Draft for the PAA annual meeting. Please do not quote or circulate

The Impact of Gynecologists’ Conscientious Objection on Access to in

Tommaso Autorino ∗ ,† Francesco Mattioli *,‡ Letizia Mencarini*,§

April 2018

Abstract Abortion in Italy is free of charge and legal in a broad set of circumstances, but 70.9% of gynecologists refuse to partake in procedures leading to the voluntary interruption of on grounds of personal beliefs. We assess whether conscientious objection is linked to the inter- regional migration of women seeking an abortion. We first perform a panel data analysis at the regional level, showing that a higher prevalence of objecting professionals is associated to a higher share of women having an abortion outside the region where they reside. Secondly, analyzing individual data, we find that conscientious objection is a significant driver of the personal decision to move across regions to obtain an abortion. Overall, results suggest that conscientious objection is significantly associated to abortion-related migration.

1 Introduction

Italian law permits abortion in a broad set of circumstances, while granting healthcare personnel the right to abstain from performing for reasons of conscientious objection. Most European countries allow conscientious objection to abortion, but the Italian case is noteworthy for the well-documented prevalence of the phenomenon, involving more than 70% of gynecologists nationwide, with peaks above 85% in some regions. Conscientious objection is a topic of lively debate in Italy, with opponents arguing that the lack of non-objecting personnel limits access to abortion, imposing a constraint onto the individual choice of how to resolve pregnancy. On the other hand, the Italian Ministry of Health – commenting 2016 data – stated that the number of abortion providers is sufficient to grant easy access to the service and any difficulty in this respect is imputable to organizational shortcomings at the local level. Indeed, there is no

∗Carlo F. Dondena Centre for Research on Social Dynamics and Public Policy, Bocconi University - Milan, Italy †[email protected] ‡[email protected] §[email protected]

1 empirical evidence that confirms or rebuts a causal relation between conscientious objection and demand for abortions in Italy, but the high share of women obtaining an abortion outside of their region of residence suggests that abortions might be harder to obtain in certain parts of the country.

Many studies that provide an economic characterization of abortion indicate the availability of providers and the costs of obtaining an abortion (including travel costs) as important drivers of abortion demand. Drawing on this literature, our study aims at assessing whether conscientious objection induces women to travel in order to find an abortion provider. We analyze data at the regional level in the period between 2002 and 2016, and find that on aggregate widespread consci- entious objection is associated with more women travelling out of their region in order to obtain an abortion. This finding is confirmed with the analysis of individual data on abortions recorded in Italy between 2002 and 2014, showing that women residing in regions where conscientious objectors are more numerous are more likely to obtain an abortion in another region.

2 Previous Research

Much of the literature that aims at estimating the impact of abortion regulation on abortion rates rests on an economic model of fertility control. Within this framework, abortion is treated as a contraceptive method of last resort and the two successive decisions of whether to conceive and whether to carry a pregnancy to term respond to an array of economic variables that weigh the cost of contraception versus the opportunity cost of an additional child. The cost of obtaining an abortion encompasses the price of the medical procedure as well as indirect costs, such as the time and travel devoted to find an abortion provider, and psychic costs that may arise depending on individual and societal attitudes towards abortion. Policy that regulates abortion can affect each of these cost components, changing the terms of the individual decision on how to resolve a pregnancy, potentially resulting in large-scale impact on abortion and birth rates.

Many studies look at cross-state differences in abortion regulation in the USA. Medoff (2013) provides an extensive review of this literature, while here we present only some relevant examples. Medoff (1988) finds that abortion demand is sensitive to price, restrictions of public funding of abortion1 and socioeconomic variables. The negative impact of state funding restrictions on abortion rates is confirmed by Levine, Trainor and Zimmerman (1996) and Cook, Parnell, Moore and Pagnini (1999). This strand of research shows that even minor variations in the abortion price can affect the decision of having an abortion, but such findings are hardly applicable to the case of Italy, where in 2016 94% of abortions were performed free of charge within the public healthcare system.

However, price is not necessarily the most relevant determinant of the decision to seek an abortion. Gober (1994) criticizes Medoff (1988) for disregarding cross-state differences in terms of abortion regulation and availability of abortion providers, and suggests that abortion rates should be interpreted as a measure of availability of abortion, rather than a measure of demand. A number of studies evaluate the impact of provider availability in the USA, focusing on the link

1Following the liberalization of abortion in the USA, abortions were initially funded in all states under the Medicaid program. Later on, abortion was excluded from federal funding, but a number of states continued to subsidize abortions at their own expense.

2 between abortion supply, travel costs and abortion demand. Deyak and Smith (1976) study this aspect in a pre-liberalization setting2, analyzing individual data of women who travelled to New York to obtain an abortion. Results suggest that a state’s distance from New York significantly affected the likelihood that state residents would seek an abortion. Brown and Jewell (1996) and again Brown, Jewell and Rous (2001) obtained similar results with individual data from Texas, showing that women residing further away from an abortion provider were less likely to obtain an abortion.

Scarce abortion supply may be associated with other inconveniences beyond travel costs, such as costs for overnight lodging, absence from work, privacy concerns and difficulty to obtain information and post-abortion care (Haas-Wilson, 1993). Provider availability may affect access to abortion also through other channels, such as signaling social acceptance of abortion and discouraging the use of other means of contraception (Brown and Jewell, 1996). To assess the impact of provider availability, some models incorporate a variable measuring abortion supply directly, rather than through travel costs. For example, Haas-Wilson (1997) finds that the number of abortion providers per 1,000 women in childbearing age is positively associated with abortion rates and concludes that women travel to obtain an abortion in states with greater supply. Matthews, Ribar and Wilhelm (1997) find that provider availability contributes to explain variations in abortion rates overtime, but conclude that the major determinant of abortion demand is the demographic structure of the population. Gius (2007) combines individual-level data on , abortions and socioeco- nomic status with state-level data on abortion providers and legal restrictions to abortion, finding that provider availability has a statistically signicant effect on abortion rates, while legal restrictions have no significant impact. Blank, George and London (1996) make an important distinction between two alternative measures of abortion demand at the state level: the abortion rate by state of occurrence, based on the number of abortions taking place in each state, and the abortion rate by state of residence, which instead counts abortions from women residing in each state. Their results indicate that provider availability is a significant determinant of the first measure only, sug- gesting that abortion supply does not determine whether a woman will have an abortion, but where.

In general, the literature supports an economic interpretation of abortion demand and indicates that direct and indirect costs of abortion, together with demographic and socioeconomic character- istics of the population, affect women’s contraceptive behaviour and abortion rates. The availability of abortion providers often appears as a major determinant of abortion use, although some studies reach different conclusions. For instance, Medoff (2002) uses an index of restrictiveness of a state’s abortion policy and finds that this measure does not contribute to explain cross-state variations in abortion rates or out-of-state abortions. Medoff (2010) examines TRAP laws – i.e. laws that impose cumbersome requirements and a licensing fee on abortion providers – and concludes that these do not affect abortion rates, despite curbing abortion supply. Importantly, a number of studies suggest that disparities across states in terms of abortion regulation and supply impact on states’ abortion rates not by preventing women from having an abortion, but inducing them to move where it is easier to obtain one.

Conscientious objection to abortion is rarely mentioned in this literature. Meier, Haider-Markel,

2Abortion was liberalized nationwide in the USA in 1973, following the Supreme Court decision in Roe v. Wade. Between 1970 and 1973, New York was one of four states allowing abortions and the only state where a woman could obtain one without being a resident.

3 Stanislawski and Mcfarlane (1996) analyze the impact of 23 different state-level abortion restrictions put in place in the USA, finding that only suspension of public funding has an impact on abortion use. One of the restrictive policies examined and found to be irrelevant consists of a “conscience clause” that allows physicians to refuse performing abortions if they are contrary. However, the model incorporates only a dummy variable that indicates the existence of this clause and not a measure of physicians’ use of the conscience clause. As for Italy, little research exists on demand for abortion and the impact of conscientious objection. Bo, Zotti and Charrier (2015) find some correlation at the regional level between the workload of non-objecting gynecologists and waiting times for obtaining an abortion. We will extend their analysis, exploiting data from the same source (the annual report of the Ministry of Health on the implementation of Law 194) in order to delve deeper into the relationship between conscientious objection and abortion use in Italy. In a second step, we will resort to individual data to study the impact of conscientious objection on access to abortion.

3 Abortion in Italy3

Abortion in Italy is regulated by Law 194 of 1978 “on the social protection of motherhood and the voluntary termination of pregnancy”4. In the first 90 days of pregnancy, abortion is permitted in a broad set of circumstances, namely whenever pregnancy, birth or motherhood could undermine the mother’s physical or mental health – given her health, economic, social or familial conditions, the circumstances of conception, or the anticipation of anomalies of the child5. Article 9 of Law 194 also grants the healthcare personnel the right to refuse to perform abortions for reasons of conscientious objection, provided that this personal stance is declared formally and with adequate notice. Objectors are not required to refer women who intend to have an abortion to non-objectors, but Article 9 mandates regions to grant adequate access to abortion at the local level, including by means of staff mobility.

3Data presented in this section are found in the 2017 Annual Report of the Italian Minister of Health on the implementation of Law 194. 4Law No. 194, published in the Gazzetta Ufficiale N. 140, 22 May 1978. 5After the first trimester, abortion is allowed when childbearing may severely endanger the woman’s health.

4 Figure 1: Abortion Rates in Italian Regions, 2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2017). Within the region of Trentino-Alto Adige, the Autonomous Province of Bolzano shows an abortion rate of 4.8, whereas the Autonomous Province of Trento an abortion rate of 5.9.

5 Figure 2: Abortion Rates in NUTS-1 Areas, 2002- 2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2004-2017).

Following the liberalization of abortion in 1978, the abortion rate peaked to 17.2 cases per thousand women in childbearing age (15-49) in 1982, then declined until reaching 6.5 in 2016. The abortion ratio sharply declined overtime as well, from 380.2 voluntary abortions per thousand live births in 1982 to 182.4 in 2016. The abortion rate for women aged 15-44 in Italy is 8.0, higher than in Switzerland (6.3) and Germany (6.8) but lower than in Spain (10.4), the USA (14.6), England and Wales (16.0) and France (18.1)6. However, there is considerable heterogeneity in terms of abortion rates at the regional level. In 2016, abortion rates ranged from 4.5 cases per thousand women in childbearing age in Basilicata to 8.8 in . As shown in Figure 1, abortion rates tend to be higher in the North West (7.0) and in the Centre (6.9) of Italy, and lower in the North East (6.3), South (6.2) and particularly in the Islands (5.4)7. Such differences have existed for years and have narrowed overtime as abortion rates declined in all parts of the country (2). Abortion rates are higher in age groups between 20 and 34, lower for women aged 30-39, and particularly low amongst women younger than 20 and older than 40 (3). In 2015, the abortion rate for Italian citizens was 5.7, compared to 15.7 for women of foreign citizenship.

6This comparison is based on the most recent available figures reported by the Ministry of Health. Figures refer to 2016 for England and Wales, Germany, Italy and Switzerland, 2015 for Spain, 2014 for the USA, and 2013 for France. 7Here we adopt the NUTS 1 classification that groups Italian regions as follows: North West (Aosta Valley, Liguria, Lombardy, Piedmont), North East (Emilia-Romagna, Friuli-Venezia Giulia, Trentino-Alto Adige/S¨udtirol, Veneto), Centre (Lazio, Marche, , Umbria), South (Abruzzo, , Basilicata, Calabria, Campania, Molise), and Islands (Sardinia, ). NUTS 1 figures are calculated by the authors, based on regional figures.

6 Figure 3: Abortion Rate by 5-year Age Group, 2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2017).

The regional abortion rate presented above, counting abortions in the region where they occur, might be a misleading indicator of abortion demand from a region’s population. In fact, some of the abortions performed in a region are requested by women residing elsewhere, while some women residing in a given region obtain abortions elsewhere. An alternative measure of abortion use is the abortion rate by region of residence, i.e. the number of abortions obtained by women who reside in a given region – regardless of where the abortion occurs – per thousand women in childbearing age residing in that region. Substantial percentage differences between the two measures (close to 10% in some cases, with peaks of 14% in Molise and 20% in Basilicata) indicate that it is relatively common for a woman to obtain an abortion out of her region of residence, and this may signal difficulties in finding abortion providers in some parts of the country. However, other factors – including social stigma related to local attitudes towards abortion – may induce women to seek an abortion out of their region and create a discrepancy between the two measures. Another source of divergence is interregional migration, as those who move across regions to study or work do not necessarily change their official place of residence. If interregional migration were concentrated in age groups where abortions were more frequent, its impact on abortion rates would be significant. The difference between abortions by region of residence and abortions by region of occurrence, divided by abortions by region of occurrence provides an indicator of the net outflow of women seeking an abortion, a “migratory balance of abortions”. If the measure is positive (negative), more (less) residents of the region have obtained an abortion out of the region than residents of other regions have in the region. In general, this measure is larger for Southern regions, while most Central and Northern regions have an inflow of women seeking abortions (4).

7 Figure 4: Migratory Balance of Abortions, 2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2017). Within the region of Trentino-Alto Adige, the Autonomous Province of Bolzano shows a net migration rate of -3.2%, whereas the Autonomous Province of Trento is at -9.4%.

The legal obligation for conscientious objectors to declare their stance formally if they wish to be exempted from performing abortions allows the Ministry of Health to collect accurate data on conscientious objection. The percentage of conscientious objectors among Italian gynecologists remained below 60% between 2002 and 2005, then jumped to 69% in 2006, mainly in response to a sudden increase in the number of objectors in the South and in the Islands. Since 2006, the share of objectors stabilized around 70% and in 2016, 71% of Italian gynecologists were objectors. Considerable heterogeneity across regions exists also in this case (Figure 5), with the percentage of objectors ranging from 18% in Aosta Valley to 97% in Molise. With the exception of Aosta Valley and Emilia-Romagna (where 48% of gynecologists are objectors), the majority of gynecologists in each region are objectors. Objection is less common in the North East (60%), more in the North West (67%) and in the Centre (70%), and particularly widespread in the Islands (78%) and in the South (84%) (6). Anesthetists and non-medical staff of gynecology and obstetrics wards can also refuse to attend to abortions. In 2016, the percentage of objectors in these categories was 49% and 44% respectively. At the regional level, the share of objectors is highly correlated across professional categories.

8 Figure 5: Percentage of Objecting Gynecologists in Italian Regions, 2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2017). Within the region of Trentino-Alto Adige, the Autonomous Province of Bolzano records 84.4% of objecting gynecologists, whereas in the Au- tonomous Province of Trento the share is equal to 58.7%.

9 Figure 6: Percentage of Objecting Gynecologists in NUTS-1 Areas, 2002-2016

Source: Annual Report of the Italian Minister of Health on the implementation of Law 194 (2004-2017).

4 Data and Methodology

After the Law on the voluntary termination of pregnancy came into force in 1978, the Italian Institute of Statistics (Istat) launched the survey on voluntary terminations of pregnancy to collect information on the aborting women, in coordination with the Italian regions and autonomous provinces, the Italian Ministry of Health and the Italian Institute of Health. As the aim of the survey is to gain better knowledge of the phenomenon and to understand how to prevent it, the collection focuses on a series of sociodemographic information about women, on the services involved in authorizing and conducting abortions, and on technical details of the operations. Hospitals and other health facilities take care of the data collection at the individual level. The survey unit is any episode of voluntary interruption of pregnancy, whose characteristics are surveyed through a template 8 filled and signed by the physician who performs the abortion. The regions monitor the collection process and receive the individual data from the health facilities. The Italian Institute of Health is entitled to check the data quality. Eventually, the data is transmitted to Istat, which is responsible for the data management. Once aggregated, Istat publishes the data on its online data warehouse, while making individual data available on request for research and statistical analyses, with due regard for the protection of sensitive personal information.

On an annual basis, Istat elaborates the data and creates tables showing the regional frequencies of abortions by sociodemographic characteristics of the women and by features of the operation. According to article 16 of Law 194/78, every year the Minister of Health presents to the a report that addresses the enforcement status of the Law and highlights the trends of voluntary abortion in Italy, attaching the aforementioned tables. In this paper, we use data retrieved from the ministerial reports, which are publicly available on the Ministry’s website9 and present data referring to two years before (the latest report, from 2018, presents data from 2016).

8http://www.istat.it/ws/fascicoloSidi/263/Modello%20D12.pdf. 9http://www.salute.gov.it/

10 The same data source has already been employed by Bo et al. (2015). The reports include regional figures on abortion and abortion rates, broken down by women’s characteristics (age, marital status, education, labor force status, citizenship), place of residence (the same region where the abortion takes place, another region, abroad), previous history of pregnancies and abortions, and technical details of the operation (when and by whom the entitlement certification is released, the urgency of abortion, the type of facility where it takes place, the type of anesthesia and operational technique, complications). Numbers and percentages of conscientious objectors among gynecologists, anes- thetists and non-medical personnel are also included in the reports, as notified by regions. In light of the prevalence of conscientious objection and of the significant share of abortions taking place outside of the women’s region of residence, the aim of our analysis is to estimate the relationship between these two variables, to assess whether an association exists between the two.

4.1 Region-level Analysis In order to establish whether there is a relationship between conscientious objection and out-of- region abortions, we use as a dependent variable a measure of abortion demand that takes into account inter-regional movements by women seeking an abortion provider. One such measure is the “migratory balance of abortions”, the difference between the number of abortions by the region’s residents and the number of abortions that take place in the region, expressed as a percentage of the latter. We also combine data on women’s residence to calculate for each region two alternative dependent variables that measure inter-regional travel. One is the “emigration for abortion” – i.e. the number of abortions obtained by residents of region A outside region A, per thousand women in childbearing age residing in region A – measuring the outflow of women seeking an abortion. The other is the “immigration for abortion” – i.e. the number of abortions performed in region A on women residing outside region A, per thousand women in childbearing age residing in region A – measuring the inflow of women seeking an abortion. The ministerial tables specify in how many cases the woman’s residence was not recorded: we drop observations where the percentage of abortions with non-recorded origin is higher than 10% 10.

Percentages of conscientious objection by professional category represent our main independent variables. Being the percentage of objectors highly correlated across categories, we estimate the impact of conscientious objection in each category separately, and with a composite indicator that captures the prevalence of conscientious objection in the healthcare environment. The indicator is derived from the first principal component of the three variables, and is highly reliable overtime (the average value of Cronbach’s alpha is 0.84). We control for the workload of non-objectors (the average number of abortions per non-objector) to account for the possibility that the reduction in abortion supply induced by conscientious objection is compensated by a higher number of abortions performed by non-objectors.

Following the reviewed literature, we enrich the analysis with measures that describe the socioeconomic context of Italian regions throughout the period, which are likely to drive abortion demand when an economic model of fertility control is posited. GDP per capita at current market prices is taken as proxy for income, while the labor market situation is accounted for through the unemployment rate among women aged more than 14. The incidence of economic disadvantage

10Results remain very similar by moving upwards or downwards this threshold, and with or without imputing the abortions with unrecorded origin to migration from other regions.

11 is controlled for by including the share of households in relative poverty. A proxy for the degree of religious observance is provided by the share of population declaring to have never attended a place of worship during the year. Given the diverging abortion behaviors followed by and foreigners, the regions’ share of women in childbearing age with foreign citizenship is also included. These data are retrieved from the Istat online data warehouse 11. In the study of conscientious objection vis-`a-vis abortions performed in or out-of-region, we control for the number of live births per thousand women in childbearing age, aiming at disentangling the confounding effects brought about by the general fertility patterns, which might be related with abortion trends.

We conduct a panel data analysis on 21 Italian regions 12 spanning the period 2002 to 201613. The region-level models presented in this section are OLS regressions including regional and year fixed effects to account for unobserved time-invariant regional characteristics and for trends overtime that are common to all regions. The estimated standard errors are robust to heteroscedasticity and allow for arbitrary intra-region correlation.

4.2 Individual-level Analysis Abortion rates aggregated at the regional level do not allow deepening the analysis by taking into consideration individual heterogeneity. Women differing in terms of intrinsic characteristics might be more or less inclined to move across regions in order to find an abortion provider. Accounting for individual heterogeneity is mandatory both to draw a socio-demographic profile of those women who abort out of their region, and to test whether and to what extent the relation between conscientious objection and out-of-region abortions is driven by these characteristics. The survey on the voluntary interruptions of pregnancy is a source of information at the individual level. The survey collects data on women’s age, place of birth, citizenship, marital status, education and employment status.

The survey also covers women’s place of residence and where the abortion took place: combining these data provides an indicator of whether a woman traveled to a different region to obtain an abortion. We create a binary variable taking value one in case the woman’s region of residence is not the same where the abortion takes place, and zero otherwise. We will use this dummy as a dependent variable in an individual-level probit model. To the best of our knowledge, we are the first to present a quantitative individual-level analysis of the phenomenon in Italy. The dataset employed contains information on all women who obtained an abortion in Italy over the period 2002 to 201414: more than a million observations are therefore available for the analysis.

We model the probability that each woman moves to a different region to obtain an abortion as a function of her individual socio-demographic characteristics. The regressors at the regional level are also introduced in the model, allowing to estimate the strength of the relationship between region-wide conscientious objection and the individual decision to travel to other regions in order to abort. By adding year fixed effects to the probit model we are able to isolate time patterns affecting

11http://dati.istat.it/ 1219 Italian regions (NUTS-2) plus 2 autonomous provinces of Bolzano and Trento (NUTS-3), referring to the region of Trentino Alto-Adige. 13We consider 14 years, with the exception of year 2004 for lacking data on religious observance. 14Year 2004 is excluded from the analysis due to the lack of data on religious observance.

12 conscientious objection and inter-regional migration by aborting women. Moreover, the addition of fixed effects capturing region of birth, residence and operation minimizes the scope for selection into inter-regional migration based on women’s origin or destination for reasons other than seeking an abortion. The standard errors in the model are robust to heteroscedasticity and clustered by region of residence.

5 Results

Table 1 summarizes the regressions of the net outflow of women in search of an abortion on conscientious objection by different categories of professionals and key control variables. The dependent variable is the “migratory balance of abortions”: the difference between the number of abortions by women residing in a certain region, regardless of the place where they occur, and abortions obtained in the same region by women residing anywhere (in the region, out of the region, or abroad), expressed as a percentage of the latter. This variable is a convenient measure of abortion-related travel: a positive (negative) value for a region implies that abortions by the region’s residents are more (less) than the abortions performed in the region. Columns 1 to 4 suggest that widespread conscientious objection is positively associated, on aggregate, with a positive net outflow of women seeking an abortion. Percentages of objectors appear to have a significant impact across all specifications (except the one amongst non-medical staff) indicating that in regions with more objectors the outflow of women seeking an abortion is greater than the inflow. The results hold when objection by different categories is collapsed into a unique indicator in column 4. Columns 5 to 8 exclude abortions by women residing abroad. The coefficients increase in magnitude, while the objection by non-medical personnel gains significance. This may indicate that women living in Italy are more informed than women living abroad about the incidence of conscientious objections across regions, and choose where to abort accordingly (less Italian women choose to abort where the share of objectors is higher).

As expected, non-objectors’ workload is always negatively related to the net outflow of aborting women, as busier non-objectors conduct more abortions in the region where they operate, absorbing more abortion demand. GDP per capita is strongly associated to lower (possibly negative) net outflows. This result can be interpreted hypothesizing that richer regions attract more migrants from other regions and from abroad, including women in childbearing age, resulting in more abortions by women who migrated there but are officially resident elsewhere. The share of foreign women is related to a higher net outflow of women travelling to obtain an abortion, possibly because of the greater abortion use by women of foreign citizenship. The fertility rate has a positive and significant association with the dependent variable, while other socioeconomic control variables do not show any statistically significant impact.

The previous findings are largely confirmed when switching to dependent variables that measure separately the outflow and the inflow of women seeking an abortion provider. Columns 1 to 4 of Table 2 have the number of residents’ abortions obtained out of region per thousand resident women in childbearing age as their dependent variable. In particular, the objection by anesthetists and non-medical personnel is associated with more pronounced outflows. By symmetry, negative coefficients are attached to conscientious objection when the dependent variable counts the number of abortions obtained in the region by women residing in other regions, excluding those residing abroad. As can be noticed in columns 5 and 7, more widespread objection by gynecologists

13 Table 1: Abortion Migratory Balance and Conscientious Objection

Abortion migratory balance Abortions by residents abroad included Abortions by residents abroad excluded (1) (2) (3) (4) (5) (6) (7) (8)

% Objecting gynaecologists 0.214*** 0.226*** (0.0684) (0.0688) Workload by non-obj. gynaecologists -0.0343** -0.0357** (0.0144) (0.0135) % Objecting anesthetists 0.306** 0.324*** (0.112) (0.106) Workload by non-obj. anesthetists -0.105*** -0.110*** (0.0369) (0.0367) % Objecting non-medical personnel 0.186 0.219* (0.108) (0.106) Workload by non-obj. non-medical personnel -0.0579** -0.0645** (0.0257) (0.0247) Indicator of conscientious objection 0.339** 0.385** (0.158) (0.156) Indicator of workload by non-objectors -0.0677** -0.0735** (0.0285) (0.0274) GDP per capita -3.193** -2.801** -2.920** -2.956** -2.825** -2.413** -2.513** -2.561** (1.235) (1.173) (1.205) (1.228) (1.235) (1.155) (1.161) (1.198) Households in poverty -0.147 -0.134 -0.0903 -0.136 -0.195 -0.182 -0.139 -0.188 (0.293) (0.322) (0.326) (0.298) (0.284) (0.316) (0.319) (0.290) Female unemployment rate 0.514 0.488 0.379 0.596 0.757 0.732 0.601 0.844 (0.574) (0.598) (0.603) (0.654) (0.544) (0.566) (0.569) (0.616) Lack of religiosity 0.211 0.178 0.151 0.208 0.353 0.318 0.307 0.368 (0.261) (0.234) (0.242) (0.244) (0.266) (0.241) (0.246) (0.244) Share of foreign women 15-49 1.149 1.039 0.860* 0.955 0.754 0.636 0.492 0.603 (0.675) (0.607) (0.483) (0.572) (0.722) (0.650) (0.490) (0.588) Share of married women -0.951 -1.583 -1.262 -1.515 -0.358 -1.027 -0.624 -0.892 (0.793) (0.965) (1.053) (1.006) (0.734) (0.908) (0.999) (0.940) Fertility rate 0.948* 1.238** 1.030* 1.059* 0.964* 1.270** 1.000* 1.035* (0.490) (0.504) (0.542) (0.533) (0.472) (0.477) (0.511) (0.503)

Observations 261 261 254 254 261 261 254 254 Adjusted R2 0.78 0.79 0.78 0.79 0.76 0.77 0.77 0.77 Region and Year FE Yes Yes Yes Yes Yes Yes Yes Yes Note: for each regional observation, the dependent variable is the difference between the number of abortions obtained anywhere by women residing in that region, and the number of abortions obtained in that region by women residing both in the region and either out of the region (in different italian regions and abroad - Columns 1 to 4) or only in different italian regions (Columns 5 to 8), expressed as a percentage of total abortions obtained in the region. Robust standard errors in parentheses, clustered by region. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.

14 and non-medical personnel pre-empts inflows from other regions. In both cases, the composite indicator follows the expected direction, highlighting that conscientious objection is a region-wide phenomenon involving all professional categories.

Table 2: Abortion Rates of Women Travelling Across Regions and Conscientious Objection

Rate of emigration for abortion Rate of immigration for abortion (1) (2) (3) (4) (5) (6) (7) (8)

% Objecting gynaecologists 0.00551 -0.00880** (0.00453) (0.00393) Workload by non-obj. gynaecologists -0.00129*** 0.00108 (0.000323) (0.000911) % Objecting anesthetists 0.0129*** -0.0139 (0.00453) (0.00809) Workload by non-obj. anesthetists -0.00411*** 0.00373 (0.00101) (0.00250) % Objecting non-medical personnel 0.00735** -0.00549*** (0.00352) (0.00185) Workload by non-obj. non-medical personnel -0.00280*** 0.000911** (0.000792) (0.000343) Indicator of conscientious objection 0.0124** -0.0128** (0.00582) (0.00588) Indicator of workload by non-objectors -0.00276*** 0.00203 (0.000757) (0.00137) GDP per capita -0.0954*** -0.0800*** -0.0843*** -0.0850*** 0.0236 0.00854 0.0128 0.0146 (0.0233) (0.0230) (0.0245) (0.0240) (0.0717) (0.0748) (0.0718) (0.0731) Households in poverty 0.00613 0.00655 0.00830 0.00707 0.0312 0.0317 0.0291 0.0316 (0.0106) (0.0114) (0.0106) (0.0106) (0.0215) (0.0214) (0.0219) (0.0219) Female unemployment rate 0.0272 0.0323 0.0217 0.0292 -0.0158 -0.0183 -0.0115 -0.0211 (0.0193) (0.0199) (0.0170) (0.0185) (0.0171) (0.0171) (0.0195) (0.0180) Lack of religiosity 0.00124 -0.000328 0.00424 0.00617 -0.0297 -0.0285 -0.0252 -0.0276 (0.0133) (0.0129) (0.0124) (0.0124) (0.0188) (0.0172) (0.0178) (0.0183) Share of foreign women 0.00713 0.00331 -0.00321 -0.000824 -0.0392 -0.0333 -0.0239 -0.0290 (0.0293) (0.0259) (0.0230) (0.0247) (0.0233) (0.0218) (0.0225) (0.0224) Share of married women 0.00329 -0.0293 -0.0127 -0.0219 -0.0114 0.0176 0.00116 0.0123 (0.0256) (0.0337) (0.0319) (0.0309) (0.0486) (0.0362) (0.0531) (0.0454) Fertility rate 0.0228 0.0338** 0.0206 0.0220 -0.00267 -0.0154 -0.00662 -0.00704 (0.0141) (0.0137) (0.0150) (0.0151) (0.0210) (0.0196) (0.0215) (0.0216)

Observations 261 261 254 254 261 261 254 254 Adjusted R2 0.76 0.77 0.79 0.79 0.74 0.74 0.73 0.74 Region and Year FE Yes Yes Yes Yes Yes Yes Yes Yes Note: for each regional observation, the dependent variable is the number of abortions by women residing in that region but obtained in different regions only over thousands of female population in childbearing age in the region of residence (Columns 1 to 4), and the number of abortions by women residing in different italian regions only but obtained in that region over thousands of female population in childbearing age in the region of abortion (Columns 5 to 8). Robust standard errors in parentheses, clustered by region. Significance levels: *** p<0.01, ** p<0.05, * p<0.1.

Conscientious objection is not only associated with inter-regional travel, but it is also related to the timeliness of access to the abortion service. Building on Bo et al. (2015), we broaden the analysis of conscientious objection vis-`a-vis waiting time between the issuance of an abortion certificate (a mandatory prerequisite granting the entitlement to obtain an abortion) and the actual operation. Table 3 summarizes these results. The first four columns show specifications where the dependent variable is the percentage of abortions performed less than two weeks after the issuance of the certificate. There are fewer timely operations taking place in regions where conscientious objection is more widespread, in particular among anesthetists and non-medical personnel. Symmetrically, as revealed in the last four columns, a greater percentage of abortions oc- curring more than four weeks after the certificate is issued is found in regions with more objectors15.

Results from the various region-level models indicate that on aggregate more out-of-region abortions and longer waiting times are associated to widespread conscientious objection, suggesting that conscientious objection induces women to travel across regions to obtain an abortion. The impact of most other explanatory variables appears to have scarce statistical significance at this

15The positive association between conscientious objection and waiting times holds when choosing different high waiting time thresholds.

15 Table 3: Waiting Time for Abortion and Conscientious Objection

Percentage of Abortions by Number of Weeks between Issuance of Certificate and Operation Less than 2 More than 4 (1) (2) (3) (4) (5) (6) (7) (8)

% Objecting gynaecologists -0.127 0.0613** (0.0856) (0.0242) Workload by non-obj. gynaecologists 0.0341** -0.00519 (0.0142) (0.00407) % Objecting anesthetists -0.353*** 0.109*** (0.0862) (0.0311) Workload by non-obj. anesthetists 0.0748** -0.0164** (0.0329) (0.00758) % Objecting non-medical personnel -0.196** 0.0741*** (0.0811) (0.0254) Workload by non-obj. non-medical personnel 0.0966** -0.0197* (0.0437) (0.0110) Indicator of conscientious objection -0.361*** 0.139*** (0.117) (0.0336) Indicator of workload by non-objectors 0.0820*** -0.0163** (0.0274) (0.00689) GDP per capita 2.093 1.722 2.018* 1.925 -0.847** -0.762** -0.804*** -0.789** (1.335) (1.348) (1.145) (1.232) (0.375) (0.350) (0.278) (0.303) Households in poverty -0.378* -0.362* -0.395* -0.391* 0.0546 0.0450 0.0666 0.0638 (0.215) (0.207) (0.216) (0.204) (0.0809) (0.0768) (0.0811) (0.0739) Female unemployment rate -0.329 -0.485 -0.189 -0.433 0.243* 0.264** 0.165* 0.264** (0.405) (0.410) (0.331) (0.370) (0.126) (0.117) (0.0949) (0.104) Lack of religiosity -0.0590 -0.0294 -0.0581 -0.110 0.0677 0.0656 0.0537 0.0633 (0.277) (0.258) (0.309) (0.289) (0.101) (0.0949) (0.101) (0.0949) Share of foreign women 1.256 1.319 1.288 1.297 -0.105 -0.155 -0.0875 -0.0666 (0.858) (0.884) (0.868) (0.841) (0.266) (0.272) (0.219) (0.205) Share of married women -0.969 -0.342 -0.954 -0.592 -0.161 -0.316 -0.0808 -0.195 (1.173) (1.223) (1.194) (1.218) (0.331) (0.313) (0.249) (0.263) Fertility rate 0.0351 -0.151 0.0933 0.0673 0.0577 0.126 0.0772 0.0851 (0.726) (0.643) (0.693) (0.678) (0.230) (0.211) (0.201) (0.196)

Observations 237 237 232 232 237 237 232 232 Adjusted R2 0.71 0.73 0.74 0.74 0.55 0.57 0.59 0.60 Region and Year FE Yes Yes Yes Yes Yes Yes Yes Yes Note: for each regional observation, the dependent variable is the percentage of abortions for which a period of less than 2 weeks (Columns 1 to 4) or more than 4 weeks (Columns 5 to 8) occurred between the date of issuance of the certificate and the date of operation. Robust standard errors in parentheses, clustered by region. Significance levels: *** p<0.01, ** p<0.05, * p<0.1. stage of the analysis.

The results obtained through panel data analysis of regional figures find a confirmation when the analysis is brought at the individual level. Table 4 shows a probit model where the binary variable set as dependent takes value one in case the woman is operated in a region that is not the one where she resides.

A higher share of objecting gynecologists is significantly associated with a higher probability that women move to another region to obtain an abortion. This relationship is not influenced by the individual sociodemographic characteristics of aborting women, which are controlled for in the specification. We find that the probability of abortion-related migration increases with age, but less so at later stages of the childbearing age range. Aborting women of foreign citizenship migrate significantly less than their Italian counterparts: this is probably suggestive of the fact that Italian women are exposed to higher psychological costs related to the social context they face, would their abortion be made public if obtained in the place where they reside (e.g. more exposure to social stigma in their neighborhood, negative consequences on interactions with people in their network, etc.). Compared to single women, no-longer married women are less likely to travel to seek an abortion; married women are the most likely. Women with higher levels of education move significantly more to other regions than those with only primary or no education, whereas women with lower secondary education move the least. As for the labor force status, employed women are the least likely to obtain an abortion out of their region of residence. The most likely

16 are students (probably reflecting the fact that they are formally residing in a different region than where they happen to live during their years as students) and those out of the labor force, while unemployed women, housewives and women searching for their first occupation are slightly more likely to migrate than employed abortion seekers.

Table 4: Out-of-region Abortions and Conscientious Objection

Woman aborting out of her region of residence

% Objecting gynaecologists 0.00077** (0.00037) Workload by non-obj. gynaecologists -0.000276*** (0.0000521) Age 0.0618*** (0.00266) Age squared -0.00102*** (0.000043) Foreign -0.190*** (0.0110) Single Ref.

Married -0.191*** (0.00536) Divorced/separated -0.0907*** (0.00913) Widowed -0.105*** (0.0318) Not educated/primary education Ref.

Lower secondary education -0.0809*** (0.0100) Upper secondary education 0.0610*** (0.0100) Tertiary education 0.282*** (0.0114) Employed Ref.

Unemployed 0.0550*** (0.00641) Seeking first-time job 0.0512*** (0.0178) Housewife 0.0258*** (0.00607) Student 0.217*** (0.00756) Other 0.419*** (0.0244) Log GDP per capita -0.424*** (0.135) Households in poverty -0.00438*** (0.00126) Female unemployment rate 0.00916*** (0.00163) Lack of religiosity 0.00374** (0.00152) Share of foreign women -0.675** (0.276)

Observations 1,233,692 Pseudo R2 0.181 Birth Region FE Yes Residence Region FE Yes Abortion Region FE Yes Year FE Yes Note: the dependent variable is binary taking value one if the region of abortion of the woman differs from her region of residence, and zero if the region of abortion and residence coincide. Robust standard errors in parentheses, clustered by region of residence. Significance levels: *** p<0.01, ** p<0.05, * p<0.1

Aggregate-level controls for the regional context are also included. Similarly to what was ob- served in the regional analysis, women are less likely to travel outside regions with a higher GDP per capita. There is a lower probability that a woman leaves her place of residence if she resides in region with a larger prevalence of poverty (a result somehow in opposition to the effect of mean income). In addition, regions in which religious observance is lower are those from which women tend to move less. The model incorporates fixed effects capturing time-invariant confounders in

17 the women’s region of birth (e.g. cultural factors, or the tendency to move out from the region of origin) and region of residence (e.g. local attitudes towards abortion). Fixed effects for the region where the abortion takes place are included to take into account time-invariant characteristics that might be related to the demand for abortions in the region (e.g. regional health systems with higher reputation attracting more women in search of an abortion, and students’ inflows in regions with more universities). Year fixed effects account for time-varying factors affecting all regions, such as the economic cycle.

6 Conclusion

The aim of this study was to assess whether conscientious objection to abortion by Italian gynecol- ogists and other healthcare personnel is linked to the phenomenon of abortion-related migration, contributing to explain why many obtain an abortion outside the region where they reside. To provide an answer we first analyzed aggregate data, constructing new regional indicators of abortion-related migration and conscientious objection. Then, individual-level data allowed us to shift the focus on the individual choice of where to seek an abortion. The results consistently indicate that inter-regional travel by women seeking an abortion is related to the prevalence of conscientious objection. A positive association between these two variables exists at the aggregate level, and holds when considering different measures of objection and abortion-related migration. Furthermore, the relation is confirmed at the individual level, suggesting that conscientious objection affects the likelihood that a woman travels out of her region of residence to obtain an abortion. A variety of controls and fixed effects were included in the models, to reduce the scope for confounding effects potentially stemming from individual characteristics of women, economic and demographic characteristics of the regions, and migration that is unrelated to abortion. Drawing on Bo et al. (2015), we also found an association between the prevalence of conscientious objection and waiting times for abortions.

Our findings support the hypothesis that conscientious objection induces women to travel across regions to obtain an abortion. Further research will be necessary to establish to what extent conscientious objection actually constrains access to abortion, and whether limited availability of abortion providers affects the decision of having an abortion or just the decision of where to have one. However, abortion-related migration indicates that conscientious objection affects to some extent the provision of abortion in Italy, possibly creating disparities across different parts of the country and imposing additional time and travel costs on some women.

18 Blank, R. M., George, C. C. and London, R. A. (1996). State abortion rates: the impact of policies, providers, politics, demographics, and economic environment. Journal of Health Economics, 15(5): 513-553.

Bo, M., Zotti, C.M. e Charrier, L. (2015) Conscientious objection and waiting time for voluntary abortion in Italy. The European Journal of Contraception Reproductive Health Care, 20(4): 272-282.

Brown, R. W. and Jewell, R. T. (1996). The impact of provider availability on abortion demand. Contemporary Economic Policy, 14(2): 95–106.

Brown, R., Jewell, R. T., and Rous, J. (2001). Provider availability, race, and abortion demand. Southern Economic Journal, 67(3): 656–671.

Cook, P. J., Parnell, A. M., Moore, J. and Pagnini, D. (1999). The effects of short-term variation in abortion funding on pregnancy outcomes. Journal of Health Economics, 18(2): 241-257.

Deyak, T. A. and Smith, V. K. (1976). The economic value of statute reform: the case of liberalized abortion. Journal of Political Economy, 84(1): 83-100.

Gius, M. (2007). The impact of provider availability and legal restrictions on the demand for abortions by young women. The Social Science Journal, 44(3): 495-506.

Gober, P. (1994). Why abortion rates vary: A geographical examination of the supply of and demand for abortion services in the United States in 1988. Annals of the Association of American Geographers, 84(2): 230–250.

Haas-Wilson, D. (1993). The Economic Impact of State Restrictions on Abortion: Parental Consent and Notification Laws and Medicaid Funding Restrictions. Journal of Policy Analysis and Management, 12(3): 498-511.

Haas-Wilson, D. (1997). Women’s reproductive choices: the impact of Medicaid funding restrictions. Family Planning Perspectives, 29(5): 228-233.

Levine, P. B., Trainor, A. B. and Zimmerman, D. J. (1996). The effect of Medicaid abortion funding restrictions on abortions, pregnancies and births. Journal of Health Economics, 15(5): 555-578.

Matthews, S., Ribar, D. and Wilhelm, M. (1997). The effects of economic conditions and access to reproductive health services on state abortion rates and birthrates. Family Planning Perspectives, 29(2): 52–60.

Medoff, M. H. (1988). An economic analysis of the demand for abortions. Economic Inquiry, 26(2): 353-359.

Medoff, M. H. (2002). The determinants and impact of state abortion restrictions. The

19 American Journal of Economics and Sociology, 61(2): 481-493.

Medoff, M. H. (2010). State abortion policies, Targeted Regulation of Abortion Provider laws, and abortion demand. Review of Policy Research, 27(5): 577-594.

Medoff, M.H. (2013). Social Policy and Abortion: A Review of the Research. The Open Demography Journal, 6: 18-27.

Meier, K., Haider-Markel, D., Stanislawski, A. and Mcfarlane, D. (1996). The impact of state-level restrictions on abortion. Demography, 33(3): 307–312.

Ministero della Salute (2017). Relazione del Ministro della Salute sulla attuazione della Legge contenente norme per la tutela sociale della maternit`ae per l’interruzione volontaria di gravidanza (legge 194/78): dati definitivi 2016.

20