Clarifying the issue

Let’s begin by clarifying the first question:

There is a long-standing theological consensus about what the process of sanctification is supposed to accomplish or bring about, either in part or entirely, namely some form of improvement in the Christian’s moral life. What form does that improvement take? Here too there is a leading answer, namely improvement in the cultivation of virtue. This comes across clearly when we think of passages such as:

What is being cultivated here are virtuous character traits.

Many contemporary scholars also subscribe to this view of sanctification and virtue. Thus, we find the following:

And the consensus does not end there. Many have held that the Holy Spirit is involved during sanctification in some way, either partially or wholly, in bringing about character change in believers. After all, as the Galatians passage indicates, the virtues being cultivated are the Fruit of the Spirit.Footnote ³

So with this framework in mind, we can return to our first question and interpret it as follows:

This formulation of the question also brings out one more point worth highlighting, namely that sanctification is taken to be restricted to all and only Christians, however precisely that term is to be understood. Atheists, agnostics, and followers of other non-Christian religious traditions would presumably not have the Holy Spirit operative, in whole or in part, in sanctifying them in this life. We will return to this issue again in the ‘Lingering issues’ section before the conclusion.

Let’s end this section with a few words about the nature of the virtuous character traits whose cultivation is the end goal of sanctification. Virtues are excellences of character, and examples include those enumerated in Galatians 5 as well as others like honesty and fairness. They are multifaceted traits of character, with cognitive, motivational, and behavioural components. For instance, the trait of compassion leads to behaviour that helps those in need. But the compassionate person needs to be able to recognize that need when it exists, know what means to take to help address it, balance helping this one instance with being able to help in other instances as well, and so forth. Furthermore, it matters why she is helping those in need. It wouldn’t primarily be for motives like improving one’s social media following, relieving feelings of guilt, or advancing in one’s job. Rather than self-interested motivation, the compassionate person is primarily moved by genuine love, care, and concern for others.

Two other features of moral virtues are worth highlighting. One is that a virtue leads to behaviour across a variety of situations. The compassionate person doesn’t just help pick up dropped papers, to take a simple example. And the other is that the impact of a virtue is felt over time. In other words, there is stability to the virtuous person’s behaviour from week to week and month to month.Footnote ⁴

With this background in place, we can try to make some progress in actually answering the first question.

Generating predictions

So if Christianity is true, what can we reasonably predict about the impact of sanctification, including the work of the Holy Spirit, on the cultivation of virtuous character traits in Christians? Alas, there isn’t a straightforward answer to this question, as we can formulate a variety of predictions. Here are some initial options:

Let me say something briefly about the first two answers before focusing the remainder of this paper on the third.

According to the Conversion Answer, for those who do genuinely convert to Christianity, we should expect to see their character improve, at least on average, in the months after their conversion as compared to the months prior to their conversion. Not all conversions need to bring about dramatic changes in character, but according to this way of thinking, it would be surprising if there were not some significant character improvement after, say, 12 months.Footnote ⁵

The Temporal Answer has a broader focus. It pertains to all Christians, whether they were converts or not, and holds that, at least over long periods of time, Christians should on average see significant improvement in their characters. Here, the time frame might be measured in decades rather than days or weeks. But it would be surprising if, by the end of their lives, Christians were not on average more virtuous than they were in their earlier years.

The Conversion and Temporal Answers are depicted in this illustration by Wayne Grudem (Figure 1). Here we see at work the assumptions that virtue (holiness) will increase both at the time of conversion and during the course of the Christian life.

Figure 1.

Illustration of the Conversion and Temporal Answers.Footnote ⁶

Having illustrated these two answers, I will put them to one side in the remainder of the paper for two reasons. First, I take them to be less controversial than the Comparative Answer, which, as we will see later in this section, gives rise to a number of concerns. Second, there isn’t a body of empirical evidence at present, which can be used in assessing whether, in fact, the changes to character predicted by the Conversion and Temporal Answers come about in the life of the Christian. But in the case of the Comparative Answer, there is a body of empirical evidence which is at least somewhat illuminating.

So let’s turn to the Comparative Answer. Right off the bat, we can note that there are certain ways of comparing the moral character of Christians versus non-Christians which do not have a chance of being plausible, either as commitments of Christianity or as being consistent with the empirical data. Here are two examples:

C. S. Lewis noted one reason why these are implausible claims, as it might be that a particular Christian (or small group of Christians) is just starting from a worse place morally speaking (i.e., has a more vicious character) to start with (Reference Lewis2009, Book IV, Chapter 10). Hence, a new convert at age 20 might have been dishonest and cruel during his teenage years, as compared to a non-Christian at age 20 who received an excellent moral education. Even post-conversion, this Christian might be less virtuous than the non-Christian.

To stand a chance, a comparative prediction will have to range over Christians in general (or at least over huge numbers of them). Hence, we get this formulation:

Now even here, the same concern could apply. In other words, it might be that Christianity has special appeal to those who are especially in need of salvation and struggle especially deeply with sin.Footnote ⁷ If so, then even post-conversion they might have a worse character than non-Christians, due to how bad their character was to begin with. And this comparative difference could persist after conversion as well.

This is certainly a possibility that cannot be ruled out ahead of time. To be sure, it is not the case that all, or even most, Christians are converts. Many were raised from childhood in the faith. And there is the complicating factor of prevenient grace, which allegedly prepares a person for salvation. It is meant to precede justifying or salvific grace. If there is such prevenient grace, it would be natural to expect it to beneficially impact at least part of a non-Christian’s character prior to conversion, thereby calling into question how bad the Christian’s character really would be pre- and post-conversion.

Also, it is an empirical question whether pre-conversion Christians really did have a worse character than non-Christians. So in principle, we could do studies which asked converts to Christianity to self-report what their character used to be like (perhaps with input from observers who knew them previously, such as family and friends). That could tell us something about whether we have a mere possibility or an actuality.

And finally, even if the scenario outlined above does reflect the facts on the ground as far as post-conversion Christians on average having a worse character because of where they started from in the first place, that character is still supposed to improve via sanctification over time. So, there might still be an important Group Comparative Prediction to consider, which is restricted just to the end of the lifespan for Christians versus non-Christians, once the Holy Spirit has had plenty of time to operate.Footnote ⁸

Nevertheless, despite all these points, with the information available to us right now, we can’t rule out the possibility in question. So we get a revised prediction as follows:

Having noted this conditional version, I will stick with the original Group Comparative Prediction for the remainder of the paper, as the exception noted above will not be necessary to consider when looking at the empirical evidence and seeing whether it is compatible with the comparative claim.

What then should we make of the Group Comparative Prediction? When I ask this question, I do not mean to be asking whether there is any evidence that Christians on average have a more virtuous character. That is the question we will take up in two later sections. Instead, what I mean here is precisely this – is greater virtue among Christians indeed what you would predict if Christianity were true? For instance, if someone were to study the New Testament carefully for the first time, would they likely come away with the expectation that – if Christianity were true – Christians would be better people, morally speaking, than non-Christians?

In the next section of this paper, I will explain why I answer yes. But I readily acknowledge that this is a contentious matter. I can imagine at least two different reasons for why someone might challenge the Group Comparative Prediction:

This second doubt can be developed into what has been called the Meager Moral Fruits Argument. The argument can be formulated in various ways, and we do not need to get into the details.Footnote ¹¹ Here is one version that is particularly relevant to our discussion:

1. (1) According to the Group Comparative Prediction, if Christianity is true and the process of sanctification is operative in the life of Christians either partially or entirely through the work of the Holy Spirit, then we have good reason to predict that Christians on average will have a more virtuous character than non-Christians.

2. (2) There is sufficient empirical evidence to conclude that Christians in general do not have a more virtuous character than non-Christians in general.

3. (3) Therefore, there is good reason to think that Christianity is not true.Footnote ¹²

Other things being equal, Christians cannot accept this conclusion.Footnote ¹³ If premise (2) is defensible, then they would need to reject the Group Comparative Prediction.

These concerns should give Christians pause before accepting the Group Comparative Prediction.

Defending the Group Comparative Prediction

And yet, Christians should not give up this prediction quite yet. Why not? Because of the following:

Now granted, if the Holy Spirit is not involved and sanctification is meant to be entirely of the Christian’s own doing, then all bets are off. Given how human beings tend to mess things up, it is not at all clear what should be expected about the character of Christians.

But as we already noted, it has been widely accepted throughout the Christian tradition that the Holy Spirit does play an integral role in sanctifying the Christian, and so thereby is involved in improving that person’s character. Furthermore, if the Holy Spirit is involved, it is hard to accept that the Spirit would have no impact, or even just a marginal one, on the moral character of Christians. Presumably this impact would be magnified over significant periods of time. Hence, I do not see a way to avoid the conclusion that, given Christian claims about sanctification and the Holy Spirit, they predict that those in whom the Spirit is operative will on average have a better moral character than those in whom the Spirit is not.Footnote ¹⁴

These points also go some way towards addressing the first set of doubts about process. Admittedly, there is little clarity about the rate at which sanctification is supposed to occur, and it is left open whether the Holy Spirit works in the same way in each believer. But again, if we are talking about the third person of the Godhead, and if the claim is that God qua Holy Spirit is causally operative in bringing about character change in the life of every Christian believer, the natural expectation is that – on average – God would move the needle of positive character change to a significant degree.Footnote ¹⁵

These preliminary observations can be developed more carefully by considering the leading models for better understanding how the Holy Spirit could be at work in the lives of Christians. Here I follow the careful development of these models in William Alston’s classic paper, ‘The Indwelling of the Holy Spirit’. Alston (Reference Alston and Morris1988) focuses on three models in particular, and fortunately for our purposes we do not have to worry about which is the most plausible since it will turn out that all of them are going to take us to the same conclusion.Footnote ¹⁶

Let’s begin with the Fiat Model. It holds that the process of sanctification involves God directly making changes to the believer’s psychology (Alston Reference Alston and Morris1988, 128). For example, the Holy Spirit might weaken certain lustful or avaricious desires. Or the Spirit might strengthen the Christian’s commitment to help the poor or to donate to the church. Or there might be both a weakening of some parts of the Christian’s psychology and a strengthening of others. As Biblical support, advocates of the model appeal to Philippians 2:13: ‘You must work out your own salvation in fear and trembling; for it is God who works in you, inspiring both the will and the deed, for his own chosen purpose.’

Suppose something like the Fiat Model is correct. What should Christians then predict about sanctification? It seems clear that they should expect better character on average from individuals in whom the Holy Spirit is directly involved in this way, as compared to individuals in whom the Holy Spirit is not involved. That is in line with the Group Comparative Prediction.

Next, turn to the Interpersonal Model. Here the central claim is that the process of sanctification involves the Holy Spirit influencing the believer in a better direction without directly causing changes in the believer’s psychology. Such influencing might take the form of divine callings, communications, and exhortations, presentations of God as a role model, the making of his love more obvious, loving encouragements, and so forth (Alston Reference Alston and Morris1988, 132). As a supporting verse, advocates of the model appeal to John 14:26, whereby the Spirit will ‘teach you everything, and will call to mind all that I have told you’.

What prediction should Christians make based on this model? The same one, i.e., they should expect better character on average from individuals in whom the Holy Spirit is involved, as compared to individuals in whom the Holy Spirit is not involved.Footnote ¹⁷

Finally consider the Sharing Model, which Alston himself supports. On this model, the process of sanctification involves being drawn into the divine life and living it to some limited extent (Alston Reference Alston and Morris1988, 139). We participate or partake in the divine nature at least partially. As support, advocates of the model cite 2 Peter 1:4 and the idea that through the indwelling of the Holy Spirit, we ‘come to share in the very being of God’.

What is a reasonable prediction given this model? The same one again. Based on the Sharing Model, Christians should expect better character on average from these individuals when compared to those in whom the Holy Spirit is not involved.

This section has thus provided what I take to be the strongest case for the Group Comparative Prediction. But before ending, there is another piece of support worth noting.Footnote ¹⁸ Recall the Conversion and Temporal Predictions. They are each well-supported as predictions of Christianity. And, it would seem, they each can serve to increase the plausibility of the Group Comparative Prediction.

Take the Conversion Prediction first. It says that for converts to Christianity, they will on average tend to have a more virtuous character post-conversion. But non-Christians don’t benefit from this particular instance of character improvement. Hence, other things being equal, converts to Christianity should have a better moral character than non-Christians, which provides some support for the Group Comparative Prediction.

Now take the Temporal Prediction. It says that over time, Christians on average will tend to develop a more virtuous character. For non-Christians, this could also be the case, but there is no guarantee that it will be the case. It will be a highly contingent matter. Hence, other things being equal, as more time passes, there is a greater likelihood that Christians will have a better moral character than non-Christians, which also provides some support for the Group Comparative Prediction.

Finally, the Conversion and Temporal predictions can be combined to provide joint support for the Group Comparative Prediction. Admittedly, they don’t entail it, either singly or jointly. But if you accept these first two predictions, it becomes more reasonable to accept the third as well.

To sum up, I think we are left with the conclusion that the Group Comparative Prediction is plausible, despite the doubts that were raised in the previous section. But then does it follow that Christianity is false? That was the conclusion of the Meager Moral Fruits argument. It held that there is sufficient empirical evidence to conclude Christians in general do not have a more virtuous character than non-Christians in general.

It is time to turn to some of the empirical evidence and see what it suggests.

Turning to the empirical literature

What kind of empirical evidence should we look to in order to evaluate whether Christians on average have a more virtuous character than non-Christians? Here individual testimonies and stories of the virtuous lives of particular Christians have limited relevance. We are looking for differences in averages involving massively large groups. Also, there are plenty of individual cases of non-Christians who have lived very admirable lives.

Perhaps more promising are historical records of morally good behaviour by large groups of Christians throughout history. For example, the church historian John Woodbridge claims that:

When we see Christians behaving in a starkly better way than others in their historical surroundings that does seem to offer some supporting evidence for the Group Comparative Prediction.

The complicating factor here is that there are also numerous historical examples of groups of self-described Christians behaving badly. The Crusades, the Inquisition, and the Salem Witch Trials are familiar examples, but there are plenty of others as well. It is hard to know how to collectively add up the historical pluses and minuses when it comes to Christians and their moral character as exhibited in outward behaviour.

What would be more helpful than these approaches is to look to the existing empirical research to see if there is relevant work which compares the moral behaviour of Christians and non-Christians. By drawing mostly on studies from sociology, psychology, and economics, this is indeed what we will do in the remainder of this section.

Before we dive in, two preliminary notes are worth making about these studies. The first is that they do not all employ the same measure of religiosity, since there is no single widely endorsed measure. Instead, we find measures ranging from self-reported strength of belief in God, to attendance at religious services and frequency of prayer.Footnote ¹⁹

The second preliminary note is that the studies tend to be heavily skewed towards American and European populations. So whatever conclusions we might be tempted to draw should be tempered until greater cross-cultural research is conducted.

With these preliminaries out of the way, what trends, if any, do we find in the empirical literature examining religiosity and moral outcomes? There is no way to review all the empirical studies across the various academic disciplines. Instead, I will look selectively at four virtues – compassion, non-malevolence, honesty, and self-control – and see what a significant body of the studies in those areas seems to indicate.Footnote ²⁰

Compassion. The virtue of compassion pertains to aiding those in need, and especially to the alleviation of suffering. Here is some of the relevant data on charity donations and volunteering to help people in need:

• • Those who regularly attended religious services were 25% more likely to give to charity than those who did rarely or said they were not religious (91–66 percent). They also were 23% more likely to volunteer (67–44 percent). In 2000, they gave away 3.5 times more money per year ($2,210 versus $642) and volunteered more than twice as much (12 times versus 5.8 times) (Brooks Reference Brooks2006).

• • ‘In 2002, religious people were far more likely to donate blood than secularists, to give food or money to a homeless person, to return change mistakenly given them by a cashier, and to express empathy for less fortunate people … religious people were 57 percent more likely than secularists to help a homeless person at least once a month’ (Brooks Reference Brooks2006, 39).

• • Similarly, other studies found that, for instance, over 20% of fundamentalists, evangelicals, mainline, and liberal Protestants self-reported giving away ‘a lot’ of money to organizations which help the poor in the past two years, whereas only 9.5% of nonreligious respondents did. In general, the latter group was 0.70 times as likely to give to such organizations as religious people were, regardless of what type of religion was subscribed to (Regnerus et al. Reference Regnerus, Smith and Sikkink1998).Footnote ²¹

• • In their literature review of 77 studies, Yasin and colleagues found that ‘religion – measured as group affiliation, values, beliefs, and the importance of faith to the believer, as well as religious service attendance and private prayer – most often positively relates with giving to outgroup and secular organizations’ (Yasin et al. Reference Yasin, Adams and King2020, 11).Footnote ²²

Non-malevolence. While compassion has to do with helping, non-malevolence is a virtue concerned with not harming others. Donating to charity would fall under the scope of compassion, whereas something like not assaulting a person would fall under the scope of non-malevolence. Here is some relevant data in this area:

• • Ellison and Anderson found that based on self-reports, men who attend religious services once a week or more are 60.7% less likely than nonattenders to commit domestic violence. Partner reports show a 48.7% decrease (Ellison and Anderson Reference Ellison and Anderson2001).Footnote ²³

• • Johnson and Siegel found that ‘the probability of committing a non-drug crime was 39 percent smaller for youth who attended church more than once a week compared to youth who did not attend, and the difference for drug use was found to be 46 percent. Finally, the probability of youth selling drugs was 57 percent smaller among regular church-attendees than non-attendees’ (Johnson and Jang Reference Johnson and Jang2010, 119, citing Johnson and Siegel Reference Johnson and Siegel2002).

• • Using a large sample of German adolescents, Dirk Baier found that 30% of Muslims versus 18% of Christians self-reported engaging in at least one instance of violent behaviour in the past 12 months (Baier Reference Baier2014, 114). And higher religiosity decreased violence for Christians but not in a statistically significant way for Muslims (Baier Reference Baier2014, 117).

• • Religious attendance was correlated with reduced self-reported performance of 43 different crimes (Evans et al. Reference Evans, Cullen, Dunaway and Burton1995). In another study, religiosity was correlated with decreased delinquency using a self-report measure with 35 different delinquent behaviours; indeed the authors found some evidence for reciprocal causation (Johnson et al. Reference Johnson, Jang, Larson and De Li2001, 35–36). Other studies also found negative relationships between various measures of religiosity on the one hand and crime and violence on the other (Lipford et al. Reference Lipford, McCormick and Tollison1993; Hull and Bold Reference Hull and Bold1995; Hull Reference Hull2000; Salas-Wright et al. Reference Salas-Wright, Vaughn and Maynard2014).

• • A 2001 meta-analysis of 60 studies found ‘religious behavior and beliefs exert a significant, moderate deterrent effect on individuals’ criminal behavior’ (Baier and Wright Reference Baier and Wright2001, 14). A systematic review in 2000 of 40 studies found that ‘Religious measures are generally inversely related to deviance, and this is especially true among the most rigorous studies’ (Johnson et al. Reference Johnson, De Li, Larson and McCullough2000, 32). In 2010, a review of 270 studies found ‘an inverse or beneficial relationship between religion and some measure of crime or delinquency’ in 90% of the studies (Johnson and Jang Reference Johnson and Jang2010, 120).

Honesty. This virtue covers many behaviours, but here we will just focus on lying (for a detailed study of honesty, see Miller Reference Miller2021). Unfortunately, there are not many studies, but these are a few results:

• • Students at a university in Canada played an economic game with a partner where they could lie to get a bigger financial pay-off for themselves. Students for whom religion was more important tended to lie a bit more (Childs Reference Childs2013). But in another study, this time at a Catholic university in the U.S., overall religiosity, frequency of prayer, and the importance of religion in daily life all increased the odds that a student would send an honest message in a similar situation (Christie Reference Christie2019).

• • A study found that undergraduate students at a school in the U.S. who tended to agree with items like ‘Lying is no big deal’ and ‘What people don’t know can’t hurt them’ also tended to be less religious. It was concluded that lying acceptability correlates negatively with religiosity (Oliveira and Levine Reference Oliveira and Levine2008).

• • Finally, a study examined counties in the U.S. that were more or less religious. It found that companies in more religious counties tended to be less likely to misrepresent their financial statements, to employ tax sheltering, and to withhold bad news in financial disclosures (Dyreng et al. Reference Dyreng, Mayew and Williams2012).

Self-control. This virtue includes acting with moderation and resisting temptation to do something bad or wrong. Here is a quick summary of various trends:Footnote ²⁴

• • Measures of religiosity have been significantly linked to reduced rates of suicide (Stack Reference Stack1983; Donahue Reference Donahue and Benson1995), lower drug use (Gorsuch Reference Gorsuch1995; National Center on Addiction and Substance Abuse 2001),Footnote ²⁵ increased health care utilization (Schiller and Levin Reference Schiller and Levin1988; Benjamins and Brown Reference Benjamins and Brown2004), reduced smoking (Koenig et al. Reference Koenig, George and Cohen1998; Gillum Reference Gillum2005), reduced alcohol abuse (Clarke et al. Reference Clarke, Beeghley and Cochran1990; Cochran Reference Cochran1993; Koenig et al. Reference Koenig, George, Meador, Blazer and Ford1994; Cochran et al. Reference Cochran, Beeghley and Bock1998), healthier lifestyles (Hill et al. Reference Hill, Ellison, Burdette and Musick2007), the promotion of mental health (Larson et al. Reference Larson, Sherrill, Lyons, Craigie, Thielman, Greenwold and Larson1992; Levin and Chatters Reference Levin, Chatters and Koenig1998),Footnote ²⁶ and even mortality rates (Hummer et al. Reference Hummer, Rogers, Nam and Ellison1999).

• • For instance, with mortality rates, Hummer and colleagues found that the life expectancy estimates at age 20 were 55.3 years for those who did not attend religious services, compared to 62.9 years for those who attended more than once per week. For African Americans, the difference was 46.1–60.1 years, respectively (Hummer et al. Reference Hummer, Rogers, Nam and Ellison1999).

• • As another example, a 2020 study by Chen and colleagues found that among ‘66,492 female registered nurses and 43,141 male health care professionals in the US, attendance at religious services at least once per week was associated with a 68% lower hazard of death from despair among women and a 33% lower hazard among men compared with never attendance.’Footnote ²⁷ An earlier 2016 study similarly found that among ‘89,708 women aged 30 to 55 years who participated in the Nurses’ Health Study, attendance at religious services once per week or more was associated with an approximately 5-fold lower rate of suicide compared with never attending religious services’ (VanderWeele et al. Reference VanderWeele, Li, Tsai and Kawachi2016, 845).

Finally, as a study that cuts across multiple virtues, consider the results of the 2002 Research Report of the National Study of Youth and Religion (2,478 twelfth-grade students):

Note the marked difference in behaviour in every case, with more positive outcomes linked to weekly religious attendance.

Some observations and limitations

What should we make of the results highlighted above, in light of the Group Comparative Prediction? I think that they are consistent with the prediction, but we need to be careful not to claim too much.

As far as the consistency goes, the prediction is that there will be character differences between Christians and non-Christians, and presumably those would manifest as behavioural differences. It turns out that the studies reviewed here did find behavioural differences. Furthermore, the prediction is that those differences should tend to reflect an alignment between higher religiosity and either acting better (donating more, volunteering more, etc.) or refraining from acting worse (less crime, lower drug use, etc.). Again, that is exactly what the results tended to show. Hence, there is strong alignment between what the Group Comparative Prediction would lead Christians to expect, and what emerged in the data.

Furthermore, we can now return to the empirical concerns from the ‘Generating predictions’ section, and in particular the Meager Moral Fruits argument. There we said that, since Christians cannot accept the conclusion of the argument, they will have to reject the Group Comparative Prediction if it is true that,

1. (2) There is sufficient empirical evidence to conclude that Christians in general do not have a more virtuous character than non-Christians in general.

But now we have seen that this premise is problematic. The empirical evidence, at least at present, does not warrant this conclusion. In fact, if anything, it points in the opposite direction and suggests there is at least some empirical evidence for thinking that Christians in general have a more virtuous character than non-Christians. Hence, there are resources available to the Christian for responding to the Meager Moral Fruits argument without having to give up the Group Comparative Prediction.

So far so good for the Group Comparative Prediction. Now come all the qualifications. There are five of them.

First Qualification. The empirical studies do not always point in the same direction. For example, in a study of Canadian subjects, Brinkerhoff et al. (Reference Brinkerhoff, Grandin and Lupri1992) did not find a significant difference in domestic violence rates. Similarly, using a church-based sample of women in Los Angeles, Fox et al. (Reference Fox, Pitkin, Paul, Carson and Duan1998) found little difference when it came to breast cancer screening.Footnote ²⁸ And in two studies with small numbers of participants, Shariff and Norenzayan (Reference Shariff and Norenzayan2011) did not find a difference between religious believers and nonbelievers using a lab measure of cheating. Overall, though, these studies tend to go against general trends in their respective areas of empirical research.

Second Qualification. The studies that were reviewed here were connected to four virtues: compassion, non-malevolence, honesty, and self-control. Indeed, these virtues were chosen precisely because there seems to be a sizable body of relevant empirical research that relates to each of them. But then, that leaves open the possibility that, with respect to other virtues, non-Christians might end up measuring higher on them than Christians. Possibilities here could include open-mindedness, intellectual humility, tolerance, scrupulosity, and bigotry. Limitations of space here prevent trying to cover all or even a majority of the virtues, but in a book-length discussion that would be a worthwhile project to undertake.Footnote ²⁹

Third Qualification. Another factor to keep in mind is whether the studies are conducted in such a way as to pick out all and only genuine Christians. This point has two parts – the ‘genuine’ and the ‘Christians’.

On the ‘genuine’, there are of course theological disputes about what it takes to be a genuine believer in the first place. And whatever the requirements are, it is unlikely that the studies are tracking them carefully. For instance, someone might be a Christian but not attend church regularly. Or the opposite. Similarly, someone might self-report a wavering belief in God, and yet be a Christian. Or again the opposite.

On the ‘Christian’, many of the studies in sociology, psychology, and economics do not filter their participants in such a way that we get a clean comparison between Christians and non-Christians (or even between Christians and the non-religious). When the measures of religiosity are things like self-reported strength of belief or frequency of prayer, they will often be used for practitioners of other faiths like Judaism and Islam as well.

Now, as a matter of fact, the studies happen to be done in locations where it is likely that those who are high on a religiosity measure also tend to be Christians. But for the Group Comparative Prediction, what we need is a clean divide between Christians and non-Christians, not between the religious and the non-religious.

Fourth Qualification. We need to remember that these are only correlational studies. They do not tell us about the direction of causation. For instance, it could be that those who are already higher on compassion, non-malevolence, honesty, and self-control tend to gravitate towards Christianity.Footnote ³⁰

Furthermore, even if we overlook this point and take there to be some causal relationship resulting in improved character, the studies reviewed above underdetermine what it would be that is causally responsible for that improvement in character. Here are some explanatory possibilities:Footnote ³¹

1. (i) Improved character is due entirely to the internal workings of the Holy Spirit.

2. (ii) Improved character is due entirely to participation in Christian practices like tithing, prayer, fasting, ministering to the poor, and so forth.

3. (iii) Improved character is due entirely to the religiosity of the participants.

4. (iv) Improved character is due entirely to participants practising the dominant religion in their political community.

5. (v) Improved character is due to some combination of factors above.

Hence, the lesson needs to be that the data are consistent with the Group-Level Prediction, but do not uniquely support it. Other explanatory stories are left on the table as well.Footnote ³²

Fifth Qualification. Finally, it is important to remember that better behaviour on average does not amount to greater virtue. As we said in the ‘Clarifying the issue’ section, a virtue is a multifaceted character trait, of which external behaviour is only one component. Virtuous thoughts and virtuous motivation are both essential as well. Hence, it is possible that even if Christians are found empirically to behave better than non-Christians, they could be doing so for less virtuous motives.

To take an example, suppose it is established fairly conclusively that Christian belief tracks higher donations to charity. What is motivating Christians on average to make such donations? All the usual egoistic possibilities would be on the table, like getting a tax donation or making a good impression on others. In addition, there are egoistic possibilities within a Christian framework as well, such as getting rewards in the afterlife or avoiding hell. If motivation on average tends to be less virtuous for Christians than it is for non-Christians, then even if Christians tend to behave better, it is not clear how things will turn out in the final analysis.

Now I am not at all suggesting that such a scenario is actually the case, only that it is possible. Since we do not have good empirical evidence about what primarily motivates Christians when they perform morally admirable actions, a key piece of the puzzle is still missing.

Lingering issues

Empirically, then, Christians should be careful not to claim too much from the empirical literature. It appears as if the data is, for the most part, in line with what they should expect if the Group Comparative Prediction were true. But at the same time, it is not sufficient by itself to demonstrate that there is greater moral virtue on average in Christians as compared to non-Christians (nor does it show the opposite).

In this final section, I want to raise some challenges to my approach and briefly address them, although they each deserve more space than is available here. The challenges have to do with common grace, infused virtue, anonymous Christians, and group-level sanctification. Let’s take each in turn.

Common grace. One challenge appeals to common grace, which is supposed to enable those suffering from sin (whether original sin or personal sin or both) to be able to perform any good action. As Louis Berkhof writes,

Note that since common grace is not limited to Christians, it might level the playing field between Christians and non-Christians as far as comparative virtue is concerned. Hence, it could give Christians reason to reject the Group-Level Prediction.

Now the existence of common grace is controversial in Christian theology, and many Christians deny that it exists. But even for those who accept common grace, it does not directly impact the discussion in this paper. For even if everyone receives common grace, there is in addition the specific work of the Holy Spirit during the process of sanctification (sanctifying grace). And that is not supposed to extend to all humans in this life, but only to genuine Christians. Thus, the combination of common grace and sanctifying grace still leads to the Group-Level Prediction.Footnote ³³

Infused Virtue. Another potential challenge could arise in connection with infused virtues. Some argue, especially in the Thomistic tradition, that at least certain virtues are only bestowed upon Christians by God, and cannot be acquired by means of human agency alone.Footnote ³⁴

By itself, the claim that there are infused virtues does not change any of the claims made in this paper. We would need to know which virtues these are supposed to be, and then check whether there is empirical data that bears on the comparative possession of those virtues in Christians and non-Christians.

Where there could be trouble is if the entirety of the sanctification process is supposed to involve bestowing infused virtues that would make no empirically discernible difference to human behaviour. Then, as far as the work of the Holy Spirit is concerned in sanctification, Christians would have to be agnostic about whether, on empirical grounds, one group is higher in these virtues than the other. But the idea that the infused virtues have no manifestations at all in behaviours which lend themselves to empirical observation, seems a very implausible view for any Christian to hold.

Anonymous Christians. Some theologians maintain that there are so-called ‘anonymous Christians’, who do not consciously consider themselves to be Christians (and indeed, may never have heard of Christianity explicitly before). Nevertheless, they are genuine Christians all the same, and as it relates to this paper, their character is undergoing the process of sanctification alongside that of more traditional followers of Christ. If anonymous Christians exist, they may not be caught by the measures of religiosity used in the empirical studies. Hence, the comparisons could be confounded by there being anonymous Christians in what is supposed to be the group of non-Christians.Footnote ³⁵

Now the existence of such anonymous Christians is controversial, but leave that aside. Suppose we grant that they exist, and that the Holy Spirit is working to improve their characters via sanctifying grace. Yet we saw that there is still a disparity in the data pertaining to the four virtues examined in the ‘Turning to the empirical literature’ section. So while there may be a few such Christians, that need not by itself change the main claims of this paper.

Admittedly, it might be a different story if the claim is that most people who are not explicitly Christians are actually anonymous Christians. Then while the Group Comparative Prediction might be true in theory, it would matter little in practice since everyone, or nearly everyone, is a genuine Christian in this life and so is being actively sanctified by the Holy Spirit. Such a scenario is possible, I admit, but I doubt that it has much support in the Christian tradition.Footnote ³⁶

Group-level Sanctification. Yet another issue worth considering is whether sanctification is supposed to operate primarily at the level of individual believers, or whether it is supposed to operate primarily at the level of groups such as the Church or the Body of Christ. In other words, might it turn out that what is primarily being sanctified is the Church corporately, rather than discreet individual members of the Church?Footnote ³⁷

We do not need to get into debates about the plausibility of this more group-oriented approach to sanctification here. For such a view does not deny that individual Christians are being sanctified as well. Presumably, this would be a ‘both/and’ phenomenon, rather than ‘either/or’. Furthermore, group-level sanctification does not change the terms of the Group Comparative Prediction, which after all is stated in terms of comparing two groups of people.Footnote ³⁸