Lost in Lockdown? Covid-19, Social Distancing, and Mental Health in Germany

Abstract

The COVID-19 pandemic and social-distancing and stay-at-home orders can directly affect mental health and quality of life. In this ongoing project, we analyze rich data from Telefon-seelsorge, the largest German helpline service, to better understand the effect of the pandemic and of local lockdown measures on mental-health related helpline contacts. First, looking at Germany-wide changes, we find that overall helpline contacts increase by around 20% in the first week of the lockdown and slowly decrease again after the third week. Results for different topics suggest that the increase is not driven by financial worries or fear of the virus itself, but reflects heightened loneliness, anxiety, and suicidal ideation. Second, we exploit spatial variation in policies among German federal states to assess whether the effect depends on the stringency of local measures. Preliminary evidence suggests that the average effect is indeed more pronounced in states that implemented stricter measures.

Full text

Lost in Lockdown?
COVID-19, Social Distancing, and Mental Health in Germany
Stephanie Armbruster†Valentin Klotzbücher‡
This draft will be updated regularly and feedback is very welcome
May 23, 2020, click here for the latest version
Abstract
The COV ID-19 pandemic and social-distancing and stay-at-home orders can directly affect
mental health and quality of life. In this ongoing project, we analyze rich data from Telefon-
seelsorge, the largest German helpline service, to better understand the effect of the pandemic
and of local lockdown measures on mental health–related helpline contacts. First, looking
at Germany-wide changes, we find that overall helpline contacts increase by around 20% in
the first week of the lockdown and slowly decrease again after the third lockdown week. Our
results suggest that the increase is not driven by financial worries or fear of the virus itself,
but reflects heightened loneliness, anxiety, and suicidal ideation. Second, we exploit spatial
variation in policies among German federal states to assess whether the effect depends on the
stringency of local measures. Preliminary evidence suggests that the average effect is more
pronounced in states that implemented stricter measures.
Keywords: COVID-19, Stay-at-Home Orders, Mental Health
JEL codes: I12,I3
We thank Ingo Eckert, Beat Hintermann, as well as seminar audiences at the University of Basel for helpful comments
and suggestions. We also thank the TelefonSeelsorge, in particular Ludger Storch, for help in obtaining and
interpreting the helpline data. Hannah Altena provided excellent research assistance. All remaining errors are our
own.
†
University of Basel, Department of Economics, Peter-Merian-Weg 6, CH-4002 Basel, and University of Freiburg,
Chair of Environmental Economics and Resource Management, Tennenbacher Straße 4, DE-79106 Freiburg im
Breisgau. E-Mail:
‡
Corresponding author. University of Freiburg, Department of Economics, Wilfried Guth Chair of Constitutional
Political Economy and Competition Polic, Wilhelmstraße 1b, 79085 DE-Freiburg im Breisgau. E-Mail:

1 Introduction
There is widespread concern that the current COVI D-19 outbreak is associated with increased psy-
chological distress, mental illness, and suicide (Rajkumar 2020). Evidence from global economic
crises suggests that periods of high unemployment rates are followed by significant increases
in suicide (Parmar et al. 2016), and an exceptionally large number of suicide deaths occurred
at the time of the SARS epidemic in 2003 (Yip et al. 2010).
4
Social distancing measures such as
stay–at–home orders are effective in containing the spread of COVID-19 (Fang et al. 2020), but,
in addition to the outbreak of a pandemic itself, potentially cause severe mental illness (Brooks
et al. 2020). A better understanding of mental health trends during the COV ID-19 outbreak, and
particularly the implications of social-distancing policies, is essential to inform policy in the
current situation, where the net benefit of releasing lockdown measures is unclear (Layard et al.
2020).
In this study, we focus on Germany, where various social-distancing policies were enacted
on the national level as well as by the 16 federal states. The majority of shops were closed on
March 17th, and on Sunday, March 22nd, Germany implemented national-wide social distancing
and contact restrictions (further referred to as the “lockdown week”). In contrast to most other
European countries, the stringency of measures differs substantially between states: For example,
while Bavaria banned “leaving the house without a reason”, spending time outdoors was still
allowed in neighboring Baden-Württemberg. The central aim of this paper is to find out if the
demand for psychological assistance increased due to the general COVID-19 outbreak and lock-
down measures (hypothesis H1), and second, if the effect is stronger in those states that imposed
stricter measures (H2).
We test our hypotheses by using data from Germany’s largest online and telephone coun-
seling hotline, the “TelefonSeelsorge”, (TS, TelefonSeelsorge 2020) for the period 01/01/2019
– 04/31/2020, combined with data on reported daily COV ID-19 cases and deaths (RKI 2020a),
state–wide policy measure data for Germany (Armbruster and Klotzbuecher 2020), as well as
state-level unemployment (Bundesagentur fuer Arbeit 2020). After analyzing the development
4
Parmar et al. (2016) found a strong increase in suicide after the 2008 global economic crisis; there were about 4900
excess suicides in the year 2009 alone compared with those expected based on previous trends. Yip et al. (2010)
examine the case of Severe Acute Respiratory Syndrome (SARS) and suicide among older adults in Hong Kong,
finding that social disengagement, mental stress, and anxiety at the time of the SARS epidemic among a certain
group of older adults resulted in an exceptionally high rate of suicide deaths.
1

graphically, we take the data to an (i) event study framework to quantify the effect over time (H1)
and to a (ii) Difference–in–Difference (DiD) model, where we try to disentangle the effects of
the pandemic and the mitigation policies on mental health by comparing strict and less–strict
lockdown states and by controlling for infection rates that differ across states (H2).
Our main findings can be summarized as follows. During the week of the lockdown, demand
for counseling increased by around 20%, and started to slowly decrease again after the third week.
Results for different problem issues reveal that the spike in helpline contacts is mainly driven
by mental health issues, such as loneliness, fear and depression. Our results are robust to using
alternative econometric approaches and different specifications. Regarding H2, preliminary
evidence suggests that the effect is indeed stronger where stricter measures were implemented:
We find a significantly stronger increase in helpline contacts for strict lockdown states in the week
of the lockdown, in particular for contacts concerning mental health issues.
Our paper relates to several strands of interdisciplinary literature. We contribute to the current
medical and psychological research on the effect of COVI D-19 and mental health (Rajkumar
2020).
5
For China, Wang et al. (2020); Xiao (2020) and Liu et al. (2020) suggest that anxiety is a very
common individual mental health symptom. Our study offers valuable insights into the mental
health issues prevailing during the COVID-19 pandemic in Germany.
We further add to the fast growing literature analyzing the social (see, e.g. Brodeur et al. 2020;
Knipe et al. 2020; Brülhart and Lalive 2020) and economic impacts of the COVID-19 outbreak (see,
e.g. Alon et al. 2020). Focusing on lockdown measures, evidence shows that people’s behavior
towards compliance with prevention recommendations and lockdown policies can depend on
media exposure and misinformation (Bursztyn et al. 2020), political leader’s communication
(Ajzenman et al. 2020), people’s expectations about the length of the lockdown (Briscese et al.
2020) or the economic endowment of a living area (Wright et al. 2020).6
One closely related study by Brodeur et al. (2020) uses Google Trends data to a analyze the
consequences of COV ID-19 lockdowns implemented in Europe and America on well-being and
5A further living systematic map of the evidence is online available under: COVID-19: living map of the evidence
6
Bursztyn et al. (2020) focus on misinformation in the U.S. concerning the COV ID-19 risk and find that provision of
misinformation in the early stages of a pandemic affects precautionary behavior and downstream health outcomes.
Ajzenman et al. (2020) show that when Brazil’s president publicly dismisses the COV ID-19 risks, recommended
prevention practices were reduced. Briscese et al. (2020) study the role of expectation about the length of the
lockdown in Italy and resulting compliance with Stay–at–home orders. If the lockdown is longer than expected, there
is a lower willingness to comply. Wright et al. (2020) show that compliance with local Stay–at–Home orders depends
on the economic endowments and that low income areas comply less than areas with stronger endowments.
2

mental health. Findings suggest that there is evidence for severe mental health implications:
levels of fear are rising and searches for loneliness, worry and sadness increase substantially
under lockdown.
7
The main advantage of Google Trends over survey data is, in addition to the
availability of daily data for different countries before and after the pandemic, the fact that online
search intensity reveals the actual interest of the population. On the downside, older segments of
the population are less likely to search online and it is not possible to distinguish individuals by
age, gender or other characteristics. Tran et al. (2017) provide evidence and a review of research
on using Google Trends to forecast suicide and conclude that the validity of the approach is rather
low and depends very much on the specific search terms chosen.
The closest match to our approach is a preliminary analysis of Brülhart and Lalive (2020), who
analyze calls to Switzerland’s most popular helpline “Die Dargebotene Hand” during the COVID-19
outbreak. They show that anxiety did not increase substantially in response to lockdown measures,
and that only calls related to the pandemic threat,
i.e.
elderly individuals who worry about the
risk of infection, increased. They do, however, find that calls about relationship issues, as well as
addiction and suicidality, have been increasing during the lockdown. Our paper provides new
evidence from Germany, looking more closely into the development for different relevant topics
and further uses spatial variation in lockdown measures across states to analyze the effect of the
lockdown itself.
The remainder of this paper proceeds as follows. Section 2provides background on the
chronology of the COV ID-19 outbreak in Germany and on the TS and summarizes the data and
illustrates descriptive time trends. Section 3describes the econometric approach. Section 4
presents the empirical findings. Section 5concludes.
2 Background and Data
In this section, we provide background information on the timeline of the COVID-19 pandemic
in Germany (section 2.1) as well as on Germany’s largest psychological telephone and online
counseling service, the “TelefonSeelsorge” (section 2.2). In section 2.3, we describe our combined
dataset, followed by descriptive time trends in Section 2.4.
7
The findings of Knipe et al. (2020), who also study changes in online search behavior using Google Trends are
broadly similar
3

2.1 COVID-19 in Germany
In December 2019, SARS–COV–2, a new virus from the family of corona viruses, appeared in
China. The virus causes the lung disease COVID-19 with typical symptoms such as fever, cough,
breathing problems, sometimes runny nose and diarrhea. The infection is usually less severe but
in particularly difficult cases, life–threatening pneumonia can develop. The disease developed
into an epidemic in China in January 2020 and ultimately spread worldwide.
8
On March 11, 2020,
the WHO officially declared the previous epidemic a pandemic.
In Germany, which is the focus of our study, the first official case occurred on January 27,
2020. The Robert Koch Institute (RKI), the government’s central scientific institution in the field of
biomedicine initially rated the risk of the COV ID-19 pandemic for the population in Germany on
February 28, 2020 as “low to moderate”, since March 17 as “high” and since March 26 as “very high”,
especially for risk groups. Risk groups are classified based on a higher risk of severe symptoms,
which mainly occurs for individuals from about 50–60 years (87% of those who died of COVID-19
in Germany were
≥
70 years old (median age: 82 year), for smokers (weak evidence), very obese
people and individuals with certain medical conditions (RKI 2020b). On February 25, the first
cases were documented in Baden-Württemberg and North Rhine-Westphalia. As of May 5th 2020,
there are 166,877 confirmed cases in Germany, 132,700 recovered and 7,110 persons died with
COVID-19 (RKI 2020a).
As a response to the COVID-19 pandemic, Germany enacted various mitigation policies on
the national as well as on the federal state level (
i.e.
a large number of laws, ordinances, general
directives and other regulations). As the stringency index provided by Hale et al. (2020) makes clear,
these measures were relatively liberal compared to the lockdown in neighboring countries such as
France or Italy. On the national level, Germany started on March 8th with a recommendation to
cancel events with more than a thousand participants, followed by an entry stop for third-country
nationals, a global travel warning, and restrictions to within EU travel. Most shops, as well as
schools and kindergartens, were closed on March 17th. On March 22nd, Germany implemented
national–wide social distancing and contact restrictions. Both the “economic lockdown” of March
17th and the “social lockdown” on the 22nd were announced roughly two days before. We further
8
On 30 January 2020, the World Health Organization (WHO) announced the international health emergency in order
to counteract the spread to countries without efficient health systems. From February 28, 2020, the WHO’s reports
assessed the risk at global level as “very high”.
4

call the week of March 16th – 23th as the “lockdown week.” The goal of the social lockdown was to
reduce physical contact as much as possible, requiring a minimum distance of at least 1.5 meters
in public spaces. Restaurants and services in the field of personal care,
e.g.
hairdressers, cosmetic
studios, massage practices and tattoo studios, were closed, with exceptions only for medically
necessary services.
However, each of the 16 federal states in Germany regulated the lockdown details differently.
9
The different lockdown measures by each federal state are presented online and are regularly
updated, see Armbruster and Klotzbuecher (2020).
10
In particular, we classify the federal states
of Bavaria, Saarland, Berlin and Brandenburg and Sachsen–Anhalt as “strict lockdown states”,
as they implemented not only contact-restriction measures but also a stay–at–home order, not
allowing individuals to leave the house “without a reason.” Our data availability leaves us with
Bavaria, Saarland, and Sachsen–Anhalt as the strict states, see figure 1and section 2.3. In these
states, leaving one’s home was only allowed if there were good reasons. Such reasons included the
way to work, to emergency care, participation in necessary appointments, as well as individual
sport and exercise in fresh air. All other outside activity, however, such as resting in parks, were
not permitted.
Since April 10th, a 14–day domestic quarantine requirement for returnees from abroad was
implemented. Re-opening slowly started on April 28th, when the Saarland Constitutional Court
overturned parts of the restrictions: encounters with family members and spending time outdoors
were possible again. Around the Easter weekend, demands for further re–opening became louder
and since May 4th, school started to re–open, although daycare centers remained closed. Re–
opening of playgrounds, hairdressing salons, church services, museums and zoos started on May
6th on the national level. On the state level, Bavaria allowed to meet or visit a person outside
of the own household and close family members since May 5th. Five people can meet again in
Saxony–Anhalt, even if they do not belong to a common household and Lower Saxony decided to
gradually reopen restaurants and coffee shops from May 11. National contact restrictions and
mask requirements were generally extended until June 5th, but federal states are supposed to
take on more responsibility and decide about the regionally appropriate level of restrictions.
9
In Germany, authority between the federal government and the states is divided by sixteen partly–sovereign states,
see Ter-Minassian (1997) for an overview on the German system of fiscal federalism.
10The now widley used Hale et al. (2020) data base does not contain sub–national state level data for Germany.
5

2.2 Psychological counseling by the TelefonSeelsorge
With over 100 helpline–centers in Germany, the TS is by far the largest telephone and online
crisis helpline offered in Germany. It is free, anonymous, partly government funded, and the only
facility in Germany to offer telephone conversations day and night for people in crisis. The TS
is a pastoral service under responsibility of the Evangelical and the Catholic Church and can be
reached around the clock by telephone at the nationwide toll-free numbers +49 0800 1110 111
(Protestant), +49 800 1110 222 (Catholic), and 116 123, as well as online via webmail and a chat on
the central website telefonseelsorge.de. Online search for relevant topics, such as “kill yourself”
on Google lead individuals in Germany directly to the TS hotline, see figure A.1. Around 7,500 fully
trained volunteers (TS counselors) with a wide range of life and professional skills are available to
help those seeking advice in 105 local counseling centers.
Figure 1: Lockdown stringency and helpline–centers. Black dots represent the approximate
locations of TS helpline–centers, red shading indicates strict-lockdown states.
2.3 Data
Since 2019, the TS has been implementing a contact tracking system and we have access to
anonymized data on contacts to the TS for the period of 01/01/2019 – 04/28/2020 (Telefon-
Seelsorge 2020). The dataset includes information on the date, time, and duration, and type of
counseling (telephone, mail, chat, on–site), as well as the the organizational unit. Moreover, a
number of individual characteristics are recorded, and we know the gender, approximate age,
occupation, living situation (living alone, in marriage /partnership, in a family, in an institution, in
a shared apartment), as well as whether the contact was the first contact of the respective person
or a repeated one. Further details include known psychological diagnoses, suicidal ideation,
6

and up to three conversation topics per data record.
11
Table A.2 provides an overview over the
available variables. We drop records where people hung up, as well as those that are labeled as
jokes or irrelevant.
Out of the available information on conversation topics, we classify the following broader
categories, which are potentially overlapping and thus non–exclusive:
• Mental and physical health: depression, grief, suicide, self-harming behavior, fears, anger,
confusion, addiction, loneliness, other mental health, and physical constitution
• Violence: physical and sexual violence
• Social issues: relationships, religion, society
→
Relationships: life with partner, parenting, pregnancy, everyday relationships, family
relations, separation, virtual relationships
→Religion: Belief/values, church, religion
→Society: Society/culture
• Economic issues: Finance and economics
→Finances/inheritance, poverty, living situation
→Work situation, unemployment, job search
If a person seeking advice calls the TS or makes contact via the Internet, he or she will be
connected to a location that is as close as possible to one’s current location.This allows us to
track counseling by helpline–center, and therefore by federal state. Table A.1 gives an overview of
the helpline–centers by state. After some initial cleaning, where we drop erroneous records and
helpline–centers that start using the tracking system only later, we are left with 91 helpline–centers
and we concentrate on mail, chat and telephone contacts. On–site contacts are dropped as they
are not tracked consistently.
We combine our data set with information on state–level policy measures for Germany that
we compile together with collaborators (Armbruster and Klotzbuecher 2020). The data includes
information on the federal state level about the onset of the lockdown, the social distancing
policies, bans on social interaction in group settings (restaurants, movies, gymnasiums etc.), zoo,
11
During a contact is made, the TS counselor picks a maximum of three topics out of an available list with problem
topics.
7

kindergarten and school closures as well as shop closures. We use the national announcement
date of the social contact restrictions on the state level as our “lockdown” date.
We further complement the data set on the helpline–center level (i.e. the community of the
helpline–center) with daily COV ID-19 cases and deaths caused by COVID-19, provided by the RKI
(RKI 2020a). Suspected COVID-19 cases, as well as evidence of SARS-COV-2 are reported to the
responsible health authorities. The data is transmitted electronically by the health department to
the state authorities and from there to the RKI at the latest on the next working day where the data
is validated using largely automated algorithms.
12
The cases are assigned to the federal state or
county from which the case was transmitted, which usually corresponds to the place of residence
or habitual residence of the cases and not the place where the person was probably infected. Note
that as our main goal is to control for the fear caused by locally reported cases, it is not important
whether the numbers reflect the actual prevalence of the disease but rather captures the alert
level transported in local media. Moreover, we also use monthly unemployment rates on the state
level from (Bundesagentur fuer Arbeit 2020).
2.4 Graphical Analysis of Helpline Contacts in 2020
Figure 2a shows the development of the daily number of helpline contacts around the social
lockdown date (03/22/2020) in Germany. Overall, contacts sharply increased around one week
before the national social lockdown, from around 1800 to almost 2400 contacts per day. After
around 3 weeks, the number of contacts starts to decrease again, but remains elevated at around
2200 daily contacts at the end of April.
Figure 2b shows the mean number of daily contacts for a helpline–center around the same
time, distinguishing strict and less–strict lockdown states. Before the lockdown, an average center
received about 22–25 contacts each day, with no substantial difference between the two groups.
Around the lockdown date, the average number increases by around 5 contacts per day, and the
increase appears to be slightly stronger in strict lockdown states.
Looking into the development of contacts by topic,
i.e.
contacts concerning mental and physi-
cal health (figure 3a), violence (figure 3b), and social issues (figure 3c) allows us to gain a better
understanding of what is behind the strong overall increase. Figure 3a illustrates that the overall
12
Only cases in which laboratory diagnostic confirmation is available regardless of the clinical picture are published.
8

1600 1800 2000 2200 2400
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Contacts by center, daily
(a) Daily contacts in Germany, 2020
20 25 30 35
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Strict states (BY,SA,SL) Other States
(b) Mean daily contacts per center, strict and non–strict states
Figure 2: Helpline contacts before and after lockdown. The upper graphs shows the daily number
of total contacts in Germany. The solid line is fitted using kernel-weighted local polynomial
regression, dashed lines represent the 95% confidence intervals. The lower graph shows the
average daily contacts by helpline–center in strict lockdown states in red and in all other states in
blue.
9

increase is driven by contacts related to mental health issues have risen sharply from around 1400
daily contacts to around 1800. Contacts dealing with the work and financial situation remained
roughly constant with a slightly decreasing trend (figure 3d). While not as strong as the increase
in mental health–related contacts, we also see a light uptick in contacts who talk about physical
and sexual violence. Note that the true prevalence of domestic violence might be higher than
figure 3b suggests, as victims might not be able to contact the helpline while in lockdown with
their tormentor.
1200 1400 1600 1800
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Mental and physical health
(a) Mental and physical health
30 40 50 60 70 80
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Violence
(b) Violence
700 800 900 1000 1100
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Social issues
(c) Social issues
100 150 200 250 300
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Work and financial situation
(d) Economic issues
Figure 3: Daily helpline contacts in Germany by topic, before and after lockdown. The solid line
is fitted using kernel-weighted local polynomial regression, dashed lines represent the 95%
confidence intervals.
In figure 4, we show mental health–related contacts further broken down into subcategories
(see section 2.3 for details) and look at the topics of loneliness, addiction, and suicidal ideation.
Loneliness, i.e. the perceived discrepancy between desired and actually existing relationships, is
as a key concern regarding the effect of social distancing policies. We see a sharp increase from
around 400 to 550–600 daily contacts during the week of the national lockdown. Contacts peaked
after around two weeks and started to decline again, but did not fully revert to the pre–lockdown
10

level. Addiction–related contacts (figure 4b) seem to decrease immediately before the lockdown,
from around 60 contacts to a little over 50, but then increase with a delay of around one week to
around 70 daily contacts. As the COVID-19 pandemic is challenging for many people (
e.g.
fears of
subsistence, social isolation, overwhelmed with home office and childcare), some might get used
to drinking regularly, and functional addicts might further loose control without the daily routine
of work. Closed borders might additionally lead to illegal drugs becoming more expensive.
300 400 500 600 700
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Issue: Loneliness, isolation
(a) Loneliness
40 50 60 70 80 90
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Issue: Addiction
(b) Addiction
150 200 250 300
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Suicide
(c) Suicide
200 300 400 500
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Issue: Fears
(d) Fear
Figure 4: Daily helpline contacts in Germany, mental health–related issues, before and after
lockdown. The solid line is fitted using kernel-weighted local polynomial regression, dashed lines
represent the 95% confidence intervals.
The demand for suicide counseling (
i.e.
contacts relating to suicidal thoughts, intentions,
or even suicide attempts) shows a similar development as overall mental health, with a sharp
increase in the week of the national lockdown, from around 230 to 280 contacts per day (4c). Also
interesting is the development of fear–related contacts shown in figure 4d: Already four weeks
before the lockdown, we see an increase from 250 to 350, probably reflecting fear of the pandemic
itself. Around the lockdown, these contacts further increase to around 450 per day.
In principle, people are not good at enduring insecurities. Humans have a basic need for
11

consistency, and the experience of coherence in our lifestyle is demonstrably related to life satis-
faction. Due to the multiple uncertainties caused by the corona crisis, this cannot be guaranteed,
which contributes to the individual and collective reduction in mental well-being (Grevenstein
et al. 2018).
3 Empirical Approach
In this section, we elaborate on the empirical method we use to test our main hypotheses. To
quantify the magnitude and statistical significance of the previously described effects more
precisely, we apply an event study design to assess the dynamic movements of helpline demand.
In order to capture differences in lockdown measures and other local factors such as the locally
reported number of COVID-19 infections, we analyze the effect in a daily panel of helpline–centers.
In total, our panel covers 91 helpline–centers and the period from 1/1/2019 up to 28/4/2020. The
baseline specification we estimate to test H1 takes the following form:
Contactsi,j,t =α+
5
X
τ=−9
βτweekτ
t+γX0
i,t +δZ0
j,t +ξi+θt+µt+υt+εi,j,t (1)
The dependent variable is the number of contacts (general and later by subcategory) per helpline–
center
i
in the federal state
j
on date
t
. The dummies
weekτ
t
takes the value of one if date is
within
τ
weeks before/after the lockdown week (March 16–22), and zero otherwise.
X0
i,t
is a vector
of community–level control variables (COVID-19 cases),
Z0
j,t
contains controls on the state level
(unemployment rate).
ξi
represent helpline–center fixed effects that capture constant factors
on the helpline–center and state level,
e.g.
the size of the helpline–center, quality of counseling
service, or local culture. We also include a weekly linear time trend
θt
to capture the long-term
increase in contacts, as well as year and weekday indicators, denoted
µt
and
υt
. The constant is
represented by αand εi,j,t is the error term.
To learn if there is a higher demand for psychological counseling in stricter states (H2), we
extend the event study and estimate the following model:
Contactsi,j,t =α+
5
X
τ=−9λτweekτ
t×strictj+γX0
i,t +δZ0
j,t +ξi+ϑt+εi,j,t (2)
12

where we include again helpline–center fixed effects
ξi
and control for local COVID-19 infections
and unemployment, and where strictjis defined as follows:
strictj=




1 if j =Bavaria/Saarland/Saxony–Anhalt
0 else
Berlin and Brandenburg also fall under this category of strict states but are dropped from the
analysis because of incomplete coverage. Importantly, in this specification we include daily date
fixed effects
ϑt
that non-parametrically capture all common time effects (e.g. chancellor Merkel’s
speech on March 18), allowing us to isolate the differential effect in strict-lockdown states. For the
ease of interpretation, we provide simple OLS estimates even though our dependent variable is
the non-negative count of contacts and a count data model is therefore more appropriate (Greene
2003). We obtain qualitatively similar results when we estimate the model using a Poisson Pseudo
Maximum Likelihood (PPML) estimator (Correia et al. 2019a,b).
4 Results
In this section, we present our effects estimates for H1 and H2. Our outcome of interest is the
change in helpline–center contacts across different time periods and problem categories.
4.1 Helpline Contacts Before and After Lockdown
In Table 1, we show the results of model
(1)
, estimated using OLS for helpline contacts in levels or
logs, as well as PPML. As all specifications show similar results, we focus on the most simple model
and plot the coefficients from column (1) in figure 5: The results confirm the interpretation from
the graphical analysis, indicating that the introduction of lockdown measures is associated with a
significant increase in helpline contacts. H1 is confirmed. In the first four weeks of the lockdown,
approximately four to five additional daily contacts were recorded at an average helpline–center:
After the fourth week contacts decrease again, and although they remain elevated, the difference
is not statistically significant. We find a significant time trend but no discernible effect of local
infections or unemployment.
Estimation results for the four main groups by problem category are presented in table 2.
13

−2 0 2 4 6 8
daily helpline contacts
−9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5
days from 22/3
Figure 5: Event study results. The graph shows point estimates from table 1, column (1) with 95 %
confidence intervals. Zero is the week of the lockdown and the coefficients are estimated relative
to the time before.
For ease of interpretation, we present OLS estimations with count of contacts as the dependent
variable. PPML estimates are identical in terms of significance and sign. See table A.5 and A.6.
The picture which is emerging in the descriptive trends can be confirmed. The effect is most
pronounced for mental health–related contacts, which significantly increase starting in the first
lockdown week and peak in the second week. Additional contact related to physical and sexual
violence are positive and highly significant in the first week after the lockdown and then appear to
flatten out. For social and economic issues, we do not find a significant increase on the demand
for psychological counseling during the lockdown period.
A more detailed analysis of the increase in demand for advice on health problems reveals that
the increase is driven by loneliness and fear. As the results presented in table 3show, contacts
concerning loneliness significantly increase in the lockdown week and remain high until the
fourth week. For fear we find a significant increase already four weeks before the lockdown,
capturing the effect of the pandemic rather than the lockdown. After the lockdown, these contacts
increase further, and remain significantly higher four weeks after the lockdown. Unexpected
demand on suicidal ideation is most pronounced in week one after the lockdown, but flattens out
more quickly in the weeks after. For addiction, we find no significant increase after the lockdown.
14

Table 1: Event Study Results – Alternative Specifications
OLS PPML
Contacts log(Contacts) Contacts
Week -9 −0.038 0.015 −0.002
(0.262) (0.017) (0.011)
Week -8 0.041 0.012 0.001
(0.272) (0.018) (0.012)
Week -7 0.310 0.041* 0.013
(0.337) (0.021) (0.015)
Week -6 0.171 0.032 0.006
(0.462) (0.033) (0.021)
Week -5 0.602 0.047 0.024
(0.550) (0.043) (0.025)
Week -4 0.960* 0.072* 0.039*
(0.527) (0.040) (0.024)
Week -3 0.342 0.037 0.011
(0.593) (0.047) (0.027)
Week -2 0.536 0.041 0.020
(0.630) (0.048) (0.028)
Week -1 0.867 0.059 0.033
(0.642) (0.044) (0.029)
Week of lockdown 4.244*** 0.182*** 0.167***
(0.677) (0.041) (0.027)
Week 1 4.951*** 0.181*** 0.186***
(0.950) (0.056) (0.039)
Week 2 3.797*** 0.129 0.143***
(1.192) (0.073) (0.049)
Week 3 3.722*** 0.133 0.137***
(1.202) (0.079) (0.052)
Week 4 1.948 0.067 0.069
(1.321) (0.085) (0.057)
Week 5 1.101 0.031 0.034
(1.333) (0.085) (0.058)
C19 cases 0.163 0.009 0.009*
(0.119) (0.007) (0.005)
Unemployment 0.615 0.032 0.015
(0.640) (0.049) (0.036)
Trend 0.060*** 0.005*** 0.003***
(0.013) (0.001) (0.001)
Constant 15.805*** 2.499*** 3.011***
(3.380) (0.259) (0.190)
Helpline center FE Ø Ø Ø
Year FE Ø Ø Ø
Weekday FE Ø Ø Ø
# Helpline centers 91 91 91
# Observations 34,199 34, 199 34,199
Note: Results from estimation equation (1), standard
errors in parentheses are clustered at the state level.
***, ** and * denote statistical significance at the 1%, 5% and 10% level.
15

Table 2: Event Study Results – Issues
Health Violence Social Economic
Week -9 0.089 0.027 −0.152 0.408***
(0.254) (0.030) (0.140) (0.076)
Week -8 0.066 0.026 −0.059 0.377***
(0.297) (0.051) (0.130) (0.126)
Week -7 0.278 −0.044 −0.190 0.437***
(0.311) (0.061) (0.145) (0.109)
Week -6 −0.002 0.038 0.011 0.514***
(0.408) (0.064) (0.273) (0.170)
Week -5 0.661 0.003 0.022 0.587***
(0.538) (0.042) (0.238) (0.100)
Week -4 0.884* −0.003 0.322 0.351***
(0.416) (0.063) (0.292) (0.085)
Week -3 0.617 −0.083 −0.451 0.344***
(0.529) (0.065) (0.312) (0.112)
Week -2 0.924* −0.065** −0.315 0.230*
(0.493) (0.027) (0.272) (0.109)
Week -1 0.199 −0.142*** −1.175*** 0.195***
(0.521) (0.041) (0.339) (0.060)
Week of lockdown 2.490*** −0.078 −0.480 0.099
(0.529) (0.051) (0.334) (0.085)
Week 1 3.389*** 0.129** −0.097 0.106
(0.765) (0.044) (0.388) (0.113)
Week 2 2.554** 0.053 −0.254 −0.025
(0.956) (0.100) (0.579) (0.182)
Week 3 2.325** 0.109 0.402 −0.159
(0.933) (0.084) (0.503) (0.157)
Week 4 1.119 0.051 −0.527 0.094
(1.101) (0.091) (0.643) (0.206)
Week 5 0.387 0.052 −0.797 0.067
(1.074) (0.078) (0.585) (0.199)
C19 cases 0.141 0.005 0.083* 0.024*
(0.095) (0.003) (0.045) (0.013)
Unemployment 0.500 −0.016 0.321 0.017
(0.444) (0.043) (0.217) (0.150)
Trend 0.045*** 0.001 0.027*** 0.001
(0.011) (0.001) (0.006) (0.003)
Constant 11.845*** 0.646** 7.436*** 2.242**
(2.373) (0.228) (1.180) (0.785)
Helpline center FE Ø Ø Ø Ø
Year FE Ø Ø Ø Ø
Weekday FE Ø Ø Ø Ø
# Helpline centers 91 91 91 91
# Observations 34, 199 34,199 34,199 34, 199
Note: Results from estimation equation (1), standard errors in parentheses are clustered
at the state level. ***, ** and * denote statistical significance at the 1%, 5% and 10% level.
16

Table 3: Event Study Results – Mental Health Issues
Loneliness Suicide Addiction Fear
Week -9 −0.180 −0.022 0.004 0.052
(0.110) (0.092) (0.039) (0.122)
Week -8 −0.201 −0.184* −0.033 0.108
(0.167) (0.094) (0.044) (0.144)
Week -7 −0.003 −0.059 0.015 0.247**
(0.135) (0.080) (0.048) (0.095)
Week -6 −0.060 −0.101 −0.010 0.133
(0.132) (0.129) (0.054) (0.109)
Week -5 0.035 −0.111 0.051 0.124
(0.191) (0.130) (0.065) (0.158)
Week -4 0.112 −0.021 0.014 0.087
(0.161) (0.160) (0.063) (0.128)
Week -3 0.146 −0.099 −0.035 0.485***
(0.159) (0.233) (0.053) (0.148)
Week -2 0.129 0.054 0.047 0.673***
(0.137) (0.173) (0.066) (0.196)
Week -1 −0.142 −0.201 −0.104** 1.049***
(0.116) (0.174) (0.048) (0.188)
Week of lockdown 1.206*** 0.113 −0.110* 2.134***
(0.226) (0.090) (0.056) (0.209)
Week 1 1.639*** 0.385** −0.032 1.970***
(0.303) (0.141) (0.047) (0.182)
Week 2 1.378*** 0.257 0.034 1.312***
(0.355) (0.286) (0.047) (0.241)
Week 3 1.178*** 0.214 0.069 1.015***
(0.309) (0.248) (0.057) (0.318)
Week 4 0.552 −0.086 −0.043 0.668**
(0.354) (0.207) (0.065) (0.269)
Week 5 0.230 −0.124 −0.072 0.383
(0.366) (0.197) (0.065) (0.262)
C19 cases 0.037 0.019 0.009** 0.023
(0.025) (0.012) (0.003) (0.028)
Unemployment 0.322 0.157* 0.028 0.031
(0.203) (0.084) (0.035) (0.122)
Trend 0.018*** 0.007** 0.002* 0.008***
(0.004) (0.003) (0.001) (0.002)
Constant 2.105* 1.376*** 0.433** 2.457***
(1.096) (0.464) (0.183) (0.663)
Helpline center FE Ø Ø Ø Ø
Year FE Ø Ø Ø Ø
Weekday FE Ø Ø Ø Ø
# Helpline centers 91 91 91 91
# Observations 34,199 34,199 34,199 34,199
Note: Results from estimation equation (1), standard errors in parentheses are clustered
at the state level. ***, ** and * denote statistical significance at the 1%, 5% and 10% level.
17

4.2 Differential Effect by Lockdown Stringency
Table 4presents the estimation results for H2 specified by equation
(2)
. We find a significant
positive difference (5% level) between the strict and non–strict federal states in the week of the
lockdown (Week of the lockdown
×
strict) of around four additional calls per helpline–center per
day than in the less–strict states. Results for selected topics suggest a significantly higher increase
in the demand for health related contacts as well as for violence and economic issues.
When we further break down the category of mental health–related contacts in table
??
, we
see a positive difference for loneliness and fear, and an even stronger difference for contacts
concerning suicidal ideation. After the second lockdown week, we find no significant differential
increase in stricter states for any of the topics.
While this preliminary evidence speaks in favor of H2, we can not be certain what is behind the
stronger average effect in Bavaria, Saarland and Saxony-Anhalt. As a next step in our project, we
will not only classify states as “strict” and “less strict”, but also take a closer look at the individual
measures of the federal states to assess whether there are certain measures that people find
particularly difficult to cope with.
5 Concluding remarks
In this paper, we exploit some unique design features of the COV ID-19 lockdown in Germany
in order to bring new evidence to bear on two important questions. First, did the demand for
psychological assistance increase as a response to the outbreack of the COVID-19 pandemic and
the implemented lockdown measures? Second, is the increase in demand is higher in stricter
states?
We see clear evidence for substantial increase in the demand for psychological counseling
after the lockdown week, by around 20% relative to the time before. While contacts related to
financial worries and fear of the pandemic itself increase already before, the strong increase
around the lockdown date seems to be driven by heightened feelings of loneliness and other
mental health problems. For contacts concerning violence we see some increase as well. Results
are robust to using alternative estimators. Our analysis further suggests that, on average, stricter
states experience a somewhat stronger increase in helpline contacts compared to less strict states.
18

Table 4: Event Study Results, Lockdown Stringency
Total Health Violence Social Economic
Week -9 ×strict 2.029 1.899* 0.254*** 0.962 0.144
(1.641) (1.026) (0.078) (0.588) (0.215)
Week -8 ×strict 2.658 2.405* 0.099* 0.517 −0.082
(1.673) (1.251) (0.048) (0.665) (0.262)
Week -7 ×strict 2.686 2.537 0.079 0.889 0.068
(1.957) (1.462) (0.052) (0.705) (0.260)
Week -6 ×strict 2.534 2.106 0.080 0.565 0.152
(1.822) (1.196) (0.067) (0.864) (0.405)
Week -5 ×strict 3.896 2.909 0.148** 1.940** 0.629**
(2.247) (1.730) (0.053) (0.867) (0.218)
Week -4 ×strict 3.446 3.472 0.097 1.516 0.325
(2.815) (2.033) (0.063) (1.262) (0.263)
Week -3 ×strict 3.048 2.561 0.109 0.671 0.284
(2.815) (2.232) (0.097) (1.280) (0.349)
Week -2 ×strict 2.534 2.616 0.222*** 0.950 0.334
(2.509) (1.672) (0.068) (1.073) (0.195)
Week -1 ×strict 3.542* 2.613** 0.147 1.615 0.420**
(1.709) (1.068) (0.091) (1.019) (0.147)
Week of lockdown ×strict 4.378** 3.134** 0.296*** 0.971 0.723**
(1.923) (1.303) (0.066) (1.048) (0.298)
Week 1 ×strict 3.258 2.475 0.269*** 0.936 0.712**
(2.705) (2.090) (0.062) (1.115) (0.305)
Week 2 ×strict 3.674 3.274* −0.005 0.508 0.359
(2.714) (1.834) (0.083) (1.410) (0.437)
Week 3 ×strict 3.896 2.971 0.047 1.113 0.470*
(3.720) (2.379) (0.054) (1.458) (0.218)
Week 4 ×strict 2.870 2.593 −0.018 −0.141 0.103
(3.523) (2.595) (0.132) (1.660) (0.423)
Week 5 ×strict 2.840 2.168 0.107 0.413 0.171
(3.590) (2.530) (0.102) (1.555) (0.373)
C19 cases 0.239** 0.202*** 0.007** 0.120*** 0.030**
(0.085) (0.065) (0.003) (0.032) (0.013)
Unemployment 1.696*** 1.120*** −0.021 0.731*** 0.166
(0.311) (0.222) (0.050) (0.228) (0.154)
Constant 12.147*** 10.031*** 0.706** 5.949*** 1.563*
(1.661) (1.183) (0.255) (1.188) (0.793)
Helpline center FE Ø Ø Ø Ø Ø
Date FE Ø Ø Ø Ø Ø
# Helpline centers 88 88 88 88 88
# Observations 32,914 32,914 32, 914 32, 914 32,914
Note: Results from estimation equation (1), standard errors in parentheses are clustered at the state level.
***, ** and * denote statistical significance at the 1%, 5% and 10% level.
Our findings are important as they shed light on the true extent of mental health consequences
of the COV ID-19 pandemic and lockdown measures in Germany. Our results support the recent
warning by the United Nations: Launching the UN policy brief on COVID-19 and mental health
on May 13th, Secretary-General António Guterres stressed that “mental health services are an
essential part of all government responses to COVID-19”.13
This article is still work in progress as we will further analyze different groups and topics, as
13See www.un.org/en/coronavirus/mental-health-services-are-essential-part-all-government-responses-covid-19
19

Table 5: Event Study Results, Lockdown Stringency by Mental Health Issues
Loneliness Suicide Addiction Fear
Week -9 ×strict 0.552 −0.128 −0.005 0.738***
(0.458) (0.174) (0.078) (0.209)
Week -8 ×strict 1.400*** −0.005 0.074 0.893**
(0.435) (0.199) (0.108) (0.335)
Week -7 ×strict 0.988 −0.207 0.138 0.303
(0.736) (0.141) (0.080) (0.281)
Week -6 ×strict 1.094** 0.004 0.158* 0.594**
(0.449) (0.176) (0.079) (0.259)
Week -5 ×strict 1.429* −0.110 −0.059 0.883
(0.669) (0.183) (0.120) (0.531)
Week -4 ×strict 1.439 −0.251 0.335*** 0.737
(0.895) (0.391) (0.074) (0.443)
Week -3 ×strict 1.020 0.030 0.265*** 0.716
(0.697) (0.293) (0.087) (0.476)
Week -2 ×strict 0.851 −0.066 0.249 0.291
(0.545) (0.305) (0.171) (0.368)
Week -1 ×strict 0.836** 0.617*** −0.149** 0.433
(0.358) (0.161) (0.061) (0.308)
Week of lockdown ×strict 0.863* 0.302* 0.033 1.128**
(0.443) (0.156) (0.111) (0.377)
Week 1 ×strict 0.890 0.515*** −0.155 0.615*
(1.034) (0.127) (0.089) (0.306)
Week 2 ×strict 1.634** 0.653** 0.094 1.197*
(0.713) (0.289) (0.076) (0.638)
Week 3 ×strict 1.399 0.268 −0.022 0.812
(0.863) (0.290) (0.170) (0.744)
Week 4 ×strict 1.036 −0.176 −0.013 1.047*
(0.789) (0.247) (0.084) (0.576)
Week 5 ×strict 1.306 −0.010 −0.096 0.679
(0.821) (0.237) (0.119) (0.633)
C19 cases 0.050*** 0.030*** 0.012*** 0.038*
(0.012) (0.009) (0.003) (0.020)
Unemployment 0.421** 0.183 0.084** 0.137
(0.146) (0.137) (0.038) (0.103)
Constant 2.104** 1.440* 0.181 2.276***
(0.754) (0.707) (0.196) (0.537)
Helpline center FE Ø Ø Ø Ø
Date FE Ø Ø Ø Ø
# Helpline centers 88 88 88 88
# Observations 32, 914 32, 914 32,914 32,914
Note: Results from estimation equation (1), standard errors in parenthesesare clustered at the
state level. ***, ** and * denote statistical significance at the 1%, 5% and 10% level.
well as update our estimates as new data points become available, allowing us to look at the
development in May 2020. Given that the lockdown measures in Germany were far less strict
than in other countries, future research should look more closely into the mental health effects in
stricter countries such as France or Italy.
20

References
Ajzenman, N., T. Cavalcanti, and D. Da Mata (2020): More than words: Leaders’ speech and risky
behavior during a pandemic. Available at SSRN 3582908 .2
Alon, T. M., M. Doepke, J. Olmstead-Rumsey, and M. Tertilt (2020): The impact of COVID-19 on
gender equality. Technical report, National Bureau of Economic Research. 2
Armbruster, S. and V. Klotzbuecher (2020): COVID-19 German state policy
database. Available online. URL:
.1,5,7
Briscese, G., N. Lacetera, M. Macis, and M. Tonin (2020): Compliance with Covid-19 social-
distancing measures in Italy: the role of expectations and duration. Technical report, National
Bureau of Economic Research. 2
Brodeur, A., A. E. Clark, S. Fleche, and N. Powdthavee (2020): Assessing the impact of the coron-
avirus lockdown on unhappiness, loneliness, and boredom using Google Trends. IZA Discussion
paper 2020 (13204). ISSN 2365-9793. 2
Brooks, S. K., R. K. Webster, L. E. Smith, L. Woodland, S. Wessely, N. Greenberg, and G. J. Rubin
(2020): The psychological impact of quarantine and how to reduce it: rapid review of the
evidence. The Lancet 395: 912–920. 1
Brülhart, M. and R. Lalive (2020): Psychological and social strain during the coronavirus-lockdown:
Evidence from helpline calls. Mimeo. 2,3
Bundesagentur fuer Arbeit (2020): Arbeitslose, Arbeitslosenquoten, Gemeldete Arbeitsstellen:
Bundesländer, Monate. Online available at Statistisches Bundesamt (Destatis). Accessed
05/2020. 1,8
Bursztyn, L., A. Rao, C. Roth, and D. Yanagizawa-Drott (2020): Misinformation during a pandemic.
University of Chicago, Becker Friedman Institute for Economics Working Paper 2020-44. 2
Correia, S., P. Guimarães, and T. Zylkin (2019a): PPMLHDFE: Fast Poisson estimation with high-
dimensional fixed effects. arXiv preprint arXiv:1903.01690 .13
21

Correia, S., P. Guimarães, and T. Zylkin (2019b): Verifying the existence of maximum likelihood
estimates for generalized linear models. arXiv preprint arXiv:1903.01633 .13
Fang, H., L. Wang, and Y. Yang (2020): Human mobility restrictions and the spread of the novel
coronavirus (2019-ncov) in china. Technical report, National Bureau of Economic Research. 1
Greene, W. H. (2003): Econometric Analysis. Pearson Education India, 7 edition. 13
Grevenstein, D., C. Aguilar-Raab, and M. Bluemke (2018): Mindful and resilient? incremental
validity of sense of coherence over mindfulness and big five personality factors for quality of
life outcomes. Journal of Happiness Studies 19 (7): 1883–1902. 12
Hale, T., S. Webster, A. Petherick, T. Phillips, and B. Kira (2020): Oxford covid-19 government
response tracker. Online. Blavatnik School of Government. 4,5
Knipe, D., H. Evans, A. Marchant, D. Gunnell, and A. John (2020): Mapping population mental
health concerns related to COVID-19 and the consequences of physical distancing: a Google
trends analysis. Wellcome Open Research 5 (82). doi:10.12688/wellcomeopenres.15870.1. 2,3
Layard, R., A. E. Clark, J.-E. De Neve, C. Krekel, D. Fancourt, N. Hey, and G. O’Donnell (2020): When
to release the lockdown? a wellbeing framework for analysing costs and benefits. Technical
report, IZA Discussion Papers. 1
Liu, S., L. Yang, C. Zhang, Y.-T. Xiang, Z. Liu, S. Hu, and B. Zhang (2020): Online mental health
services in China during the COVID-19 outbreak. The Lancet Psychiatry 7 (4): e17–e18. 2
Parmar, D., C. Stavropoulou, and J. P. Ioannidis (2016): Health outcomes during the 2008 financial
crisis in Europe: systematic literature review. BMJ 354. 1
Rajkumar, R. P. (2020): COVID-19 and mental health: A review of the existing literature. Asian
Journal of Psychiatry p. 102066. 1,2
RKI (2020a): Covid-19 infections per 100,000 population in the German federal states, dl-de/by-2-
0. online. 1,4,8
RKI (2020b): SARS-CoV-2 Fact Sheet on Coronavirus Disease-2019 (COVID-19). techreport, RKI.
As of 04/30/2020. 4
22

TelefonSeelsorge (2020): Detailed anonymized counseling data data. confidential. 1,6
Ter-Minassian, T. (1997): Fiscal Federalism in Theory and Practice. International Monetary Fund.
5
Tran, U. S., R. Andel, T. Niederkrotenthaler, B. Till, V. Ajdacic-Gross, and M. Voracek (2017): Low
validity of Google Trends for behavioral forecasting of national suicide rates. PloS one 12 (8). 3
Wang, C., R. Pan, X. Wan, Y. Tan, L. Xu, C. S. Ho, and R. C. Ho (2020): Immediate psychological
responses and associated factors during the initial stage of the 2019 coronavirus disease (COVID-
19) epidemic among the general population in china. International Journal of Environmental
Research and Public Health 17 (5): 1729. 2
Wright, A. L., K. Sonin, J. Driscoll, and J. Wilson (2020): Poverty and economic dislocation reduce
compliance with COVID-19 shelter-in-place protocols. University of Chicago, Becker Friedman
Institute for Economics Working Paper 2020-40. 2
Xiao, C. (2020): A novel approach of consultation on 2019 novel coronavirus (COVID-19)-related
psychological and mental problems: structured letter therapy. Psychiatry Investigation 17 (2):
175. 2
Yip, P. S., Y. Cheung, P. H. Chau, and Y. Law (2010): The impact of epidemic outbreak: the case
of severe acute respiratory syndrome (SARS)]and suicide among older adults in Hong Kong.
Crisis: The Journal of Crisis Intervention and Suicide Prevention 31 (2): 86. 1
Appendix
A.1 Additional figures and tables
Table A.1: List of Helpline–centers by State
Baden-Württemberg:
Freiburg (485), Heilbronn (134), Karlsruhe (89), Konstanz (484), Lörrach (480),
Mannheim (482), Offenburg (484), Pforzheim (481), Ravensburg (326), Stuttgart (913), Tübingen (404),
Ulm (485)
23

Bavaria:
Aschaffenburg (485), Augsburg (297), Bamberg (302), Bayreuth (478), Erlangen (472), Ingolstadt
(452), München (897), Passau (334), Regensburg (443), Rosenheim (483), Weiden (485), Würzburg (484)
Berlin: Berlin (822)
Brandenburg: Potsdam (1)
Bremen: Bremen (451)
Hamburg: Hamburg (908)
Hesse:
Darmstadt (484), Frankfurt (880), Fulda (120), Gießen (441), Hanau (485), Kassel (480), Mainz
(182), Trier (126)
Mecklenburg-Vorpommern: Greifswald (459), Neubrandenburg (466), Rostock (485), Schwerin (485)
Lower Saxony:
Bad Bederkesa (394), Braunschweig (337), Hannover (416), Meppen (59), Oldenburg
(274), Soltau (1), Wolfsburg (485)
North Rhine-Westphalia:
Aachen (484), Bad Neuenahr (31), Bad Oeynhausen (485), Bielefeld (474),
Bochum (485), Bonn (454), Dortmund (469), Duisburg (312), Düren (479), Düsseldorf (362), Essen
(883), Hagen (485), Hamm (485), Krefeld (485), Köln (519), Meschede (1), Münster (485), Neuss (316),
Paderborn (484), Recklinghausen (484), Siegen (485), Solingen (470), Wesel (318), Wuppertal (485)
Rhineland-Palatinate: Bad Kreuznach (120), Kaiserslautern (484), Koblenz (136)
Saarland: Saarbrücken (397)
Saxony: Auerbach (468), Chemnitz (188), Dresden (151), Leipzig (191), Zwickau (114)
Saxony-Anhalt: Dessau (485), Halle/Saale (326), Magdeburg (124)
Schleswig-Holstein: Kiel (119), Lübeck (35), Sylt (302)
Thuringia: Erfurt (129), Jena/Gera (1)
Note: The table shows the helpline–centers by federal state, number of daily observations in parentheses.
Stuttgart, München, Berlin, Hamburg, Frankfurt, Essen, and Köln each have two separate centers.
Table A.2: Summary Statistics – Individual and Contact Characteristics
Mean S.D. Min. Max. N
Chat contacts 0.045 0.208 0 1 715,227
Mail contacts 0.070 0.255 0 1 715, 227
Phone contacts 0.885 0.319 0 1 715, 227
Duration in minutes 22.693 29.046 0 17312 715, 227
First contacts 0.199 0.400 0 1 540, 657
Recurring contacts 0.801 0.400 0 1 540, 657
24

Mean S.D. Min. Max. N
Female 0.683 0.465 0 1 697, 929
Male 0.315 0.464 0 1 697,929
Other gender 0.002 0.049 0 1 697, 929
Living alone 0.642 0.479 0 1 622,869
Living in institution 0.052 0.222 0 1 622,869
Living with family 0.137 0.344 0 1 622, 869
Living with partner 0.143 0.351 0 1 622, 869
Living in shared flat 0.025 0.157 0 1 622,869
Searching job 0.061 0.240 0 1 555,418
Employed 0.280 0.449 0 1 555,418
Disability 0.278 0.448 0 1 555,418
Not searching job 0.058 0.234 0 1 555, 418
Retired 0.234 0.423 0 1 555,418
In education 0.088 0.284 0 1 555,418
Suicide of others 0.013 0.112 0 1 714, 959
Suicidal thoughts 0.086 0.281 0 1 714, 959
Suicidal intentions 0.014 0.119 0 1 714,959
Suicide attempts 0.012 0.109 0 1 714,959
Psych. diagnosis 0.326 0.469 0 1 714,968
Table A.3: Summary Statistics – Topics
Mean S.D. Min. Max. N
Physical constitution 0.165 0.371 0 1 702, 351
Depressive mood 0.178 0.383 0 1 702, 351
Grief 0.044 0.206 0 1 702,351
Fears 0.146 0.353 0 1 702,351
Stress, emotional fatigue 0.091 0.288 0 1 702,351
Anger, agression 0.073 0.260 0 1 702, 351
Self-harming behaviour 0.014 0.115 0 1 702, 351
Confusion 0.023 0.150 0 1 702,351
Addiction 0.030 0.171 0 1 702,351
Low confidence, shame 0.068 0.252 0 1 702,351
25

Mean S.D. Min. Max. N
Loneliness, isolation 0.211 0.408 0 1 702, 351
Positive feeling 0.013 0.113 0 1 702, 351
Suicidal self 0.031 0.174 0 1 702, 351
Suicidal other 0.011 0.104 0 1 702, 351
Sexuality 0.027 0.162 0 1 702,351
Other mental issues 0.075 0.263 0 1 702,351
Partner search or choice 0.056 0.230 0 1 702,351
Life with partner 0.076 0.265 0 1 702, 351
Parenting 0.025 0.157 0 1 702,351
Pregnancy, childwish 0.004 0.063 0 1 702,351
Family relations 0.167 0.373 0 1 702, 351
Everyday relationships 0.109 0.311 0 1 702,351
Public institutions 0.024 0.154 0 1 702,351
Care, therapy 0.071 0.257 0 1 702, 351
Separation 0.034 0.181 0 1 702,351
Mortality, death 0.028 0.165 0 1 702, 351
Virtual relationships 0.002 0.046 0 1 702, 351
Migration, integration 0.002 0.048 0 1 702, 351
Physical violence 0.018 0.133 0 1 702, 351
Sexual violence 0.012 0.110 0 1 702, 351
School, education 0.018 0.131 0 1 702, 351
Work situation 0.047 0.211 0 1 702, 351
Unemployment, job search 0.017 0.128 0 1 702,351
Daily routines 0.053 0.224 0 1 702, 351
Volunteering 0.003 0.055 0 1 702,351
Poverty 0.014 0.117 0 1 702,351
Finances, inheritance 0.023 0.148 0 1 702,351
Housing situation 0.027 0.162 0 1 702,351
Belief, values 0.028 0.165 0 1 702,351
Church, religion 0.006 0.079 0 1 702, 351
Society, culture 0.012 0.109 0 1 702, 351
TS: positive feedback 0.018 0.133 0 1 702,351
TS: negative feedback 0.003 0.057 0 1 702,351
TS: agreed feedback 0.001 0.035 0 1 702,351
26

Mean S.D. Min. Max. N
TS: other feedback 0.003 0.053 0 1 702,351
Further information 0.006 0.078 0 1 702, 351
Other topic 0.014 0.116 0 1 702, 351
Current topic 0.049 0.216 0 1 702, 351
Table A.4: Summary Statistics – Age groups
Mean S.D. Min. Max. N
Age: 0-9 0.000 0.011 0 1 653, 683
Age: 10-14 0.009 0.096 0 1 653,683
Age: 15-19 0.041 0.199 0 1 653,683
Age: 20-29 0.103 0.305 0 1 653,683
Age: 30-39 0.133 0.340 0 1 653,683
Age: 40-49 0.170 0.376 0 1 653,683
Age: 50-59 0.249 0.433 0 1 653,683
Age: 60-69 0.196 0.397 0 1 653,683
Age: 70-79 0.075 0.263 0 1 653,683
Age: 80 and above 0.023 0.149 0 1 653, 683
27

Table A.5: Event Study Results – Issues (PPML)
Health Violence Social Economic
Week -9 0.005 0.041 −0.015 0.158***
(0.014) (0.046) (0.013) (0.025)
Week -8 0.003 0.038 −0.007 0.146***
(0.017) (0.079) (0.013) (0.044)
Week -7 0.015 −0.070 −0.019 0.169***
(0.018) (0.094) (0.014) (0.036)
Week -6 −0.002 0.055 −0.001 0.196***
(0.025) (0.098) (0.026) (0.057)
Week -5 0.036 0.003 0.000 0.221***
(0.031) (0.066) (0.023) (0.034)
Week -4 0.048** −0.006 0.028 0.139***
(0.024) (0.098) (0.028) (0.034)
Week -3 0.031 −0.137 −0.046 0.136***
(0.031) (0.100) (0.031) (0.041)
Week -2 0.049* −0.106** −0.033 0.094**
(0.028) (0.042) (0.027) (0.042)
Week -1 0.007 −0.246*** −0.120*** 0.080***
(0.031) (0.068) (0.034) (0.024)
Week of lockdown 0.131*** −0.124 −0.049 0.041
(0.028) (0.079) (0.032) (0.035)
Week 1 0.169*** 0.180** −0.016 0.042
(0.041) (0.070) (0.038) (0.045)
Week 2 0.127** 0.097 −0.029 −0.012
(0.051) (0.151) (0.055) (0.073)
Week 3 0.113** 0.173 0.026 −0.069
(0.052) (0.130) (0.049) (0.064)
Week 4 0.053 0.099 −0.056 0.032
(0.061) (0.140) (0.063) (0.082)
Week 5 0.013 0.101 −0.082 0.020
(0.060) (0.123) (0.058) (0.080)
C19 cases 0.009** 0.004 0.009** 0.010**
(0.004) (0.005) (0.004) (0.005)
Unemployment 0.019 −0.041 0.023 0.006
(0.032) (0.067) (0.024) (0.064)
Trend 0.003*** 0.002 0.003*** 0.000
(0.001) (0.001) (0.001) (0.001)
Constant 2.712*** −0.184 2.231*** 0.979***
(0.170) (0.349) (0.127) (0.336)
Helpline center FE Ø Ø Ø Ø
Year FE Ø Ø Ø Ø
Weekday FE Ø Ø Ø Ø
# Helpline centers 91 90 91 91
# Observations 34, 199 34,169 34,199 34, 199
Note: Results from estimation equation (1) using PPML, stan-
dard errors in parentheses are clustered at the state level.
***, ** and * denote statistical significance at the 1%, 5% and 10% level.
28

Table A.6: Event Study Results – Issues (PPML)
Loneliness Suicide Addiction Fear
Week -9 −0.040 −0.009 0.005 0.016
(0.024) (0.035) (0.055) (0.037)
Week -8 −0.045 −0.073** −0.049 0.033
(0.037) (0.036) (0.065) (0.043)
Week -7 −0.003 −0.023 0.019 0.074***
(0.030) (0.031) (0.069) (0.028)
Week -6 −0.017 −0.040 −0.016 0.038
(0.028) (0.052) (0.075) (0.034)
Week -5 0.003 −0.044 0.068 0.035
(0.040) (0.052) (0.085) (0.048)
Week -4 0.018 −0.009 0.018 0.023
(0.034) (0.062) (0.086) (0.040)
Week -3 0.023 −0.040 −0.054 0.135***
(0.034) (0.093) (0.076) (0.041)
Week -2 0.020 0.018 0.062 0.186***
(0.029) (0.064) (0.089) (0.050)
Week -1 −0.040 −0.080 −0.161** 0.278***
(0.027) (0.072) (0.080) (0.046)
Week of lockdown 0.214*** 0.043 −0.168* 0.506***
(0.039) (0.034) (0.089) (0.043)
Week 1 0.272*** 0.137** −0.047 0.468***
(0.056) (0.054) (0.070) (0.037)
Week 2 0.217*** 0.088 0.036 0.337***
(0.066) (0.099) (0.067) (0.059)
Week 3 0.179*** 0.076 0.075 0.270***
(0.066) (0.082) (0.077) (0.078)
Week 4 0.073 −0.023 −0.066 0.189***
(0.075) (0.072) (0.094) (0.066)
Week 5 0.012 −0.037 −0.107 0.115*
(0.078) (0.071) (0.096) (0.061)
C19 cases 0.011* 0.005** 0.011** 0.005
(0.006) (0.002) (0.005) (0.004)
Unemployment 0.055 0.063* 0.040 −0.002
(0.052) (0.034) (0.054) (0.038)
Trend 0.004*** 0.003** 0.003* 0.003***
(0.001) (0.001) (0.002) (0.001)
Constant 1.214*** 0.581*** −0.596** 1.110***
(0.277) (0.191) (0.291) (0.207)
Helpline center FE Ø Ø Ø Ø
Year FE Ø Ø Ø Ø
Weekday FE Ø Ø Ø Ø
# Helpline centers 91 91 90 91
# Observations 34,199 34,199 34,164 34,199
Note: Results from estimation equation (1) using PPML, stan-
dard errors in parentheses are clustered at the state level.
***, ** and * denote statistical significance at the 1%, 5% and 10% level.
29

Figure A.1: Searching for help
1000 1200 1400 1600 1800
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Female
500 600 700 800
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Male
Figure A.2: Daily helpline contacts, by gender
30

100011001200130014001500
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Recurring contacts
200 250 300 350 400 450
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
First contacts
Figure A.3: Daily helpline contacts, repeated and first contacts
90010001100120013001400
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Living alone
150 200 250 300 350
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Living with family
150 200 250 300
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Living with partner
60 80 100 120 140 160
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Living in institution
Figure A.4: Daily helpline contacts, by living situation
31

300 350 400 450 500 550
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Employed
300 350 400 450 500 550
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Disability
250 300 350 400 450 500
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Retired
100 120 140 160 180
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
In education
Figure A.5: Daily helpline contacts, by occupation status
10 20 30 40
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Suicide of others
100 150 200 250
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Suicidal thoughts
10 20 30 40 50
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Suicidal intentions
10 20 30 40
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Suicide attempts
Figure A.6: Daily helpline contacts, different degrees of suicidal ideation
32

0 10 20 30 40
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 10−14
40 60 80 100 120
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 15−19
100 150 200 250 300
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 20−29
150 200 250 300 350
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 30−39
Figure A.7: Daily helpline contacts, by age group
33

200 250 300 350 400
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 40−49
350 400 450 500 550
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 50−59
300 350 400 450 500
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 60−69
100 150 200
daily helpline contacts
0−7−14−28−21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 70−79
Figure A.8: Daily helpline contacts, by age group (continued)
34

30 40 50 60 70
daily helpline contacts
0−7−14−28 −21−35−42−49−56−63−70−77 7 14 21 28 35
days from 22/3
Age: 80 and above
Figure A.9: Daily helpline contacts, by age group (continued)
35