The Psychological Effect of COVID-19 and Lockdown on the 1 Population: Evidence from Italy 2

The COVID-19 pandemic and the lockdown measures adopted to prevent the spread of the disease 26 had a huge impact on a personal, social, and economic level for the world population. In Europe, 27 Italy was one of the frontrunner countries dealing with an emergency that significantly affected 28 people’s lives. Previous research on the psychological impact of the pandemic revealed an increase 29 in anxiety, depression, and feelings of distress; however, these studies were conducted on non- 30 representative samples of the population reached through social media channels, a method that is 31 likely to lead to many forms of statistical and methodological bias. For the first time to our 32 knowledge, we measured depressive symptoms on 6,700 Italian individuals, representative of the 33 Italian population in terms of age, gender, and geographical areas revealing higher scores of 34 depressive symptoms in females, younger adults, people reporting professional uncertainty and 35 lower socio-economic status. A positive correlation was also found for individuals living alone, 36 those who could not leave home for going to work, and people with a case of COVID-19 in the 37 family, whereas the region of residence was not a significant predictor of depressive symptoms. 38 These findings underline the importance of considering the psychological effects of COVID-19 and 39 providing support to individuals seeking mental health care. 40


Introduction
Method 120 To evaluate the effect of the pandemic on mood and feelings of the Italian population a 121 psychometric approach was adopted. Due to the special nature of the period which prevented to 122 conduct experimental studies meeting participants face-to-face, a psychometric self-reporting 123 methodology was chosen. We adopted the short version of the Mood and Feelings Questionnaire 124 (SMFQ; Messer et al., 1995) which includes 13 items indicating how much individuals have felt or 125 acted depressed during the last few weeks (e.g., "I felt miserable or unhappy", "I didn't enjoy 126 anything at all"). The answers are given on a three-point scale where respondents are asked to 127 decide if the statements are "true", "sometimes true", or "not true". Scoring of the SMFQ is 5 obtained by summing together the point values of responses for each item. The response choices 129 and their designated point values are as follows: "not true" = 0 points, "sometimes true" = 1 point, 130 "true" = 2 points. Higher scores on the SMFQ indicate more severe depressive symptoms. The 131 range of scores on the SMFQ varies from 0 to 26. A score of 12 or higher may indicate the presence 132 of depression in the respondent (see among others Thabrew et al., 2018). As reported by Jarbin et 133 al. (2020) suggested cut-offs on the SMFQ self-report have been divergent with cut-offs ranging 134 from 4-5 in studies with just fair AUC and younger subjects to a high of 10-12 in studies with good 135 AUCs and older subjects. 136 The SMFQ has been validated with children and young people aged between 6 and 19, however a 137 study by Turner et al. (2014) showed that it is a useful and valid diagnostic tool for studying 138 depression within a community-based sample in late adolescence and that it relates well to an adult 139 measure of depression, namely the Clinical Interview Schedule-Revised form (CIS-R; Lewis et al., 140 1992). The SMFQ has a number of important (psychometric and implementation) features (i.e. 141 internal consistency, test-retest reliability, validity, sensitivity to change as to the former, and 142 brevity, availability, ease of scoring as to the latter), thus making it a useful tool for analysing mood 143 and feelings attitude, especially during a pandemic when other methodologies such as experimental 144 lab studies present severe and objective constraints.

145
In this work, we focused on young adults (16-24) and adults (25+) so as to proceed with a self-146 administered questionnaire that was completed by the sampled respondents 1 . The final sample was 147 composed of 6,692 Italian individuals, representative of the Italian population in terms of age, 148 gender, and geographical area. More specifically, the sample design and the stratification were 149 based on the following variables: i) age (7 age groups: 16-17; 18-24; 25-34; 35-44; 45-54; 55-64: 150 65+); iii) gender, iii) geographical breakdown (all Italian regions and size of the residential 151 community; 7 classes), iv) education (2 classes: graduates and non-graduates). The field data 152 collection was conducted in June 2020 (from the 4 th of June to the 19 th of June) with a mixed 153 technique CATI (Computer Assisted Telephone Interviewing) and CAWI, (Computer Assisted Web 154 Interviewing) as to limit any risks in terms of sample's distortion and self-selection.

155
All methods were carried out in accordance with the Declaration of Helsinki. The experiment was 156 approved by Autorità per le Garanzie nelle Comunicazioni. Informed consent was obtained from all 157 participants and from a parent and/or legal guardian for participants under 18.

158
Since the SMFQ has mainly been validated for children and young adults (Turner et al., 2014), in 159 our analysis we separated young adults (aged between 16-24) from adults (25+). hospital or outside and were tested before or after dying. Regardless of their goodness and accuracy, 175 these statistics are those that have been most widely disseminated by health institutions and the 176 news outlets, thus potentially influencing the mood and feelings of Italian citizens.

179
The first step was to analyse the distribution of the SMFQ score among Italian population, just after 180 the lockdown period, i.e., at June 2020 (see Figure 1). As we would expect, the total distribution of 181 SMFQ score is asymmetric and skewed to the right (also by a skewness test), with mean at 5.2. It 182 presents a mode near 0, and a slight bump around 12. If we define a cut-off at 12 that may indicate 183 the presence of depression in the respondent (see Thabrew et al., 2018), then, at June 2020, 14.4% 184 of the Italian population lied above this threshold. In Figure 2, we compare the score distribution of 185 the two age groups: young adults (16-24) vs. adults (25+). Both distributions follow a similar 186 pattern (skew to the right with a mode near 0 and a long tail thereafter), however the scores across 187 groups age do not have the same distribution function 2 . In particular, the mean score of the two 188 groups is significantly different (7.04 for young adults and 4.97 for adults), and the probability of 189 scoring a higher value than the cut-off (SMFQ ≥ 12) is significantly higher for the youngest 190 (24.17% vs. 13.33%; Table 2).

191
The second step was to run econometric analysis on the overall population, correlating the SMFQ 192 score to some major determinants. Following the existing literature (see above), we included socio- censoring (left at 0 and right at 26) has been estimated 3 . Model (1) in Table 5 shows results of the 198 count model with aforementioned "traditional" explanatory variables. Younger individuals are 199 confirmed to be more exposed to depressive mood (the coefficient of age is negative and significant

212
We then turned to the core of the analysis, by adding variables related to the pandemic phase into 213 two stages. First, we added variables concerning the experience of the individual during the 214 pandemic (see model 2 of Table 5). COVID-19 is a dummy variable which is equal to 1 when the 215 individual or someone in her/his family tested positive to coronavirus, 0 otherwise 6 . Moreover, we 216 looked at the effect of lockdown introducing a dummy variable that is equal to 1 whenever the 217 individual kept on going to the workplace during the lockdown period and is 0 in the opposite 218 situation of people staying at home 7 . Results of model 2 show that COVID-19 is the variable that 219 displays the greatest effect on the probability of falling into depression (.733 and std. dev. .0562).

220
To obtain a more analytical account of this effect, we calculated a probit model of scoring more 221 than the cut-off (SMQF ≥ 12; see Table A.2) and then plot the probability of depression as a

229
In the third step (models 3-6 of Table 5), we introduced variables related to the diffusion of the 230 COVID-19 pandemic in the local area of residence of Italian citizens (either the county or the 231 region) 9 . In particular, we used the number of confirmed cases (at county level) and deaths (at a 232 regional level) in June 2020, in terms of absolute and relative (as a % of local population) values.

233
Indeed, these statistics were the most used by media outlets and public institutions to inform Italian 234 citizens about the spread of the pandemic 10 . Results of models 3-6 of Table 5  that the attention of individuals to the pandemic is not related to the local spread but to the national 239 diffusion (Castriota, Delmastro and Tonin, 2020). In sum, results show that anxiety and depressive 240 6 From our data, 7.6% of Italian population has experienced directly or indirectly negative health effects from coronavirus (see Table 3). This value seems to be coherent with official statistics. In fact, at June 2020, confirmed cases are nearly 0.5% of Italian population. This value must be times a factor of 7.3 if one considers only the close family network (see Istat, 2018). When one takes into account also the largest family network (second-and third-degree family members), one may reach percentages around 7-8% of Italian population. 7 Of course, by comparing scores of individuals who experienced total lockdown with people that were forced to keep going to their workplace we greatly underestimate the effect of lockdown. Indeed, the second category of individuals suffered severe stress and anxiety from the particular situation and type of job they did (by definition public interest work in hospitals, supermarkets, pharmacies,…), so that are not able to compare lockdown with a "normal" situation. 8 However, as mentioned (see previous footnote), the coefficient of the variable "no lockdown" underestimates the effect of lockdown. Indeed, this result tells us that individuals who were forced to go to the workplace during this period and, presumably, suffered from stress and anxiety due to the situation and type of work they did, score significantly lower values of SMFQ than people under lockdown. Therefore, the effect of lockdown on depression may be significantly higher than that detected. 9 At county level, the only available official statistics related to the number of COVID-19 confirmed cases. At regional level, there are other statistics such as the number of deaths and recovered individuals. 10 In any case, the use of other local variables on the pandemic (recovered, intensive care, …) confirms results of Table  5. symptoms are not connected with the local spread of the virus (but only with family cases), since 241 they are widespread nationwide (and this may also be the effect of a national lockdown).

242
As a final step, we split the sample into two age groups (young adults and adults) and we ran count 243 statistical models for the two groups, separately (see the Appendix for linear models, Table A.3).

244
Results are reported in Table 6 11 (in column 1 we report results for all individualsas in model 2 of 245 Table 5 -, in column 2 those of young adults and in column 3 those of adults) and confirm previous 246 outcomes for both age groups, in particular with regard to the significance and size of the effect of 247 COVID-19 (whose coefficient is equal to .485std. dev. .087for young adults and to .77std.  The higher vulnerability of younger adults was observed in previous research as well in relation to 278 anxiety and distress (Forte et al., 2020;Mazza et al., 2020). Our results complement this picture and 279 reveal an association with depressive symptoms. This result is of high importance when thinking about policy and health measures to be adopted to help younger generations to overcome the 281 individual and social loss that they experienced during the pandemic.

282
Regarding work and financial conditions, our results confirm that professional uncertainty and low 283 socio-economic status are related to depression, as found by Rossi and colleagues (2020) for people 284 with working discontinuity and financial struggles.   The present research stresses the need to take into account the psychological consequences of the 316 COVID-19 pandemic and lockdown, aiming at the implementation of a holistic approach that 317 considers both physical and mental health and well-being. The society as a whole, and in particular 318 vulnerable groups such as children, older adults, people with existing mental health disorders, and front-line health-care workers, call for support to overcome this difficult time. Next months will be 320 characterised by uncertainty, financial insecurity and worry, therefore it is pivotal to provide help 321 through mental health care which could also make use of the Telemedicine, e.g., telehealth and app 322 tools (see Kontoangelos et al., 2020). Indeed, online interactions can promote a sense of connection 323 and improve psychological well-being, as highlighted by Van Bavel and colleagues (2020). 324 Future studies should consider the long-term effects of the pandemic on mental health, adopting a 325 longitudinal design to measure change over time. Additional work should aim at comparing the 326 experiences of the different countries affected by the pandemic in order to understand the size of the 327 psychological impact and the potential risk and protective factors. Importantly, the data on people 328 who seek for mental health assistance should closely monitored to prevent a second pandemic of 329 psychological distress. ncov-v3.pdf?sfvrsn=195f4010_6