lifting of most containment measures, there were significant signs of full recovery concerning energy, but the scores for the other well-being components remained relatively lower than those observed before the onset of COVID-19.

Introduction

Restrictive measures introduced by national and local governments to reduce the spread of SARS-CoV-2 (hereafter COVID-19) infection have led to a drastic disruption of “lives as usual”, with potential unintended consequences on health and well-being. For instance, the implementation of lockdown has been linked to a significant deterioration in mental health (Banks and Xu, 2020; Pierce et al., 2021; Lucchini et al., 2021) and sleep quality (Alimoradi et al., 2021). Regarding the change in subjective well-being (SWB) before and during the COVID-19 emergency, overall stability was reported in cognitive measures, such as life satisfaction, while some variability was observed in indicators assessing the frequency of positive and negative emotions (Ebert et al., 2020; Fujiwara et al., 2020; Helliwell et al., 2020; Schmidtke et al., 2021). That said, an important and still overlooked issue concerns how SWB evolved across the pandemic, during confinement and beyond. Studies investigating the trajectories of SWB at times of COVID-19 have generally found strong and immediate negative effects on its affective and cognitive components, with subsequent improvement once the immediate dangers were brought under control and the situation stabilized (Aknin et al., 2021; Cielo et al., 2021). However, due to the cross-sectional nature of several datasets employed, the findings are not straightforward. SWB improvement was reported either from the very beginning of lockdown (Foa et al., 2020; Recchi et al., 2020) or later, when containment measures were lifted (Cheng et al., 2020; Layard et al., 2020; Pierce et al., 2021; Richter et al., 2021; Schmidtke et al., 2021; Zacher and Rudolph, 2021; Zhou and Kan, 2021). Hence, it is difficult to assess the timing and overall pattern of the SWB recovery.

Against this background, this manuscript draws on a subsample (N=851) of respondents to ITA.LI - Italian Lives (hereafter ITA.LI) – a recently established panel study on a probability sample of individuals aged 16+ living in Italy – to track changes in the affective (positive and negative emotions such as calm, energy, and sadness) and cognitive (life satisfaction) components of well-being in Italy at different COVID-19 policy phases. To the best of our knowledge, this study is among the first to use panel data to examine the trajectories of SWB during the pandemic, in Italy and internationally. Italy is a relevant case study: it was the first country, after China, to be severely affected by COVID-19, with containment measures introduced very early (late February 2020). Most evidence collected in the country used convenience and cross-sectional sampling to assess the impact of lockdown and related containment measures (e.g., Mazza et al., 2020; Rossi et al., 2020; Epifanio et al., 2021). However, these studies were unable to assess the nature and magnitude of changes in SWB associated with the severity of key government responses. Furthermore, this manuscript does not use a composite index of SWB but relies on separate measures of positive and negative affect and life satisfaction. This may prove useful to policymakers and practitioners, as long as the analyses provide evidence of the extent to which different components of well-being were more or less sensitive or resilient to the pandemic and key policy responses.

Literature review and research hypotheses

According to a widely accepted definition, SWB is a self-reported measure of well-being consisting of distinct components (Diener and Emmons, 1984; Diener et al., 1999): an affective component, which refers to the experience of pleasant (e.g., joy, calm, happiness, and contentment) and unpleasant (e.g., sadness, anxiety, and anger) emotions; and a cognitive component, which is based on an evaluative account of one’s life relative to an ideal state and is usually assessed through measures of overall life satisfaction and satisfaction with specific life domains, such as work, health, family life. By definition, the affective component of SWB tends to capture the short-term emotional reaction to specific events or circumstances (Diener et al., 1999; Kahneman et al., 1999) and is usually more volatile than satisfaction judgments (e.g., Eid and Diener, 2004). Research conducted during the COVID-19 emergency confirmed this argument (Aknin et al., 2021; Cielo et al., 2021). Studies on affective well-being consistently indicate an increase in the frequency of negative emotions throughout the pandemic and a concomitant decline in the frequency of positive feelings and moods (Fujiwara et al., 2020; Schmidtke et al., 2021; Wettstein et al., 2021; Zacher and Rudolph, 2021). However, regarding the cognitive component, evidence of resilience in life satisfaction was collected in cross-sectional studies (Helliwell et al., 2021) and panel data studies (Kivi et al., 2021; Schmidtke et al., 2021; Wettstein et al., 2021). Accordingly, we may anticipate (Hypothesis 1) negative COVID-related consequences on affective well-being and a certain degree of stability in cognitive well-being.

Several studies have been conducted on the change in SWB following the COVID-19 outbreak. On the one hand, evidence suggests that the decline in SWB scores was associated with concerns about the infection’s spread or the pandemic’s severity further reinforced by media alarm, especially in the first months after the pandemic outbreak. For instance, Recchi and colleagues (2020), who collected longitudinal data in France, found that shortly after the implementation of a nationwide lockdown, sampled individuals reported a rapid improvement in affective well-being, assessed by a composite index combining positive and negative emotions. The authors linked this improvement in well-being scores to an “eye of hurricane” paradox: people who had not been infected tended to perceive their lives more optimistically than before the pandemic. Similarly, Foa and colleagues (2020), whose analyses used weekly cross-sectional data collected from a large sample in the UK, noticed that the magnitude of negative change in life satisfaction, and positive and negative moods decreased significantly once the national government implemented lockdown measures. On the other hand, evidence suggests that the implementation of containment measures was the main reason for reduced well-being in its various components. For instance, Anglim and colleagues (2021) observed that well-being scores declined significantly after weeks of confinement compared to pre-COVID levels. In particular, there was a deterioration in items assessing the experience of pleasant and unpleasant feelings; change in life satisfaction was comparatively smaller but significant. Research conducted in Europe and the USA by Brodeur and colleagues (2020) further proved that restrictions were associated with a preponderance of unpleasant feelings, such as worry and sadness.

More recent empirical work has focused on SWB trajectories and investigated patterns and predictors of resilience: the ability to cope with and adapt to a new situation. In particular, increasing attention has been paid to disentangling the effects that the implementation and easing of policy measures to control the spread of COVID-19 may have had on the different components of SWB. Most research on the consequences of COVID-related containment measures suggests that policy severity has generally been associated with lower levels of affective and cognitive well-being, and that any easing of restrictions could lead to significant improvements (Banks et al., 2021; Cheng et al., 2020; Ebert et al., 2020; VanderWeele et al., 2020; Zacher and Rudolph, 2021). Available literature also indicates temporal variation in adaptation and different recovery patterns for affective well-being items compared to cognitive ones (Clark and Lepinteur, 2021; Schmidtke et al., 2021). Hence, we expect (Hypothesis 2) a general decline in SWB during the early stages of the pandemic, following the implementation of lockdown measures, and a U-shaped recovery thereafter, with SWB scores returning or approaching pre-COVID-19 baseline level after significant relaxation of restrictions. Furthermore, we may anticipate (Hypothesis 3) different patterns in the recovery trends of affective and cognitive components.

Data and Methods

Sample

Analyses draw on a subsample of respondents to ITA.LI - Italian Lives. ITA.LI¹ is an ongoing panel study based on a probability sample of individuals aged 16+ in Italy. The first wave of data collection (May 2019 - December 2020) gathered information on household structures and dynamics, education and employment, income and wealth, health, and well-being. From April to September 2020, all panel members who participated in ITA.LI wave 1 were invited to complete, using CAWI and CATI methods, an ad-hoc survey (hereafter ITA.LI COVID-19), which collected data on the psychological and socioeconomic consequences of the pandemic using items and measures that were already employed in ITA.LI wave 1. The response rate to ITA.LI COVID-19 was 39.3% (AAPOR response rate 1). Data collected from ITA.LI wave 1 was merged with the records of the respondents to ITA.LI COVID-19. Observations that were not matched were dropped. Moreover, respondents with missing values on any of the variables included in the models were omitted from the analysis. The final sample for this study consisted of 851 individuals (1,702 observations)². The sample profile is displayed in Table 1.

Table 1.

Sample (N= 851) and Weighted sample (N=5281) profiles

	Sample		Weighted sample
	N	Freq. (%)	N	Freq. (%)
Sex
Male	340	40.0	2372	44.9
Female	511	60.0	2909	55.1
Age
16–34	141	16.6	1009	19.1
35–54	324	38.1	1821	34.5
55+	386	45.4	2451	46.4
Level of education
Primary	268	31.5	2279	43.2
Secondary	434	51.0	2370	44.9
Tertiary	149	17.5	631	12.0
Cohabiting with partner/spouse
No	364	42.8	2391	45.3
Yes	487	57.2	2890	54.7
Children living in the household by age group
No children aged 0–14	702	82.5	4410	83.5
Pre-schoolers	71	8.3	432	8.2
7–14 years	78	9.2	439	8.3
Employment status
ITA.LI wave 1
Employed	424	49.8	2432	46.1
Unemployed	83	9.8	439	8.3
Inactive	344	40.4	2409	45.6
ITA.LI COVID-19
Employed	387	45.5	2124	40.2
Unemployed	123	14.5	764	14.5
Inactive	341	40.1	2393	45.3
Macro-region
North-West	240	28.2	1372	26.0
North-East	181	21.3	1086	20.6
Centre	173	20.3	1063	20.1
South	257	30.2	1759	33.3
Big Five	N	Mean	N	Mean
Agreeableness	851	11.3	5281	11.3
Extraversion	851	10.3	5281	10.4
Conscientiousness	851	11.7	5281	11.6
Neuroticism	851	9.1	5281	9.2
Openness	851	11.3	5281	11.2

Variables

Dependent variable: SWB.

Affective well-being was measured with two items asking how often in the last 4 weeks the respondent had felt “full of energy” (positive affect) or “sad and downhearted” (negative affect) on a 6-point scale, ranging from 1 “None of the time” to 6 “All of the time”. Overall life satisfaction (cognitive SWB) was measured on an 11-point scale ranging from 0 “Not at all satisfied” to 10 “Completely satisfied”.

Independent variables: Time.

To examine the impact of COVID-related policy phases and control for seasonal patterns in SWB, we used the data collection date and identified different periods. Observations collected in ITA.LI wave 1 (pre-pandemic period) were grouped in the following time categories: 1) 20 June - 31 August 2019; 2) 1 September - 31 October 2019; 3) 1 November - 31 December 2019; 4) 1 January - 8 March 2020. Observations collected in ITA.LI COVID-19 (pandemic period) were grouped into the following time categories, based to the severity of key national government responses: 5) 9 March - 3 May 2020 (first nationwide lockdown); 6) 4 May - 2 June 2020 (so-called Phase 2, when several containment measures were relaxed and retail shops, bars, restaurants, and personal services reopened under COVID-19 safety guidelines); 7) 3 June - 6 August 2020 (summer reopening, when most mandatory COVID-19 regulations, such as mobility restrictions in Italy and EU countries, were gradually lifted or removed). Time categories in the pandemic period were created based on the Decrees of the President of the Council of Ministers containing urgent measures for the containment and management of the epidemiological emergency.

Covariates.

Analyses were adjusted for a wide range of socio-demographic, household, and socioeconomic variables that the reference literature has found to be correlated with overall subjective well-being (Dolan et al., 2008; Diener et al. 2018) and, thus, potential confounding factors that can either increase or mitigate the consequences of COVID-19 on respondents’ well-being: gender (Batz and Tay, 2018); age at the time of the interview (16–34; 35–54; 55 and older); education level (primary, secondary, and tertiary) (Lansford, 2018); employment status (employed, unemployed, inactive - students, pensioners, homemakers) (Ford et al., 2018); cohabitation with partner/spouse (no, yes) (Nelson-Coffey, 2018); children living in the household by age (no children aged 0–14; pre-schoolers; children aged 7–14); personality traits (agreeableness, extraversion, conscientiousness, neuroticism, openness) measured with an Italian adaptation of the Big Five Inventory - Short version (BFI-S) and treated as time-constant correlates of well-being levels (Lang 2011; Lucas, 2018). Finally, we control for the macro region of residence (North-East, North-West, Center, South), which, given the uneven spread of the virus with the northern regions experiencing a higher concentration of cases especially in the early months of the pandemic, may also have been associated with heterogeneous well-being effects.

Analytical strategy

As already stated, this study aims to trace the change in the affective and cognitive components of well-being in Italy through the different COVID-related policy phases. Accordingly, an event-study design was employed, which used mixed-effects ordered logistic models – known in psychometrics as graded response models (Samejima, 1969) – to investigate the change in SWB scores. Given the nested nature of the data – repeated measurements, one before and one during the COVID-19 pandemic, within individuals – multilevel modelling was chosen as the most appropriate method. The variation in the date of the interviews, before and during the COVID-19 pandemic, was exploited to better identify the effects of the pandemic-associated containment policies, describe time trends and control for seasonal effects. In particular, rather than simply comparing the levels of well-being before and after the outbreak of the pandemic and during the different policy phases, we consider a detailed picture of 7 time periods (4 before and 3 after the health emergency) covering the entire calendar year, in order to also capture potential seasonal trends in subjective well-being assessments.

Coefficients of the random-effects ordered logistic models³ do not offer a direct interpretation. Therefore, the coefficients were transformed into odds ratios and marginal effects. Based on the regression model equation, we could interpret the factor variation in the odds of higher versus lower outcomes. In other words, we could estimate and assess whether, for a change in a given time category, the odds of reporting higher SWB scores were greater (OR>1), lower (OR<1), or equal (OR=1) than for the remaining policy phases, holding all other variables constant. However, odd ratios do not assess the magnitude of change in SWB scores. Consequently, we also computed and reported marginal effects, which contribute to grasping how large the effects of the change in the policy stage in terms of the change in the probability of reporting certain SWB scores.

The analyses used non-response weights to consider any unbalanced selection probability in the survey due to the low response rate. Since missing outcomes are unlikely to occur randomly, selection bias was corrected based on age, sex, employment status, education, and macro-region of residence (Mansournia and Altman, 2016).

Findings

The summary statistics, displayed in Table 2 below, reveal that average SWB deteriorated with the pandemic outbreak. Specifically, during the first lockdown period in March 2020, the percentage of people feeling downhearted and blue all or most of the time more than doubled from 3.7% to 8.2%, compared to before COVID-19. Over the same period, the percentage of individuals who reported they had a lot of energy all or most of the time declined from 45.7% to 33.7%. The prevalence of respondents who stated they were very satisfied (at least 8 out of 10) with their lives decreased from 57.9% to 39.8%.

Table 2.

Summary statistics of the outcome variables, by time (COVID-19 policy phases)

Phase	20 Jun - 31 Aug 19	01 Sep - 31 Oct 19	01 Nov - 31 Dec 19	01 Jan - 08 Mar 20	09 Mar - 03 May 20	04 May - 02 Jun 20	03 Jun - 26 Aug 20
Energy
1 None of the time	0.61	1.00	0.99	1.06	2.04	3.28	1.11
2 A little of the time	4.27	7.46	4.93	4.59	9.18	13.93	9.67
3 Some of the time	27.44	25.37	24.63	26.5	38.78	42.62	27.26
4 A good bit of the time	21.95	23.38	20.2	21.55	16.33	13.11	9.19
5 Most of the time	27.44	32.34	38.42	31.45	26.53	20.49	39.3
6 All of the time	18.29	10.45	10.84	14.84	7.14	6.56	13.47
Median	4	4	4	4	3	3	5
Sadness
1 None of the time	26.22	27.36	18.72	24.73	1.02	4.10	23.45
2 A little of the time	46.95	45.77	53.2	50.88	38.78	27.05	39.46
3 Some of the time	20.73	19.90	18.72	17.67	32.65	54.1	25.52
4 A good bit of the time	2.44	1.99	4.93	3.18	14.29	6.56	3.80
5 Most of the time	3.66	3.48	3.45	2.83	9.18	4.92	6.02
6 All of the time	0.00	1.49	0.99	0.71	4.08	3.28	1.74
Median	2	2	2	2	3	3	2
Life satisfaction
0 Not at all satisfied	0.00	0.00	0.49	0.35	2.04	1.64	0.95
1	0.00	1.00	0.49	0.00	1.02	0.82	0.32
2	0.00	0.50	0.00	0.00	2.04	1.64	0.63
3	0.00	1.00	1.48	1.41	2.04	3.28	0.63
4	0.00	1.00	1.48	0.35	2.04	0.82	1.74
5	2.44	4.48	3.94	5.30	15.31	15.57	5.86
6	11.59	12.44	6.90	8.48	13.27	17.21	10.94
7	28.05	19.40	19.21	23.32	22.45	23.77	26.78
8	34.76	38.81	46.80	39.58	34.69	25.41	25.52
9	12.80	14.93	12.32	14.84	3.06	6.56	13.31
10 Completely satisfied	10.37	6.47	6.90	6.36	2.04	3.28	13.31
Median	8	8	8	8	7	7	8

In the first six months of the pandemic, there were signs of SWB deterioration during the nationwide lockdown and little change in the subsequent phase of initial relaxation of containment measures. As of June 2020, the percentage of individuals reporting high energy all or most of the time peaked at 52.8% surpassing pre-pandemic levels; the prevalence of sadness (all or most of the time) was 7.8%, still higher than before the onset of COVID-19, while 52.1% of respondents, still fewer than before the pandemic, reported high levels of life satisfaction.

To test Hypothesis 1, we applied random-effects ordered logistic models that estimated the probability of reporting higher SWB scores within the first six months of the pandemic compared to pre-pandemic levels, after adjusting for a set of control variables. The slopes coefficients (Table 3) show that the odds of reporting higher levels of energy and life satisfaction were, respectively, 0.80 and 0.74 times lower after the onset of the pandemic. Similarly, the odds of reporting sadness were 1.92 times larger under COVID-19 than before⁴. Accordingly, Hypothesis 1 is confirmed: following the COVID-19 onset, affective and cognitive well-being scores deteriorated significantly.

Table 3.

Random-effects (RE) and fixed-effects (FE) ordered logistic models on SBW (Odds Ratio) - Pre-Post COVID-19 comparison

	Energy		Sadness		Life Satisfaction
	RE	FE	RE	FE	RE	FE
Time
(Ref. cat. 20 Jun - 08 Mar 20)
09 Mar - 26 Aug 20	0.798***	0.785**	1.918***	2.010***	0.743***	0.711***
	(0.029)	(0.079)	(0.073)	(0.218)	(0.027)	(0.074)

Standard errors in parentheses;

^***p<0.001,

^**p<0.05,

^*p<0.01

To test Hypothesis 2 and Hypothesis 3 concerning SWB trajectories after the onset of the pandemic, we fitted random-effects ordered logistic models that estimated the probability of reporting lower SWB scores following the implementation of containment measures and higher SWB scores after their relaxation. We used data collection dates after the start of the lockdown as a proxy for COVID-related policy phases and a pre-pandemic baseline (20 June - 31 August 2019). The results, adjusted for selected control variables, are reported in Table 4 above and displayed, as odds ratios, in Figure 1. To work out the variation in SWB scores, we computed the average marginal effects and estimated the predicted probabilities of each category of the outcome variables with respect to time or policy phases, keeping other variables at their current values. The predicted average can be interpreted as the probability that individuals in the sample belong to a certain category, that is the prevalence. To save space, only the predicted probabilities for the different policy phases within the six months of COVID-19 are reported (Table 5).

Table 4.

Random-effects (RE) and fixed effects (FE) ordered logistic models

	Energy		Sadness		Life Satisfaction
	RE	FE	RE	FE	RE	FE
Time
(Ref. cat. 20 Jun - 31 Aug 19)
01 Sep - 31 Oct 19	0.689***	0.842	1.027	1.158	0.675***	0.450***
	(0.058)	(0.251)	(0.091)	(0.391)	(0.058)	(0.138)
01 Nov - 31 Dec 19	0.847**	0.933	1.292***	1.261	0.981	0.651
	(0.071)	(0.273)	(0.114)	(0.441)	(0.084)	(0.203)
01 Jan - 08 Mar 20	0.931	1.052	0.994	0.834	0.803***	0.841
	(0.073)	(0.292)	(0.081)	(0.269)	(0.063)	(0.254)
09 Mar - 03 May 20	0.410***	0.372***	6.519***	9.609***	0.312***	0.197***
	(0.043)	(0.125)	(0.704)	(5.042)	(0.033)	(0.076)
04 May - 02 Jun 20	0.221***	0.098***	6.945***	14.600***	0.286***	0.184***
	(0.022)	(0.040)	(0.717)	(7.742)	(0.029)	(0.074)
03 Jun - 26 Aug 20	0.910	1.040	1.386***	1.388	0.805***	0.650*
	(0.063)	(0.226)	(0.100)	(0.357)	(0.056)	(0.153)

Standard errors in parentheses;

^***p<0.001,

^**p<0.05,

^*p<0.01

Figure 1.

Random-effects ordered logistic models on affective and cognitive well-being of time (or COVID-19 policy phase) with 95% confidence intervals (odds ratios)

Reference category: 20 June 2019 – 31 August 2019

Table 5.

Average marginal effects of time (COVID-19 policy phases) on energy, sadness and life satisfaction

Policy phase	09 Mar - 03 May 20	04 May - 02 Jun 20	03 Jun - 26 Aug 20
(Ref. cat. 20 Jun - 31 Aug 19)
Energy
1 None of the time	0.011*** (0.002)	0.026*** (0.003)	0.001 (0.001)
2 A little of the time	0.062*** (0.008)	0.125*** (0.01)	0.005 (0.004)
3 Some of the time	0.104*** (0.011)	0.150*** (0.009)	0.012 (0.009)
4 A good bit of the time	0.003 (0.002)	−0.015*** (0.003)	0.002 (0.002)
5 Most of the time	−0.098*** (0.012)	−0.172*** (0.011)	−0.009 (0.006)
6 All of the time	−0.082*** (0.009)	−0.115*** (0.008)	−0.011 (0.008)
Sadness
1 None of the time	−0.201*** (0.012)	−0.205*** (0.012)	−0.050*** (0.012)
2 A little of the time	−0.161*** (0.012)	−0.170*** (0.01)	−0.007*** (0.001)
3 Some of the time	0.156*** (0.008)	0.157*** (0.008)	0.036*** (0.008)
4 A good bit of the time	0.068*** (0.005)	0.070*** (0.005)	0.009*** (0.002)
5 Most of the time	0.094*** (0.007)	0.099*** (0.007)	0.009*** (0.002)
6 All of the time	0.045*** (0.005)	0.048*** (0.005)	0.003*** (0.001)
Life satisfaction
0 Not at all satisfied	0.008*** (0.001)	0.009*** (0.001)	0.001*** (0.000)
1	0.005*** (0.001)	0.006*** (0.001)	0.001*** (0.000)
2	0.005*** (0.001)	0.006*** (0.001)	0.001*** (0.000)
3	0.015*** (0.002)	0.016*** (0.002)	0.002*** (0.001)
4	0.011*** (0.002)	0.013*** (0.002)	0.001*** (0.000)
5	0.062*** (0.006)	0.068*** (0.006)	0.009*** (0.003)
6	0.070*** (0.006)	0.075*** (0.006)	0.013*** (0.004)
7	0.053*** (0.005)	0.053*** (0.005)	0.015*** (0.005)
8	−0.084*** (0.009)	−0.093*** (0.008)	−0.009*** (0.003)
9	−0.073*** (0.006)	−0.077*** (0.006)	−0.015*** (0.005)
10 Completely satisfied	−0.072*** (0.007)	−0.076*** (0.007)	−0.018*** (0.006)

Standard errors in parentheses;

^***p<0.001,

^**p<0.05,

^*p<0.01

Figure 1 shows that similar trajectories were found in the first six months of the COVID-19 pandemic in the different SWB components. In particular, affective and cognitive well-being scores declined during lockdown compared to the baseline period. Distress continued in the following period when confinement ended and the partial easing of other restrictive measures began. Only in the policy phase of summer reopening, when restrictions were further lifted or removed, SWB scores improved.

Regarding positive affect, the odds of reporting higher energy scores were 0.41 times lower during the lockdown and 0.22 times lower in Phase 2 than in summer 2019. We also interpreted the random-effects ordered logistic model by computing the marginal change in energy with respect to the different policy phases. Compared to summer 2019, the predicted probability of reporting a lot of energy all or most of the time was 17.9% lower during the lockdown and 28.6% lower following the initial easing of containment measures (Table 5). However, in the subsequent period, average energy scores significantly improved and fully recovered; indeed, the odds of reporting higher (or lower) energy were not significantly different from baseline. We further explored the effects of each policy phase on energy and the results of reverse adjacent contrasts indicated that each difference was significant. Hence, we may conclude that the positive affect component of SWB significantly declined after the enactment of a nationwide lockdown and deteriorated further in the following period but improved and bounced back to pre-pandemic levels during the summer reopening policy phase.

Concerning negative affect, the odds of reporting higher levels of sadness were 6.52 and 6.95 times higher during the confinement and Phase 2, respectively. During the summer, the odds of feeling downhearted were still 1.39 times higher than in the summer 2019 and, more generally, before the onset of the pandemic. Compared to baseline, the prevalence of sadness – as measured by the predicted probability of individuals reporting feeling downhearted and blue all or most of the time – increased by 13.9% during the nationwide lockdown and by 14.7% in Phase 2; however, during the summer the predicted probability of reporting high levels of sadness was just about one percentage point higher than a year before (Table 5). Contrasts of parameter estimates revealed that the effect of Phase 2 on sadness was not significantly different from that of the preceding policy phase; furthermore, sadness scores during the summer reopening policy phase were significantly lower than those reported during the previous phase, indicating a pattern of resilience, yet below pre-pandemic levels.

Regarding the cognitive component of SWB, the odds of reporting higher levels of life satisfaction were 0.31 times lower during the nationwide lockdown than during summer 2019. With the end of confinement, in Phase 2, the odds of reporting higher levels of life satisfaction were 0.29 times lower than at baseline. Life satisfaction levels improved further during the summer of 2020 but the recovery was not complete: the odds of reporting higher levels of life satisfaction were 0.81 times smaller than a year before. Compared to the baseline, the prevalence of individuals rating their life as very satisfactory (8 or above) decreased by 22.9% during the lockdown, by 24.6% in Phase 2, and by 4.2% in the policy phase of summer reopening (Table 5). Post-estimation contrast tests indicate that average life satisfaction declined at the beginning of the pandemic, remained stable and lower than before the pandemic at the end of confinement, and did not improve until the summer of 2020, when it increased but was still below pre-pandemic levels.

In summary, Hypothesis 2 was partially supported: affective and cognitive well-being worsened at the beginning of the pandemic, did not benefit from the partial easing of policy measures, and only started to improve in summer 2020. Hypothesis 3 was confirmed: evidence of different patterns emerged in affective and cognitive SWB items, with complete recovery of positive affect but incomplete recovery in negative emotion and life satisfaction scores.

We also estimated the effect of time (or COVID-19 policy phases) on SWB scores using fixed-effects ordered logistic modelling, which could account for any potential endogeneity arising from time-invariant individual characteristics (Table 3 and Table 4). The Wald test indicated that all variables included in the models were jointly statistically significant. Odds ratios obtained for easier interpretation of the effect sizes confirmed the results of random-effects ordered logistic modelling and proved that positive and negative affect and satisfaction with life worsened during the national lockdown period. Lower well-being levels persisted in the following policy phase for all SWB indicators, while in summer 2020 it reached a non-significant level for energy and remained negative for sadness and life satisfaction⁵.

Discussion and conclusion

Since the outbreak of COVID-19, a mounting body of research has considered how people responded to pandemic-related concerns and uncertainty in multiple life domains, as well as the imposition of non-pharmaceutical interventions to control the virus spread (Helliwell et al., 2021; Samios et al., 2022; Satici et al., 2021). In this regard, a crucial, still overlooked issue is to understand the unintended consequences of COVID-related containment measures and detect recovery patterns in health and SWB outcomes (Cheng et al., 2020; Foa et al., 2020; Layard et al., 2020; Zacher and Rudolph, 2021; Pierce et al., 2021; Richter et al., 2021; Schmidtke et al., 2021). Against this backdrop, this study is among the first to use a longitudinal sample with pre-pandemic data as a baseline to assess the short-term pandemic-related consequences on affective and cognitive well-being in Italy and to understand how people experienced and evaluated their lives and coped with increasing adversities during the different containment policy phases. To provide robust estimates and exclude that the observed changes in SWB occurred following regular patterns, we controlled for potential seasonal effects (Banks and Xu, 2020).

Consistent with the first research hypothesis, the results confirmed previous international evidence (Banks et al., 2021: Cheng et al., 2020; Schmidtke et al., 2021; Zacher and Rudolph, 2021) and indicated that average SWB worsened compared to pre-pandemic levels. Indeed, individuals experienced a significant decline in affective and cognitive well-being. Furthermore, the findings showed that SWB varied significantly across policy phases. In particular, cognitive and affective well-being declined with the implementation of a nationwide lockdown strategy. Contrary to expectations, recovery did not begin soon after its release, with further deterioration recorded for negative and positive affect during the first phase of reopening, potentially indicating accumulated fatigue.

However, in the longer run (summer 2020), with the easing of restrictions, SWB improved and returned around pre-pandemic level – as in the case of sadness and life satisfaction – or rebounded completely - as in the case of energy. These fluctuating trajectories in SWB can be explained by the human tendency to adapt to changes and the ability to tolerate or cope with adversity. It has been argued that as people move on after a traumatic event, they instinctively tend to make sense of it and, in this process of rationalization and adaptation, their initial emotional and cognitive reactions fade away, leading to weaker and weaker negative effects on SWB (Helson, 1948, 1964; Frederick and Loewenstein, 1999; Wilson and Gilbert, 2008). This process can also be seen as a return to a stable, largely heritable, neutral level of SWB that is referred to as a set point, at which individuals tend to bounce back after positive and negative life events (Lykken and Tellegen, 1996; Bartels, 2015; Nes and Røysamb, 2017). However, adaptation does not always happen, and some circumstances or recurring events may have a more lasting impact (Lucas, 2007; Clark et al., 2008; Luhmann et al., 2012; Yap et al., 2014). In this respect, only recently has literature started investigating trends in health and well-being throughout the different waves of COVID-19. For instance, Pierce and colleagues (2021) studied a sample of individuals across the first 6 months of the pandemic (April- October 2020), finding that a composite measure of mental health, distress, and well-being (GHQ-12) worsened with the COVID-19 outbreak and started to improve in July, although by October it was still below pre-pandemic levels. Zhou and Kan (2021) extended these analyses, finding that, after the enforcement of a second nationwide lockdown in November 2020, GHQ-12 scores declined to even lower levels than those recorded during the first lockdown period. Schmidtke and colleagues (2021), who examined a sample of workers in Germany, noticed partial adaptation between the lockdown periods so that containment measures during the second wave of COVID-19 had smaller effects on the different well-being indicators than the first wave of restrictive policy. However, this adaptation was more pronounced for cognitive than for affective SWB, confirming the importance of distinguishing these well-being dimensions.

To get a clear picture of the actual levels of SWB adaptation in the face of the new circumstances posed by the pandemic-related containment measures, it would be important to analyse its response also during later waves of the health emergency. However, due to the nature of the available data, it was impossible to track the changes in SWB over a longer period and verify whether the recovery pattern persisted despite the subsequent partial or full re-imposition of containment measures during successive waves of COVID-19 in Italy. This is a limit of this study that needs to be acknowledged.

Another limitation of the study is that, given the available data, we were unable to disentangle the lockdown-related consequences on respondents’ well-being levels and other potential effects of the trend in infection rates and related media alarm, especially in the first months after the outbreak of the pandemic. Moreover, due to the limited sample size, we could not estimate the heterogeneity of the COVID-19 effects on different subgroups: it would indeed be important to look behind the trajectories of average well-being levels, which could conceal significant heterogeneity in the intensity and persistence of the impact of the pandemic between different subpopulations. For instance, previous studies have shown heterogeneity based on gender and age (Helliwell et al., 2021; Ebert et al., 2020; Dawel et al., 2020; Schlomann et al., 2021; Banks and Xu, 2020; Clark and Lepinteur, 2021). Furthermore, future research, especially longitudinal one, may consider assessing the medium- and long-term effects of the pandemic, disentangling those related to containment measures from those related to its economic and social consequences as they unfold. Moreover, the evidence of persistently lower well-being levels even once the lockdown has been eased underlines the importance of assessing the effects of the pandemic on well-being that extend beyond the period of strict containment measures to subsequent phases aimed at gradually restoring normality to economic and social life. The well-being cost of the pandemic should therefore enter into the welfare calculus in the acute stages of the pandemic and when policymakers design interventions in the later phases of reopening, once infection rates begin to decline and life returns to normal. This is further supported by the fact that the way in which people live and evaluate their lives can have profound implications in terms of behaviour and life choice (think of the Great Resignation phenomenon that emerged in the aftermath of the pandemic). The affective SWB measures, which tend to grasp the short-term emotional reactions to different life circumstances, and the evaluative SWB component, which assesses people’s overall life satisfaction, can provide a kind of thermometer for overall changes in well-being during the pandemic and beyond. In this way, SWB indicators can enable policymakers to obtain valuable evidence on people’s perceived well-being and be better informed in the design of interventions able to balance concerns for their physical health and economic security with considerations of their overall quality of life.

Biographies

Egidio Riva is Professor of Social Policy and Sociology of work and employment at the University of Milano-Bicocca. His research is mainly focused on the influence of the work-life interface and working conditions on health and well-being. He is currently investigating the socio-economic impact of COVID-19 containment policies.

Mario Lucchini: is full Professor of Sociology and Research Methods at the University of Milano-Bicocca. His research is mainly focused on poverty, health, subjective well-being, and social inequality.

Marta Pancheva is a post-doctoral research fellow at the SOPHIA University Institute (Florence, Italy). She holds a PhD in Science of Civil Economy from LUMSA University (Rome, Italy) and has collaborated, among others, with the University of Milano-Bicocca as part of the longitudinal research project Italian lives (ITALI). Her research focuses on ‘non-traditional’ welfare measures (such as indicators of subjective well-being), with a particular interest in aspects concerning the measurement of subjective well-being from a multidimensional perspective, its underlying dynamics, integration into multivariate regression models, and relationship to social and relational aspects of human life.

Carlotta Piazzoni is a post-doctoral research fellow at the University of Milano-Bicocca (Milan, Italy). She earned a PhD in Analysis of Social and Economic Processes from the University of Milano-Bicocca and she has collaborated with the University of Milano-Bicocca as part of the longitudinal research project Italian lives (ITALI). Her research mainly focuses on individual health and subjective well-being.

Dean Lillard is a Professor in the Department of Human Sciences at The Ohio State University in Columbus, Ohio, USA. He earned a PhD in Economics from the University of Chicago. He is a research fellow at the German Institute for Economic Research in Berlin, Germany, and a research associate of the National Bureau of Economic Research. He directs and manages the Cross-National Equivalent File Project – a compendium of harmonized panel data from twelve (soon to be eighteen) countries. He also directs an NIH-funded project to study the economic and social consequences of COVID-induced government mitigation policies across eleven countries.

Data availability:

in accordance with EU General Data Protection Regulation (EU 2016/79), University of Milano-Bicocca internal regulations (D.R. 6256/2018, prot. 90980/18) ITA.LI – Italian Lives data are stored anonymously by UniData - Bicocca Data Archive. ITA.LI – Italian Lives wave 1 data are currently available only to researchers working at the Department of Sociology and Social Research at the University of Milano-Bicocca, while ITA.LI COVID-19 data are currently available only to researchers working at the ITA.LI – Italian Lives project. As already stated at the time of first submission, the same data will be publicly available in due time. However, the data release policy (time of public release, details of how to apply for and access the datasets, end user license, etc.) has not been formally defined yet. Due to this lack of formal policies and procedures, all data underlying the findings may be currently accessed, only for the purpose of reproducing the analyses, through the corresponding author (or any of the remaining authors) and the personal data protection officer at the University of Milano-Bicocca. The personal data protection officer can be contacted (at rpd@unimib.it or certified email rpd@pec.unimib.it) for all queries concerning personal data processing and the exercise of any rights deriving from the General Data Protection Regulation (EU 2016/79). ITA.LI – Italian Lives Data Controller is the University of Milano-Bicocca, represented by its legal representative, Rector Giovanna Iannantuoni (rettorato@unimib.it or certified email ateneo.bicocca@pec.unimib.it). All relevant materials that may be reasonably requested by others to reproduce the results will be made available upon the publication of the study.