Brexit uncertainty: fewer mergers, big business winners

While it is widely recognised that last year’s EU referendum caused significant uncertainty for markets, some early indications were that it had not reduced the level of business confidence. Our results show that the uncertainty surrounding the referendum triggered a fall in the number of mergers and acquisitions, and M&A activity has not recovered. The mergers that have suffered the largest drop have been the ones that were large in comparison to the size of the acquiring firms. Finally, as a result of the post-referendum uncertainty, the businesses that have remained more M&A active have been typically the largest ones. Our results – that are based on a careful study design of a treatment and control group – contradict some of the earlier, more optimistic discussions, which were based on simple before-after comparisons.


There have been various reports on how business confidence has evolved in the aftermath of the Brexit referendum. Some of these emphasised positive developments others painted a less rosy picture. We look at a specific angle of this story, mergers and acquisitions (M&A). Our premise is that if the referendum increased policy uncertainty, this would have negatively affected UK businesses’ M&A activity. The negative relationship between policy uncertainty and M&A activity has been shown to exist in various other circumstances, our post-referendum study offers a novel perspective to this story.

Did the referendum have a positive or a negative, or no effect on M&D activity in the UK? Even such a seemingly unambiguous question has managed to attract a range of contradicting answers. Some interpreters find positives, some highlight negative developments. Why the discrepancy? Probably because these analyses are limited to simply looking at how the numbers changed in comparison to M&A activity in the UK pre-referendum. For example, the average monthly number of M&A announcements in the 11 months before the referendum was around 430. The same average in the 11 months since the referendum has been around 350. Can we conclude that the referendum caused the monthly number of M&A announcements to drop by 80? Some might say yes, but others could argue that the drop was caused by something other than the referendum. Before-after comparisons are unable to identify causal effects and, as such, are ill suited for identifying the impact of the referendum on M&A activity.

In this post we present a preliminary attempt to identify a causal relationship between the referendum and M&A activity. We do this by offering a set of causal inference methods – methods typically used in research to establish causal relationships, such as the impact of a change in policy, or – in our case – the impact of a specific event (the referendum). The data we used (the number and value of transactions) was downloaded from S&P’s Capital IQ transaction database. Note that this is only a preliminary analysis of the data and reflects ongoing work. However, we felt confident about publishing these preliminary results as they have proven to be robust to a number of different methods and model specifications. Moreover, we feel that it is important that these results are out before the June 2017 General Elections. For the sake of accessibility to the general public, we refrain from technical explanations, and instead offer a digestible explanation of what we are doing.

Study design

Finding the right study design is pivotal for identification of causality between the referendum and M&A activity. The first, and probably most important, step is to find a Control group, against which the changes in the Treatment group (UK businesses) can be compared. The idea is that the Control group should be just like the Treatment group except it did not receive the treatment (i.e. it did not have a referendum). For our problem this translates to finding a counterfactual country, or set of countries, and compare M&A activity in the UK before and after the referendum to M&A activity in this selected set of counterfactual countries. If the selected other countries are sufficiently similar to the UK, then we can assume that these other countries represent how M&A activity would have evolved in the absence of the referendum in the UK. Therefore the difference between the UK and these other countries could be thought of as the effect of the referendum.

A simple way of selecting a Control ‘group’ would be to pick a country that is roughly similar to the UK and compare M&A activity across the two countries. For example comparing with Germany might seem like an obvious choice. For Germany to be a good comparator one would need it to be as similar to the UK as possible, and one obvious measure of this similarity is the trend in M&A activity pre-referendum. The figure below shows how the monthly number of M&A announcements changes for the two countries (the vertical dotted lines represent the start of the referendum campaign and the referendum). It appears that before the referendum there had been a slight drop in the UK number, on the other hand there seems to be a slight rise in the German one. So using the difference between these countries as an estimator for the effect of the referendum is likely to pick up some other effect as well – something that caused the difference in trends even pre-referendum.

OK, what can we do then? One could of course look at every other country one by one, until a sufficiently similar one is found. But similar in what respect? Here we are only comparing countries based on one characteristics (M&A activity). What if there are other characteristics that differ across countries and that have a strong impact on M&A activity? For example how healthy a country’s economy is, or how easy it is to do business, and so on. Such multi-dimensional comparisons would be an arduous task to do manually, and even then it would be questionable whether a single best comparator country can be selected. One thing that researchers can do in these cases is rely on something called a synthetic control group, which is a weighted average of a number of other countries that is the most similar to the Treatment group based on a comparison of many characteristics. It is as if we were saying that given a list of attributes the UK is a little bit like Germany, but also a bit like the US, and a bit like France too. When we create a synthetic Control group we allow the UK to be compared to many countries at the same time, each with a different weight.

For this preliminary study we selected 8 other countries (Australia, France, Germany, Italy, Netherlands, Spain, Sweden, and the US) to serve as the starting step for creating a synthetic Control group. Data on country characteristics were downloaded from the World Bank’s Databank. We make the comparison on the following attributes: GDP (level and growth); domestic credit to private sector (% of GDP); ease of doing business index (a World Bank index); foreign direct investment; net inflows (% of GDP); inflation (annual %); interest rate spread (lending rate minus deposit rate, %); lending interest rate (%), profit tax (% of commercial profits), research and development expenditure (% of GDP), tax revenue (% of GDP), unemployment (% of total labour force); average monthly value of transactions (million USD); total monthly value of transactions (million USD). We selected these attributes as the ones that we expected to affect M&A activity in any country.

Based on these characteristics, the composition of countries that best matches the UK is: 44% France, 3% Netherlands, 35% Spain, and 18% US. Using this weighted set of countries as a synthetic Control group, we can make a better comparison.

The effect of the referendum of M&A activity

As the figure below shows, using the synthetic Control makes the pre-referendum M&A activity lines almost perfectly parallel, which is one of the indicators that trend in the synthetic Control group is a good proxy of the trend in the UK. We could therefore assume that the trend in the synthetic group following the referendum is the trend that the UK would have experienced in the absence of the referendum. The figure shows a clear change, after the referendum, the UK line diverges from the parallel trend and drops, whilst the Control line remains roughly around the same level, implying that the monthly number of M&As dropped as a result of the referendum. This is shown clearly by the dashed line, which is our estimate of how M&A activity would have evolved in the absence of the referendum. The figure also shows the referendum campaign period, between the two vertical dotted lines. It clearly shows that the start of the referendum campaign was the main trigger, and M&A activity has not recovered since. The figure also suggests that the 2015 Bill on the referendum (28 May 2015) did not have an impact on the number of mergers (we test this formally and confirm it in our ongoing work).

Just how bad this drop was? One way of finding out is to compare the figures across the UK and the synthetic group. Formally, this can be in many different ways. Because we are looking at the difference across the two groups in the change before and after the referendum, we did this though a simple difference-in-differences regression. We knew from the data that the monthly number of M&As dropped by around 80 after the referendum. Our regression results (using the synthetic Control group) provide evidence that the effect of the referendum was a drop of 60 mergers per month. Therefore, around 75% (60/80) of the drop in the monthly number of M&As can be attributed to the referendum (that is, if you accept our study design as suitable for identification of causality).

The effect of the referendum on the composition of M&As

What about the size of these transactions? The monthly average value of announced M&As was just under 220 million USD pre, and around 220 million USD post-referendum (looking at our analysed period of 2015-2017). So can we conclude that the referendum did not affect the size of mergers because the average remained the same? Not really, it is easily possible that large mergers were affected differently from small mergers and still the average would remain the same. Our next task is to test if this is true.

When wanting to estimate how an event changes not the average M&A value but its different levels, one could look at how its quantiles changed post-referendum. Quantiles are the observations at every 5th percentile. Imagine that M&A value observations are ranked in ascending order. The first quantile of M&A values tells us the value below which lies 5% of the observations (i.e. the 5% smallest M&A values). For example, the first quantile (5%) of M&A values in the UK before the referendum was around $0.6 million. This means that 5% of the M&As in our sample were smaller than $0.6 million in value. What about the other end? The 18th quantile (90%) of M&A value was $214 million before the referendum. This means that 10 % of the mergers were larger than this amount.

Pre and post-referendum percentiles of M&A values

Looking at some summary statistics on how quantiles changed post-referendum already suggests that there was a differential impact the referendum had on M&A activity. For example, the 25th percentile of M&A values was $5.7 million pre-referendum, and $4.5 million post-referendum. What does this mean? That the smallest 25% of all M&As in the UK are now smaller than what they were before the referendum. What about the largest mergers? The 90th percentile of M&A values was $214 million pre-, and $250 million post, i.e. the largest 10% of mergers have become larger.

So it would appear that in terms of merger value, the small mergers have become smaller, and the largest ones became larger. But these are just simple readings of the data. Does this difference hold under a similar design we introduced above, i.e. do these differences hold when compared to our synthetic Control group? To the extent that the answer is yes, we can attribute these differences to the referendum.

To compare the changes to the synthetic Control group, the idea was the same as above (except here we used a quantile difference-in-difference method). We assume that the synthetic Control represents how the change in quantile values would have developed in the UK without the referendum. The figure below shows that the referendum had a different effect on M&As of various sizes (i.e. confirms our reading of the summary statistics). To help understand what these figures mean, take the figure at 0.25, which is -1.86. This implies that post-referendum, the value of the merger that is at the 25th percentile is around $2 million dollars smaller (we control for exchange rates changes, and we also control for time fixed effects for any other to make other effects time independent). Therefore the smallest 25% of mergers have become smaller as a result of the referendum. We can also see that this estimated $2 million difference is larger than the observed difference of $1.2 million. This suggests that the actual effect of the referendum is larger than the observed effect (the value at the 25th percentile would have increased without the referendum).

The results show that the effects up to the 75th percentile are significantly negative. For the 80th and 85th there is a lot more variation, and the very largest mergers seem to have become larger as a result of the referendum (see the positive values at the 90th and 95th percentiles). This is in line with what we saw in the summary statistics, small mergers became smaller and the largest mergers became even larger.

Looking at the transaction value alone doesn’t give a full picture about the underlying changes that happened after the referendum. For example, given the size of the acquiring company, is it the larger or smaller deals that suffered from the referendum? First we calculated the relative value of the transaction in comparison to the size of the acquiring business – where this data was available. Then we looked at how the quantiles of the proportionate value of the transaction were affected by the referendum. Using the same method as above, we estimated the quantile effects. The figure below shows that given the size of the business, it was mainly the larger deals that were foregone after the referendum (the deals that were small in comparison to the acquiring business’ size were not significantly affected). This makes sense, with increased uncertainty and risk, firms would be most cautious about the deals that a large relative to their own size.

Finally, we looked at which businesses were most likely affected by the referendum. We used the total revenue of the acquiring company in order to find out if smaller or larger firms were the more likely to lessen their M&A activity. The figure below reveals following the referendum, the distribution of firms that engage in M&A changed, larger businesses have become more M&A active(for example, the minimum size of the top 10% of businesses is now around $3600 million larger than before the referendum).

These results – as emphasised in the introduction – are preliminary, although given that the design of our study is already in place, we believe it is unlikely that these figures will qualitatively change with further refinement of the estimated models. Moreover, we have already experimented with various different model specifications and the headline results seem robust to all changes.

What do these results tell us?

First of all, the referendum caused a drop in M&A activity in the UK. This is bad news. The vast majority of mergers (unless they have a significant adverse effect on competition) have the potential to contribute to social welfare for example by reducing transaction costs, or by enhancing the efficiency of the merging businesses. If competition is left undisturbed, these benefits are passed on to consumers in the form of lower prices. When there is a setback in M&A activity (and a 15% setback is a rather sizable one), it means that some of these potential benefits are foregone.

Moreover, it appears that not all mergers were affected the same way. The finding that the distribution of mergers has changed as a result of the referendum implies that some mergers were more affected than others. Small mergers have become smaller and some of the very big mergers are now even bigger. Moreover, the largest businesses became more likely to carry on with their M&A activity post-referendum.

One interpretation is that cross-border mergers – that tend to be the larger in value – were less affected by the uncertainty.[2] However, there is a potentially more worrying interpretation. Large businesses have better means for rent-seeking behaviour. When we find that the largest transactions and the largest businesses were not hindered (and in some cases were even spurred) by the referendum, one inevitably worries: how much of this differential effect is due to the fact that these larger businesses are cushioned from the increased uncertainty, thanks to their rent-seeking behaviour.

Transitional periods are never good for businesses and consumers, but what makes it even worse is that businesses do not seem to be equally exposed to the same risk from the increased policy uncertainty, and those that are more likely to have political influence seem more protected from these risks.



[1] Special thanks to Ioannis Pappous and Luke Garrod for the stimulating discussions on this work.

[2] We do not believe that this was a case of capitalising on the falling GBP for two reasons: (1) most of the change happened before the pound fell, and (2) we control for the change in exchange rates in our estimated models.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: