Social capital, quality of institutions and lockdown. Evidence from Italian provinces.

This paper uses the exogenous shock generated by the COVID-19 pandemic and the relative government response as an informative case in investigating the factors able to affect policy efficacy. Lockdown measures have been widely adopted to limit the diffusion of COVID-19, indirectly supporting the capacity of the hospital system to face the pandemic. Lockdown obliges people to change their social behaviour significantly, and consequently is a matter of serious concern amongst the population. For this reason, how people react to lockdown is of the utmost importance, since failure to observe it properly will be of little benefit in reducing contagion. In this rationale, factors correlated with individuals' behaviour could affect the efficacy of such measures. The aim of this paper is to investigate whether differences in institutional quality and social capital are correlated with the efficacy of lockdown measures, taking the Italian provinces as a case study. Using a quantitative analysis employing F-GLS estimators, our results suggest that both local social capital and institutional quality have affected the efficiency of lockdown measures in Italian provinces. In general terms these factors contribute to forming the set of incentives able to promote individual behaviour that is in closer compliance with government choices. In this perspective, institutional quality and social capital can be considered factors able to influence the efficacy of policies.

Introduction

When the outbreak of a new coronavirus infectious disease (COVID-19), initially observed in the Chinese province of Wuhan, spread quickly across the world, it was accompanied by a dramatic and growing number of cases of infections and deaths. Governments everywhere have tried to strengthen the capacity of their hospital systems, and at the same time adopted policies aimed at reducing the probability of their citizens’ contracting the virus. More specifically, lockdown measures and social distancing are the most common examples of policies aimed at easing the strain on health services by slowing down the outbreak (Hamzelou, 2020) and implementing a more effective epidemic control (Shao, 2020).

In this context, Italy has been a very peculiar case. After an initial phase in Europe when a few small clusters of the virus were localized in Germany, France and the UK, a major outbreak erupted in northern Italy. The entire area, and in particular the regions of Lombardy, Emilia Romagna, and Veneto, has been severely affected by COVID-19. Starting on 23 February, 10 municipalities in Lodi province (in the region of Lombardy) and one in Padua province (in the region of Veneto) were locked down (Riccardo ). On 11 March 2020, the central government decided to extend the lockdown across the entire country, a measure that was put into action the very next day.

Despite a growing debate regarding the optimal point in the trade-off inherent in this measure (i.e. the choice between safeguarding citizens’ health and minimizing the negative economic impact of the lockdown), some early contributions to this fresh strand of the literature have highlighted the efficacy of such a measure in terms of reducing the contagion, by means of both single-country studies (De Figureido ; Lau ) and a cross-country approach (Alfano and Ercolano, 2020; Cruz and Dias, 2020). Nevertheless, as pointed out by Dhanwant and Ramanathan (2020), despite the fact that lockdown and social distancing are the only proven solutions when it comes to halting the spread of COVID-19, different infection rates may be observed, even when looking at the diffusion of the virus within a single country. For example, Sardar , using a predictive analytical study that incorporates the lockdown measures implemented in India, conclude that the positive effects of lockdown can be observed only in certain provinces.

It is thus interesting to offer some initial evidence regarding the factors capable of affecting the efficacy of such measures, especially in the case of countries that for a certain period of time adopted a homogeneous measure at national level, i.e. a country-wide policy. In this perspective, Italy is a particularly interesting case to study, because it has very high within-country regional disparities (Ercolano, 2012) that highlight the existence of wide cross-regional heterogeneity in economic terms. Moreover, it has been severely affected by the virus and experienced over two months of national lockdown. Indeed, after the first localized policies, aimed at containing the infection within a restricted number of municipalities, the Italian government felt the need to change strategy and restrict people across the country from going out and meeting other people. On 12 March 2020 a national lockdown was put in place. It was continued until 3 May of the same year, when less draconian measures were adopted. Fig. 1 shows the incidence (total cases over population) of COVID-19 in the different provinces.

Fig. 1

. COVID-19 Incidence in Italian Provinces (total cases divided by population) on May 3. .

It can easily be understood that these kinds of measures oblige people to effectuate a profound change in their social behaviour, which may provoke serious concern amongst the population, insofar as these sorts of policies recall troubling historical precedents (Lau ). For this reason, how people decide to observe lockdown is crucial if the measure is to have an effective outcome. In this perspective, different factors that correlate with individual behaviour could affect the efficacy of such measures.

In the present paper, we try to address the following research question: are differences in institutional quality and social capital correlated with differences in the efficacy of lockdown measures, when looking at the Italian provinces? In more general terms, using the random shock generated by the COVID-19 pandemic as an informative case, the aim of the paper is to contribute to the literature regarding policy effectiveness, investigating how institutional quality and social capital could affect the effectiveness of lockdown measures.

There are indeed reasons to believe that these dimensions could affect the efficacy of lockdown measures. With regard to institutional quality, according to Kaufmann , governance consists of the traditions and institutions by which authority in a country is exercised. Following this framework one could state that governance quality tends to facilitate individual compliance by means of the political trust capable of furthering the effectiveness and legitimacy of government action (Letki, 2006; Marien and Hooghe, 2011). In this perspective we can formulate the following research hypothesis:

H1: The perception of institutional quality positively affects the effectiveness of lockdown.

On the other hand, looking at social capital as proposed by Putnam (1995), which includes features of social organization such as networks, norms, and social trust, it is easy to imagine that this may facilitate coordination and cooperation for mutual benefit. However, this is a concept that must be approached with caution. Indeed, although a large strand of the literature deals with this concept and its operationalization, there is still no scholarly consensus around it. Social capital is a much-debated concept in economics and social science; nonetheless, thanks to its multidimensional nature, it has been widely used by social scientists to investigate a wide range of phenomena (Crescenzi ).

Looking at policy effectiveness, it might be suspected that a higher endowment of social capital at local level, fostering trust and cooperation amongst citizens, could help local governments in communicating to citizens the necessity and adequacy of such measures, improving individuals’ compliance with lockdown. If this is the case, the second research hypothesis can be formulated as follows:

H2: A higher level of social capital, facilitating trust and cooperation amongst citizens, positively affects the effectiveness of lockdown.

The focus on the Italian case is motivated by the significant heterogeneity detected within the country, in terms of both institutional quality and social capital. Indeed, if we look at institutional quality, as pointed out by previous research (Nifo and Vecchione, 2014, 2015; Lasagni ), the Italian southern provinces appear to be weaker than their northern counterparts in all aspects of institutional quality (corruption, bureaucratisation, inefficiency in the organization of public services, and lower endowment of infrastructures). If we look at social capital, following Banfield's seminal work (1958), and the subsequent contribution of Putnam (1993), Italy has been a unique case study because of the factor's uneven distribution across the national territory. In particular, as confirmed by several contributions, southern provinces appear to be poorer in social capital than their northern counterparts (Guiso; Leonardi, 1995; Helliwell and Putnam, 1995; Crescenzi ).

The rest of this paper is organised as follows: in the next section we review the main literature, focusing on the relation between quality of institutions, social capital and policy effectiveness. Section three introduces the data and methodology used to address our research question, while section four reports and describes the main results and robustness tests. The final section concludes the paper.

Literature review

Starting from the seminal contribution by North (1991), social scientists have found a new interest in studying the role of institutions in contemporary economies. These are “the humanly devised constraints that structure political, economic and social interaction” (North, 1991, p. 97). As pointed out by Glückler and Lenz (2016), a sort of institutional turn (Martin, 2000) can be observed in economic geography (Storper, 1997; Gertler, 2010; Storper, 2011; Rodríguez-Pose, 2013), economics (Williamson, 1985; Hall and Jones, 1999; Rodrik ; Acemoglu ), organizational science (Barley, 1990; Whitley, 1992; Zilber, 2011), and political science (March and Olsen, 1984; Helmke and Levitsky, 2004).

Governance is the means through which institutions exercise authority. More specifically, following Kaufmann , governance includes: i) the process by which governments are selected, monitored and replaced; ii) the capacity of the government to effectively formulate and implement sound policies; iii) the respect of citizens and the state for the institutions that govern the economic and social interactions between them. In order to measure how the public, private, and NGO sectors perceive institutional quality, Kaufmann and his co-authors propose a set of indicators that capture six different dimensions: Voice and Accountability, Political Stability and Absence of Violence/Terrorism, Government Effectiveness, Regulatory Quality, Rule of Law, and Control of Corruption.

Institutional quality as proposed by Kaufmann can be considered in general terms as the ways in which the public interest is addressed (Koppell and Auer, 2012). Following this, one might expect governance quality to be positively correlated with individuals’ compliance, by means of the political trust capable of promoting the effectiveness and legitimacy of government action (Hooghe and Marien, 2011). Indeed, as pointed out by Nifo and Vecchione (2014), governance and institutions contribute to forming the set of incentives that underlie behaviour and individual choices. If this mechanism is at work, it is possible to suppose that individuals’ varying perceptions of institutional quality could affect the effectiveness of policies. Torgler and Schneider (2009) find that when citizens perceive that their interests are properly represented by effective institutions, characterized by low levels of corruption, they tend to be more cooperative, reducing their willingness to act in the shadow economy. If we look at the educational sector, according to Zaman (2015), institutional quality can increase the effectiveness of reforms aimed at fostering the internationalization of universities. In more general terms, as stressed by Glückler and Lenz (2016), an institutional perspective on policies could help in understanding the role of governance in the policy-making process “as well as for the analysis of the effectiveness of regional policies with regard to their intended effects” (Glückler and Lenz, 2016, p. 270).

While the idea of modelling institutions as a set of constraints and incentives that affect individuals’ behaviour is intrinsically linked to the effectiveness of policy, as pointed out by the previously cited literature, it is worth noting that the relationship between citizens and institutions is also mediated by some historical patterns, which have been able to influence the nature of this relationship. It could be argued that social capital is a (much debated) concept that tries to capture some of those patterns. Following the definition proposed by Putnam (1995), it includes “features of social organization such as networks, norms, and social trust that facilitate coordination and cooperation for mutual benefit” (Putnam, 1995: p. 67). Looking more explicitly at the correlation between social capital and public policy, the author seems to suggest that rather than being a substitute for effective policy, social capital could be a prerequisite for it (Putzel, 1997).

In this perspective, it is worth noting that, while the quality of institutions can be described as a factor capable of affecting policy effectiveness through the supply side, social capital may be a demand-side factor for policy effectiveness. Indeed, as pointed out by Boix and Posner (1998), social capital can be summarized as a set of institutionalized expectations about cooperative individuals’ behaviour, which is potentially an additional explanatory variable for government effectiveness. In order to overcome the lack of “an explicit articulation of the mechanism by which the ability of people in society to co-operate affects the performance of the governmental institutions” (Boix and Posner, 1998, p. 689), the same authors suggest possible mechanisms able to explain the correlation between social capital and policy effectiveness. Firstly, according to them, social capital by means of a more active political participation may be a driver for policy effectiveness, informing citizens about the policies put in place. Secondly, following the main findings proposed by the literature on collective actions (Olson, 1982), Boix and Posner suggest that social capital can facilitate the demand for policies because civic communities are able to articulate their interest to the government effectively, ensuring its policies are closer to the wishes of the community. Furthermore, social capital, fostering a more optimistic view about citizens’ behaviour, can reduce the cost of enforcing and implementing policies. In this perspective social capital tends to encourage “the articulation of demands on government which are to everyone's benefit rather than helping some members of society at the expense of others” (Boix and Posner, 1998, p. 691). It is worth noting that the concepts of institutional quality and social capital seem to be intrinsically intertwined. Indeed, according to Fukuyama (2002: p. 29), “the concept of social capital puts both policies and institutions in their proper cultural context, and guards us from certain naive expectations that a relatively simple policy formula will lead inevitably to economic growth.” In this sense it is interesting to recall the concept of community governance, as proposed by Bowles and Gintis (2002), whereby a set of small groups’ social interactions became one of the ingredients of effective institutions able to handle certain social issues where the non-coordinated actions of both individuals and governments tend to fail.

This link suggests that from an empirical perspective it is important to control for both these dimensions when assessing the impact of institutional settings on policy effectiveness. For these reasons, we consider them to be intrinsically intertwined, and argue that it is impossible to evaluate the impact of one dimension on lockdown efficiency without also taking the other into account.

Data and methodology

To measure the efficacy of lockdown in the different Italian provinces, following Alfano and Ercolano (2020) we utilize a panel dataset, with daily data from the Italian provinces used as the basic statistical unit of observation. In formal terms, we estimate the following equation:where are new COVID-19 cases at time t with respect to t-1 in province p. As should be clear from Eq. (1), the daily increase in the spread of infections in province p is modelled as a function of the infections measured in the same province on the previous day (). The equation also includes DLD, a dummy signalling whether Italy was under lockdown on day t. It is equal to one on the days after 11 March, and zero before. Alternatively, DLD may also be used to signal whether a lockdown was already in place for x days (this will be explained in more detail later), in order to control for the efficiency of the policy over time. Indeed, we cannot expect that the implementation of a lockdown will be immediately effective in limiting the contagion (quite the opposite, although we should expect its efficacy to increase over time, since the infections reported in time t are likely to be due to transmissions of the virus that happened 7 to 14 days beforehand).

In order to estimate equation (1), we need: the daily number of COVID-19 cases, daily data on lockdown measures, and quality of government and social capital proxies to split the dataset into subsamples and observe different coefficients in the equation. The former data are gathered from the Italian Ministry of Health dataset, reporting official data for each province and day, from 24 February, the first day for which data were collected, to 3 May, when the lockdown was (partially, but in a significant way) lifted. Data are gathered from the latest version of this document (taking us up to 4 May). This leaves a total of 70 days observed, for 106 Italian provinces (out of 110, given the listwise deletion of some provinces for which quality of government and social capital data are not available, as they did not exist at the time1 ). From this source we computed New cases, the operationalization of , as the first difference between the cases of day and , and also YCases, the operationalization of , the absolute value of cases in province p at time .

The lockdown began on 12 March, as per the decree of the President of the Council of Ministers, announced the previous day. For this we built the dummy variable Lockdown (operationalization of DLD), which assumes the value of 1 in the days after 11 March, since this was the last day before Italy implemented a complete lockdown measure affecting the entire population. DLD is also computed for a given number of days that elapse after the lockdown, from 20 days after to 40 days after. Indeed, 97.5% of those who develop symptoms do so within 11.5 days of infection, which means a confidence interval of between 8.2 and 15.6 days (Lauer ). Considering that in Italy testing for COVID-19 was available only for people with symptoms, this suggests that lockdown gives greater benefits in terms of a recorded reduction in new cases only once a certain time has elapsed after the imposition of the policy. Fig. 1 presents COVID-19 incidence (total cases over the population) for the various Italian provinces on May 3.

According to the literature on the topic (Nannicini ), social capital theoretically refers to the diffusion of civic attitudes, and in particular to the portion of people who care about aggregate, and not individual, welfare. While this has been operationalized in different ways over time, and a number of different options have been used in the literature, here, following the approach of Tabellini (2009) and Nannicini , we measure social capital by provincial non-profit entities per 100,000 inhabitants in 2011 (the last year for which ISTAT provides data from a complete census of non-profit realities).2 More specifically, the concept is operationalized as a measurable variable by computing the number of non-profit realities (associations, foundations, charities) for every hundred thousand inhabitants of the province. A heat map presenting the varying intensity of social capital in the different Italian provinces is presented in Fig. 2 .

Fig. 2

. Social capital (proxied through non-profit entities per 100,000 inhabitants registered in 2011) and Institutional Quality (proxied through IQI in 2012) in Italian Provinces.

Finally, quality of government is proxied through the Institutional Quality Index (Nifo and Vecchione, 2014), a composite index inspired by the World Governance Indicator proposed by Kaufmann . We used the last IQI update (2019)3 to divide the sample.

Indeed, we believe that the IQI, which captures overall institutional quality (Nifo and Vecchione, 2014), may be a good proxy of the efficiency of the local government in implementing lockdown and ensuring citizens respect lockdown rules. It is worth noting that the IQI tries to capture the quality of public service and the policies formulated and implemented by the local government as well as individuals’ perception of law enforcement in terms of contract fulfilment, property rights and policing. Fig. 2 presents a heat map of IQI values of Italian provinces in 2019.

Please also consider that as these variables are computed annually, they are time invariant in our dataset, and thus we cannot include them as covariates in the regressions.

For this reason, we take into account the differences between Italian provinces in terms of social capital and institutional quality by splitting the sample into different subsamples on the basis of the percentile ranking of every province included in the analysis for each of these dimensions. More specifically, we divided the sample by considering observations belonging to the best quartile (equal to or over the 75th percentile) and the worst (below the 25th percentile).4

Considering that data have several observations for each p and t, the best estimator to employ is a Feasible–Generalized Least Square (F-GLS) (Aigner and Balestra, 1988; Hsiao, 1986). Moreover, given that the spread of the virus may be due to other factors (rather than the value of infections at t-1), which are specific to each province (such as the composition of the productive and entrepreneurial forces in a given province, which may affect the number of workers who continue to move into the territory5 ), we employed fixed effects (FE), which capture heterogeneity between provinces. In other words, we estimate the average effects with respect to a single province, assuming that the “structural” heterogeneity amongst provinces does not change in the 70 days of our timespan. In this way, we implicitly control for all the characteristics that do not vary in each province throughout the timespan, (such as median age, the female to male ratio, population density, to name only a few).

The final dataset is composed of 69 daily observations (for 70 days, i.e. from 24 February to 3 May; one observation is lost in order to compute the lag in the cases) in 106 Italian provinces, giving a total of 7314 observations. It is worth bearing in mind that our estimates will be split on the basis of percentile rank of each province included in the analysis for each of the governance and social capital dimensions. For this reason, the resulting subsample of the top and bottom quartile for social capital and institutional quality is composed of 27 provinces, giving a total of 1863 observations. Descriptive statistics of the variables used are provided in Table 1 .

Table 1

. Descriptive statistics.

Variable	Label	Obs	Mean	Std. Dev.	Min	Max	Source
New COVID-19 Cases (with respect to the previous day)	Newcases	7314	28.37818	59.57336	−186	868	Minister of Health – COVID 19 report
Number of cases reported the day before	YCases	7314	878.2987	1828.192	0	19,950	Minister of Health – COVID 19 report
Dichotomous variable, equal to 1 if dummy was in place on the day, to 0 otherwise	Dummy Lockdown	7420	.7571429	.4288384	0	1	Author elaboration from DPCM 11th March 2020
Dichotomous variable, equal to 1 if dummy was in place for at least 20 days, to 0 otherwise	20 days of Lockdown	7420	.4714286	.4992218	0	1	Author elaboration from DPCM 11th March 2020
Dichotomous variable, equal to 1 if dummy was in place for at least 30 days, to 0 otherwise	30 days of Lockdown	7420	.3285714	.4697302	0	1	Author elaboration from DPCM 11th March 2020
Dichotomous variable, equal to 1 if dummy was in place for at least 35 days, to 0 otherwise	35 days of Lockdown	7420	.2571429	.4370928	0	1	Author elaboration from DPCM 11th March 2020
Dichotomous variable, equal to 1 if dummy was in place for at least 38 days, to 0 otherwise	38 days of Lockdown	7420	.2142857	.4103578	0	1	Author elaboration from DPCM 11th March 2020
Dichotomous variable, equal to 1 if dummy was in place for at least 40 days, to 0 otherwise	40 days of Lockdown	7420	.1857143	.3889061	0	1	Author elaboration from DPCM 11th March 2020
Non-Profit Entities per 100,000 inhabitants	NoProfit	7420	475.1282	158.0965	116.773	915.5414	ISTAT No-Profit Census (2011)
Institutional Quality Index	IQI	7420	.5832757	.2422294	0	1	Nifo and Vecchione (2014)

Results and robustness tests

As can be seen in Tables 2 and 3 , lockdowns were more effective in Italian provinces with higher levels of social capital than they were in those with lower levels of social capital. 30 days into lockdown the former group had fewer new cases than provinces belonging to the latter group. This difference held, and even grew in magnitude, when looking at the effect 35, 38 and 40 days into lockdown.(Table 4 and 5 )

Table 2

. F-GLS-Best quartile of Social Capital.

	(2.1)	(2.2)	(2.3)	(2.4)	(2.5)	(2.6)
	Newcases
YCases	−0.00252⁎⁎	0.00667⁎⁎⁎	0.0124⁎⁎⁎	0.0123⁎⁎⁎	0.0121⁎⁎⁎	0.0115⁎⁎⁎
	(−2.23)	(4.54)	(9.15)	(9.72)	(10.00)	(9.75)
Dummy Lockdown	20.63⁎⁎⁎
	(14.36)
20 days of Lockdown		−1.216
		(−0.77)
30 days of Lockdown			−11.15⁎⁎⁎
			(−7.22)
35 days of Lockdown				−12.91⁎⁎⁎
				(−8.33)
38 days of Lockdown					−14.38⁎⁎⁎
					(−9.11)
40 days of Lockdown						−14.58⁎⁎⁎
						(−9.01)
Constant	2.499⁎⁎	14.22⁎⁎⁎	14.40⁎⁎⁎	14.10⁎⁎⁎	13.96⁎⁎⁎	13.90⁎⁎⁎
	(2.32)	(18.06)	(19.23)	(18.96)	(18.83)	(18.74)
Observations	1863	1863	1863	1863	1863	1863

t statistics in parentheses, *p < 0.1,.

p < 0.05,.

p < 0.01.

Table 3

. F-GLS-Worst quartile of Social Capital.

	(3.1)	(3.2)	(3.3)	(3.4)	(3.5)	(3.6)
	Newcases
YCases	0.00700⁎⁎⁎	0.00899⁎⁎⁎	0.00968⁎⁎⁎	0.00973⁎⁎⁎	0.00961⁎⁎⁎	0.00949⁎⁎⁎
	(10.73)	(13.03)	(14.26)	(14.52)	(14.44)	(14.32)
Dummy Lockdown	19.71⁎⁎⁎
	(8.92)
20 days of Lockdown		−3.841*
		(−1.95)
30 days of Lockdown			−10.59⁎⁎⁎
			(−5.14)
35 days of Lockdown				−13.01⁎⁎⁎
				(−5.97)
38 days of Lockdown					−13.48⁎⁎⁎
					(−5.85)
40 days of Lockdown						−13.44⁎⁎⁎
						(−5.55)
Constant	3.250*	18.87⁎⁎⁎	20.10⁎⁎⁎	19.92⁎⁎⁎	19.54⁎⁎⁎	19.23⁎⁎⁎
	(1.74)	(14.75)	(17.58)	(18.16)	(18.15)	(18.04)
Observations	1863	1863	1863	1863	1863	1863

t statistics in parentheses.

p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Table 4

. F-GLS-Best quartile of Institutional Quality.

	(4.1)	(4.2)	(4.3)	(4.4)	(4.5)	(4.6)
	Newcases
YCases	−0.0107⁎⁎⁎	−0.00219⁎⁎	0.000793	0.000711	0.000825	0.000447
	(−12.70)	(−2.10)	(0.82)	(0.77)	(0.93)	(0.51)
Dummy Lockdown	58.88⁎⁎⁎
	(19.12)
20 days of Lockdown		−1.186
		(−0.37)
30 days of Lockdown			−18.01⁎⁎⁎
			(−5.66)
35 days of Lockdown				−20.88⁎⁎⁎
				(−6.42)
38 days of Lockdown					−25.00⁎⁎⁎
					(−7.46)
40 days of Lockdown						−24.98⁎⁎⁎
						(−7.21)
Constant	16.45⁎⁎⁎	49.44⁎⁎⁎	50.37⁎⁎⁎	49.94⁎⁎⁎	49.76⁎⁎⁎	49.60⁎⁎⁎
	(7.09)	(27.71)	(29.56)	(29.52)	(29.56)	(29.45)
Observations	1863	1863	1863	1863	1863	1863

t statistics in parentheses, *p < 0.1,.

p < 0.05,.

p < 0.01.

Table 5

. F-GLS-Worst quartile of Institutional Quality.

	(5.1)	(5.2)	(5.3)	(5.4)	(5.5)	(5.6)
	Newcases
YCases	0.000847	0.00582⁎⁎⁎	0.0119⁎⁎⁎	0.0112⁎⁎⁎	0.0106⁎⁎⁎	0.0101⁎⁎⁎
	(0.64)	(3.74)	(8.09)	(7.92)	(7.69)	(7.50)
Dummy Lockdown	7.513⁎⁎⁎
	(9.46)
20 days of Lockdown		0.262
		(0.33)
30 days of Lockdown			−5.415⁎⁎⁎
			(−6.79)
35 days of Lockdown				−5.587⁎⁎⁎
				(−6.81)
38 days of Lockdown					−5.661⁎⁎⁎
					(−6.66)
40 days of Lockdown						−5.750⁎⁎⁎
						(−6.53)
Constant	0.815	5.465⁎⁎⁎	6.172⁎⁎⁎	5.975⁎⁎⁎	5.872⁎⁎⁎	5.812⁎⁎⁎
	(1.29)	(12.52)	(15.38)	(15.11)	(14.92)	(14.80)
Observations	1863	1863	1863	1863	1863	1863

t statistics in parentheses, *p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Meanwhile, when we compare the provinces in the best and worst quartiles for institutional quality, the effect is even bigger. Indeed, in the provinces belonging to the best quartile of the IQI (i.e. over the 75th percentile), the coefficients measuring the impact on Newcases of 30, 35, 38 and 40 days of lockdown are always greater in absolute value (given that we are talking about negative coefficients) than their respective coefficients in the bottom quartile (i.e. below the 25th percentile), which grows to be about five times greater.

This finding suggests that while both social capital and institutional quality matter when it comes to improving the effects of a lockdown, the latter plays a greater role than the former in enforcing the lockdown and making it efficient and effective. This may be read as being reflective of the greater importance of institutions and authorities in enforcing the policy, rather than people respecting it due to some inner desire to contribute to the common welfare. This means that both H1 and H2 are supported by the present analysis.

As a robustness check to strengthen our findings, we chose to replicate the analysis excluding all the provinces in Lombardy from the sample and adopting a different threshold. Lombardy is the Italian (and, possibly, the European) region that has been most severely affected by the pandemic. Indeed, as an obvious outlier in terms of infections, and at the same time a region characterized by communities with good levels of social capital and institutional quality, the provinces belonging to this region may influence the result, distorting the estimates and making generalization of the results debateable.

For this reason, we replicate the division in subsamples, as previously explained, after excluding all the provinces in Lombardy from the dataset. Furthermore, in order to provide more sensitive estimates, we chose a different threshold, and divided the sample into two subsamples, the first including all the observations above the median values of social capital and institutional quality, and the second with all the observations below. The results, presented in Tables 6 and 7 for social capital and in Tables 8 and 9 for institutional quality, confirm our previous findings, and show once again how the observation of the best subsample performs better than the worst, for both indexes. In this case too, institutional quality seems to have a greater effect on the reduction of contagion due to lockdown than social capital, even if the difference is narrower in this case (possibly also because the two subsamples being compared are closer, as the threshold chosen to separate the observations is the median).

Table 6

. F-GLS - Over the median of social capital (without provinces in Lombardy).

	(6.1)	(6.2)	(6.3)	(6.4)	(6.5)	(6.6)
	Newcases
YCases	−0.000558	0.00693⁎⁎⁎	0.0129⁎⁎⁎	0.0125⁎⁎⁎	0.0123⁎⁎⁎	0.0115⁎⁎⁎
	(−0.67)	(6.62)	(13.22)	(13.52)	(13.82)	(13.34)
Dummy Lockdown	23.20⁎⁎⁎
	(17.00)
20 days of Lockdown		−0.778
		(−0.54)
30 days of Lockdown			−14.40⁎⁎⁎
			(−10.07)
35 days of Lockdown				−15.69⁎⁎⁎
				(−10.85)
38 days of Lockdown					−17.36⁎⁎⁎
					(−11.73)
40 days of Lockdown						−17.13⁎⁎⁎
						(−11.24)
Constant	3.965⁎⁎⁎	17.35⁎⁎⁎	17.92⁎⁎⁎	17.51⁎⁎⁎	17.33⁎⁎⁎	17.24⁎⁎⁎
	(3.82)	(23.24)	(25.43)	(25.01)	(24.84)	(24.67)
Observations	3381	3381	3381	3381	3381	3381

t statistics in parentheses, *p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Table 7

. F-GLS-Under the median of social capital (without provinces in Lombardy).

	(7.1)	(7.2)	(7.3)	(7.4)	(7.5)	(7.6)
	Newcases
YCases	0.00964⁎⁎⁎	0.0136⁎⁎⁎	0.0149⁎⁎⁎	0.0147⁎⁎⁎	0.0146⁎⁎⁎	0.0143⁎⁎⁎
	(17.74)	(22.99)	(26.16)	(26.27)	(26.60)	(26.23)
Dummy Lockdown	14.21⁎⁎⁎
	(12.17)
20 days of Lockdown		−6.832⁎⁎⁎
		(−6.38)
30 days of Lockdown			−13.63⁎⁎⁎
			(−12.46)
35 days of Lockdown				−14.78⁎⁎⁎
				(−12.84)
38 days of Lockdown					−16.60⁎⁎⁎
					(−13.77)
40 days of Lockdown						−16.64⁎⁎⁎
						(−13.18)
Constant	2.497⁎⁎⁎	14.50⁎⁎⁎	15.05⁎⁎⁎	14.49⁎⁎⁎	14.28⁎⁎⁎	13.97⁎⁎⁎
	(2.58)	(21.95)	(25.50)	(25.34)	(25.41)	(25.00)
Observations	3174	3174	3174	3174	3174	3174

t statistics in parentheses, *p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Table 8

. F-GLS-Over the median of IQI (without provinces in Lombardy).

	(8.1)	(8.2)	(8.3)	(8.4)	(8.5)	(8.6)
	Newcases
YCases	0.00468⁎⁎⁎	0.0131⁎⁎⁎	0.0161⁎⁎⁎	0.0155⁎⁎⁎	0.0153⁎⁎⁎	0.0146⁎⁎⁎
	(7.61)	(17.74)	(23.36)	(23.58)	(24.19)	(23.49)
Dummy Lockdown	27.42⁎⁎⁎
	(17.57)
20 days of Lockdown		−11.34⁎⁎⁎
		(−7.16)
30 days of Lockdown			−24.06⁎⁎⁎
			(−15.42)
35 days of Lockdown				−25.64⁎⁎⁎
				(−16.03)
38 days of Lockdown					−28.63⁎⁎⁎
					(−17.42)
40 days of Lockdown						−28.25⁎⁎⁎
						(−16.57)
Constant	5.574⁎⁎⁎	24.20⁎⁎⁎	24.06⁎⁎⁎	23.27⁎⁎⁎	22.95⁎⁎⁎	22.69⁎⁎⁎
	(4.60)	(28.07)	(30.35)	(29.72)	(29.59)	(29.17)
Observations	3243	3243	3243	3243	3243	3243

t statistics in parentheses, *p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Table 9

. F-GLS-Under the median of IQI (without provinces in Lombardy).

	(9.1)	(9.2)	(9.3)	(9.4)	(9.5)	(9.6)
	Newcases
YCases	0.00685⁎⁎⁎	0.0113⁎⁎⁎	0.0157⁎⁎⁎	0.0152⁎⁎⁎	0.0147⁎⁎⁎	0.0142⁎⁎⁎
	(6.66)	(9.69)	(14.00)	(13.97)	(13.86)	(13.59)
Dummy Lockdown	8.823⁎⁎⁎
	(9.26)
20 days of Lockdown		−1.048
		(−1.15)
30 days of Lockdown			−8.095⁎⁎⁎
			(−8.69)
35 days of Lockdown				−8.547⁎⁎⁎
				(−8.83)
38 days of Lockdown					−8.906⁎⁎⁎
					(−8.84)
40 days of Lockdown						−8.850⁎⁎⁎
						(−8.44)
Constant	0.949	7.001⁎⁎⁎	7.966⁎⁎⁎	7.648⁎⁎⁎	7.477⁎⁎⁎	7.345⁎⁎⁎
	(1.23)	(13.31)	(16.66)	(16.35)	(16.13)	(15.90)
Observations	3243	3243	3243	3243	3243	3243

t statistics in parentheses, *p < 0.1, ⁎⁎p < 0.05,.

p < 0.01.

Conclusions

Since the adoption of isolation measures in the Venetian Republic during the Great Plague of the seventeenth century (Alfano and Sgobbi, 2021), lockdown measures have seemed to be one of the most effective policies for containing the diffusion of pandemics, especially in the absence of a cure or a vaccine. For this reason, it is important to try to shed some light on which social and institutional factors can influence the efficacy of such measures.

In this perspective, the present contribution, adopting a country-level analysis, has tried to investigate how different endowments of institutional quality, as proposed by Kaufmann , and social capital as suggested by Putnam (1995), are associated with greater efficiency and efficacy in lockdown.

Using the Institutional Quality Index built for Italy by Nifo and Vecchione (2014) and some proxies of social capital employed in several seminal previous contributions (Guiso; Tabellini, 2009; Nannicini ), we found that both these dimensions are associated with a higher reduction in new cases of COVID-19. Moreover, we found that the Institutional Quality Index seems to be related to a higher effect on the efficacy of lockdown, when compared to that due to social capital.

From a theoretical perspective, our findings seem to suggest that social and – more especially – institutional factors, incorporating the quality of public service and policies formulated by local government as well as individuals’ perception of law enforcement in terms of contract fulfilment, property rights and policing, contribute to forming the set of incentives able to promote individual behaviour that is in closer compliance with lockdown measures. In this perspective, on the basis of the described mechanism, institutional quality and social capital can be considered as factors able to influence the efficacy of policies. These findings add to the recent contribution made by Rodriguez-Pose (2020) that highlights how institutional factors are central in determining differences amongst territories, not only in terms of economic outcomes, but more generally regarding the effectiveness of policies. Moreover, our contribution, using lockdown measures as a case study, with the advantage of their unexpected nature, also sheds some light on the possible transmission mechanism between institutions and policy effectiveness which, according to Rodriguez-Pose (2020), requires more attention.

From a policymaker's perspective, although institutions are often described as path dependant, various contributions also suggest that social capital and institutional quality may be considered as factors capable of effecting change (Rodríguez-Pose and Storper, 2006; Rodríguez-Pose and Ketterer, 2019).

Nevertheless, it is worth noting that these factors are the results of long-term policies. From a policymaker's perspective, this result suggests that institutional factors, unlike short-term political choices, which are often determined by the electoral cycle, could be an important resource for local governments in order to confront unexpected shocks.

Lockdown measures as a response to the COVID-19 pandemic, because of the random and exogenous nature of such a shock, are a challenging opportunity for researchers to investigate the links between these dimensions and policy efficacy. Further studies may focus on the effects that different dimensions of social capital have on lockdowns, or replicate the analysis with a different country to provide further evidence on this relationship, thereby helping to generalize the main findings of the present paper.

CRediT authorship contribution statement

Vincenzo Alfano: Conceptualization, Methodology, Data curation, Software, Writing – original draft, Writing – review & editing, Investigation, Formal analysis. Salvatore Ercolano: Conceptualization, Methodology, Writing – original draft, Writing – review & editing, Validation, Project administration.