“In the defined benefit space schemes are moving towards de-risking and moving out of equities. This is why factor investing in the fixed income space is quite interesting.”
Graham Wardle, BESTrustees
PI: Why are more and more investors using factor-based strategies?
Olivier Laplénie: We tend to see three categories of investors potentially using factor-based strategies. The first are the investors who are unsatisfied with the performance of their active managers. We have met quite a few of these. The new factor-based strategies seem interesting to them and, generally speaking, they are more cost efficient than their active manager. Typically, in terms of pricing, factor-based strategies sit in the middle between pure passive management, like index replication, and traditional active management. So an investor’s thinking is: “If I’m able to get a bit better performance than the benchmark but with a lower cost, why not try these strategies?”
Secondly, we are seeing investors coming from pure passive index-based strategies who are looking to enhance their return. Since they were comfortable with systematic passive strategies, systematic factor based strategies are a natural evolution of their investment process.
Finally, the third type of investor is those who have a growing awareness of factor investing. These investors are starting to analyse the global factor exposure of their portfolio. They may have discovered, for instance, that they have a value bias with their active managers and they are therefore interested in custom multi-factor strategies. This means the weight of the factors can be tilted towards the exposures that they do not have in their portfolio, such as momentum. What is interesting is that this is a move from an asset-based allocation framework to a more factor based allocation framework. We are still at the beginning, but it’s a start.
PI: So you expect to see a shift in this direction going forward?
Laplénie: Yes, but it will definitely take some time. We are still in the early stages here in Europe.
PI: How are trustees using factor investing?
Tony Charlwood: For defined contribution (DC) funds, which approach about £1bn now, we use target date funds. In the growth phase it’s predominantly invested in equities, mainly through index-tracking funds. Charges are competitively set in the DC world so using index-tracking funds helps with the issue of fees.
About a year or so ago, our investment manager suggested we move some of that pure index tracking exposure into global factor funds. We see the move slightly away from pure index-tracking funds to factor funds as a way of harvesting some additional alpha. At the same time, although factor fund fees are higher they are clearly lower than those for pure active managers. So by having a mixture of indextracking funds and factor funds, we can still offer a competitive charge to employers and members.
PI: Where does this sit in the active/passive debate?
Charlwood: It is a question of definition. Given the complications of the charge cap, we wouldn’t have pure active equity managers. So we see it as a halfway house. The factor funds we are using are global multi-factor funds where the mix of factors is managed dynamically by the manager, so that’s clearly an active element.
Kate Hollis: Like anything else in investment, it’s a spectrum. Drawing hard dividing lines is difficult. So when we started looking at fixed income smart beta we started at the passive end and designed better benchmarks. If you are designing better benchmarks you can incorporate tilts into your benchmarks. Now, whether you call that passive or whether you call it factors is a matter for discussion. As you move further away from passive and more towards the active end, the factor exposures are more explicitly used for adding excess return and then you move into full active.
From time-to-time, we have animated debates about which side of the line this or that particular strategy should fall in and it doesn’t worry us if clients want to classify it differently or use it for different things. It’s all about providing solutions that work for the clients’ portfolios in the context of the portfolio as a whole.
Andrew Peach: It doesn’t neatly fit into either bucket for me. Coming from the client’s perspective, what matters is the reference point. That reference point for a lot of clients is the market benchmark, be that equities or bonds. So if something is designed in a different way, that’s going to introduce tracking error relative to where that client is coming from. That, in itself, is an active decision. Investors, therefore, need to have patience that this will work over the long term. However, the construction can be considered passive, and that’s where the cost benefit comes in.
Graham Wardle: In the fixed income space it seems that investors have always been on the active side. There is very little passive investment there so this is a way of reducing the cost of investing in fixed income.
Hollis: There are some asset classes in fixed income, like government bonds, where it’s not worth i nvesting actively. Lots of our clients have passive gilts as part of their liability driven investing portfolio, but that’s rather different. DC clients that have a limited fee budget might get a better bang for their buck in equities or in another asset class, in which case it might make sense to do fixed income passively. Wardle: In the defined benefit space schemes are moving towards de-risking and moving out of equities. This is why factor investing in the fixed income space is quite interesting.
PI: The cost benefits sound appealing, but how are these strategies performing?
Wardle: In the fixed income active space there have been recent moves into a lot more emerging market debt, private debt, all of that sort of thing, which is undoubtedly active and quite expensive. It has produced some fairly good returns, but if you can do it in a more efficient way via factor investing then, yes, that will work.
Hollis: There are some things that active fixed income managers do consistently and systematically which can be replicated and they call it ‘alpha’. It might generate excess return, but if it can be replicated systematically it’s quite difficult, in my mind, to call it ‘alpha’. If you can do the same thing systematically at a much lower cost then that’s obviously for the benefit of the investors.
Charlwood: If you analyse a portfolio’s most active fixed income managers, they always have a bit of extra yield there as an under-pinner and that can easily be replicated.
PI: When you get beyond value, low volatility and momentum, how are the more niche factors performing?
Hollis: A good strategy which is well constructed should perform. A bad strategy probably won’t and so it’s not enough just to say: “I got factor exposures.” You have to be constructing your factors in the right way, you have to be combining them in the right way and implementing them in the right way. Momentum, for example, can be a high turnover factor, but in fixed income you can bleed away all the excess return because you’ve turned over your portfolio very quickly. That’s one of the important things about factor investing in fixed income as opposed to factor investing in equities.
Peach: The construction is as important as the factor.
Charlwood: If you look at the factor funds that we are using, the emphasis is on persistent factors that over a reasonably long time period will be expected to outperform. There might be certain market conditions where certain factors won’t, but again, it emulates back to this question of turnover. You can’t keep churning the portfolio or all the benefits are absorbed in transaction costs.
PI: Quant-based models have been used in fix income for some time and now we have factor investing. What is different this time?
Peach: If it’s active management and it’s ‘alpha’ that’s being sought, then the quant model is the intellectual property of the manager and it’s fiercely guarded. If it’s a ‘factor’, it’s a market exposure and it needs to be understood and, therefore, it needs to be more transparent. From a client’s perspective, that’s the key difference.
Laplénie: Transparency is one of the differentiators between factor investing and the older quant models. Another fundamental differentiator is that factors are based on a strong economic or financial rationale, not just statistical findings. If you think about the quality factor in the credit space, for instance, leverage, it is something that has been used for decades by managers. It is something which is widely understood, which you can replicate yourself. That’s very different from old quant models, which were mostly doing statistical arbitrage. They were running sophisticated and complicated regressions on huge universes, and finding relationships that no one understood. They ran that and expected it to work, and often it didn’t, except for some of the largest hedge funds. So this idea of using factors with an economic and financial rationale, and not only a statistical finding, is one of the key differences.
Factor-based strategies have a much lower turnover than older quantitative models because they tend to use low frequency indicators. We were talking about the leverage of a company. It changes, at best, on a quarterly basis and so it generates lower transaction costs. The other big benefit of having factors which have a low turnover is around the capacity of the strategy. An old statistical arbitrage strategy generally had a low capacity because if you put too much money into the strategy, the arbitrage disappears. If you put money into a quality factor it won’t affect the market. People have been doing it for decades and it has not disappeared, so we feel much more comfortable running these strategies with billions of assets than statistical arbitrage strategies.
What is different is that we are benefiting from decades of experience in quantitative portfolio management and this experience has been integrated into the construction of factor-based strategies, particularly on the risk management side. A good factor-based strategy pays close attention to its absolute risk, but also its risk compared to the benchmark. By using these decades of experience in equities and other asset classes we have been able to build more robust models. In a way we have learned from past mistakes, which is critical in the investment process.
PI: Does this mean investors are switching out of fundamental analysis?
Hollis: They’re doing it in a different way. Factors in many cases are based in fundamental criteria, but you are just looking at them or finding them and expressing them differently. To put it another way, people have used quant screens in fixed income for ages as an easy away to identify bonds that might be cheap. They then did fundamental analysis to find out if they were really cheap or not. Factor-based strategies will take a quantitative screen and also other indicators and using just that we’ll try and find a bond that’s cheap.
So I don’t think it’s different. It’s complimentary, but there are things like the low risk factor which you wouldn’t find explicitly in fundamental analysis at all. People tend to buy lower quality longer duration bonds, whereas low risk factors intuitively are about better quality shorter duration bonds. That’s one of the things about factor investing. It can diversify the excess returns you get from fundamental managers because it has factors in it, perhaps, that you don’t get through a common-or-garden active manager.
Charlwood: There has to be a fundamental justification for honing in on certain factors. There’s always a danger that data mining is used to produce a range of factors, but for us we would like to see a rationale underpinning the use of these factors as well as a vigorous analysis, even if it’s only using back-tested data about how they will actually perform.
PI: Liquidity is one of the differences between factor investing in fixed income and equity, but how else do they differ?
Laplénie: There are obviously some similarities in the styles or, let’s say, in the factor families that are used. Typically, we have quality, value, momentum and low risk. At BNP Paribas Asset Management we do not consider size a factor unlike others.
Wardle: In the fixed income space presumably by value you mean real yield?
Laplénie: It depends. We split fixed income into three segments: government bonds, credit and FX. We use different indicators to assess the constituents of each universe. So for credit the indicators are comparable to equities, but for FX what is a quality indicator for a currency? Also for government bonds, what is quality? For government bonds we use macro-based factors linked to the performance of the economy of a given country and factors based on the evolution of prices and credit in the economy. What is most striking within credit is if you compare the book value of a company to the price of its equity. You would typically buy equities with a low price-to-book ratio, but if you do that in the investment grade credit space you would lose a tonne of money. So in the fixed income world you want to take this factor in reverse by avoiding companies with a low price-to-book ratio. This example demonstrates that when designing good factor-based strategies you need people who have fixed income experience, not just those coming from the equity space saying: “Okay, it worked for equities so let’s apply it to fixed income.”
Hollis: The other difference is that equities are perpetual and they are at the bottom of the capital structure. In fixed income neither of those are likely to be true and fixed income can be anywhere in the capital structure. So when you’re looking at value you can’t just look at the issuer level, you have to look at the bond level. When you are looking at a 30-year bond you can’t just run its track record for 30 years because it starts as a 30-year bond and ends as a six-month bond. So you have to be able to take all of these things out at each security as you analyse it and as you run your back tests, but this is difficult. Also getting that data is difficult and expensive which means unlike equities there are fewer houses doing factor research at the stock level in fixed income. There are plenty of people doing it on futures and swaps on derivatives, but at the individual security level it’s much more difficult to do.
Charlwood: That also means, of course, there is probably, in theory, more opportunities. The breadth of Barclay’s global indices is enormous.
Laplénie: That’s between 15,000 and 20,000 bonds. So yes, more opportunities, but it is of course more difficult to analyse bonds. We launched our credit strategies later than our rates and FX strategies, because we spent a great deal of time developing our customised database that maps bonds to equities. For the equity of the issuer there has been no public database giving an accurate mapping over the last 20 years.
This information is key, as bonds are located on a different part of the capital structure. We employed a two-stage process to achieve the mapping. First, we used an algorithm, however this proved to be unsatisfactory. I then set my credit specialist portfolio manager an objective to review the mapping by hand. It was an enormous task, with a total of about 20,000 bonds. We needed to take into account all corporate actions over the past 20 years and check explicit and implicit guarantees, for which we are looking directly into the prospectuses of the issues. It took 18 months and that’s why there are less managers today providing credit strategies than in the equity space, because you need to put in the upfront effort to build this up. This is proper mapping.
Peach: The data problems inherent within fixed income markets mean that it is more complicated to do the analysis and to design and construct an index, so there are clear transparency challenges. Within equities, if you talk to a client about the ‘value’ factor and you define it as price-to-book; that makes intuitive sense to a lot of people – it is buying stocks cheaper than the book value of the company suggest that it ought to be. That comparison is harder in bonds because of the data difficulties; you can compare them to the spread of other similar bonds or to the equity volatility of the company. Given these complexities, at the end of all this, you need to make sure that there isn’t a wonderful product that is every bit as expensive as a traditional active product would have been.
Charlwood: In practice most institutional investors, certainly in corporate bond portfolios, are managed actively now. There’s been a trend towards total return investing in bonds and, again, coming up with a factor product, which is almost as expensive, is a bit of a hard sell.
Laplénie: We know that if we want to be able to sell the strategy it needs to be cheaper.
Wardle: How much research are the investment consultants doing into this?
Hollis: We’ve been doing credit smart beta for 10 years. We only started looking at factor-based credit at the security level about four years ago, because it was only about four years ago that there were products coming out. At the passive end of the strategy we’ve been doing it for 10 years. At the active end of the strategy we’ve only been doing it for about the last four.
Charlwood: Another differentiating factor is there’s now a whole body of academic research on equity factor investing, but it’s still quite light on fixed income investing.
Hollis: If you’re going to be doing the academic research you have to be able to test it, which means you need the clean scrubbed data to test it with.
Wardle: I’m not seeing much evidence yet of investment consultants putting it forward to pensions schemes.
Peach: I would concur from our experience. I don’t think that the challenge here is anything to do with factors necessarily. It’s to do with your traditional credit benchmark, which has been cannibalised to a degree by the extent of a pension scheme’s leveraged liability management processes. It’s not that there isn’t an interest in factor-based investing; it is that all of a sudden the floodgates have opened to different opportunities for earning a premium above government bonds. That may well change. Asset performance over the last five years has been better than expected, not least in the equity markets.
Pension schemes are generally better funded than they were five years ago. More schemes are now considering end-game portfolios and may well be coming back to unleveraged fixed income investments with a focus on buy-and-maintain credit. Factor-based investing may well have a large role to play there.
Wardle: Yes, that makes sense certainly and without doubt the better funded defined benefit schemes are moving away from equities and more into fixed income of some sort.
PI: How do you see the quality of the advice being given?
Wardle: There are some very good investment consultants out there. Certainly all of the major firms would be included in that. It was just that I, personally, have not seen much in the fixed income space yet where factor investing is being promoted by the consultants. I’m sure it will come because it certainly happened in the equity space. I’ve been mainly talking about defined benefit, but in defined contribution in the growth phase it’s pretty much entirely equity-type factors being promoted.
PI: So factor investing is not as prevalent as it is in the equity space. Do you see that changing?
Hollis: There’s a long and dishonourable history in the financial markets of people taking a good idea and sticking it as a label on all sorts of things that it isn’t. I’m starting to see that in fixed income, I don’t know whether any of you are, but, as we just discussed, it’s difficult to do it well and there are only a few houses capable of doing it well, but there are a lot of people starting to promote factors in fixed income.
Peach: The amount of white papers that I’ve seen on fixed income has increased. It took some time for that to get going in equities. First there was academic research, then the more commercial research followed and when it turned into something investable clients became interested. We’re not quite there yet on the journey with fixed income, but I’ve certainly seen more white papers.
Charlwood: One thing I will be asking consultants or managers on factor investing in fixed income is, is there a liquidity issue amongst those 20,000 fixed income bonds? What’s the ability to trade here?
Laplénie: A huge part of our work as quantitative portfolio managers is to make sure that these strategies make investable portfolios. In the credit space we spent quite a bit of time working on our liquidity filters and robust portfolio construction to make sure that the alpha would not be eaten by the transaction costs. To do that we integrated turnover and transaction costs at the heart of the portfolio construction process.
BNP Paribas Asset Management has been trading credit for many years and we have a huge database of trades on all kinds of forms. So we took this database and used it to build a model to estimate the total transaction cost incurred when trading a credit bond. This transaction costs model was then integrated into the optimiser used to construct the portfolio. When the optimiser switches a bond in the portfolio it will make sure that the expected alpha that you will get from switching the bond is significantly higher than the transaction cost you will incur. This transaction cost model was built a few years ago when we started with our low risk mono-factor strategy. We now have a four-year track record for this transaction cost model.
We recently did an analysis of the performance of the model and it turned out that we were in line with real-life transaction costs. We were actually lower for European credit and that’s why we feel comfortable trading credit bonds. Regarding liquidity for US denominated bonds we have TRACE (Trade Reporting and Compliance Engine) data, which is a system into which every bank and broker has to submit its transactions. Liquidity can thus be evaluated reliably at the bond level. For a euro-denominated bond, it’s a bit more complicated because we have no such systems and here we rely on the size and age of the issue.
We also rely on our specialised portfolio manager who before trading, checks if it is liquid. If it’s illiquid it’s flagged as such and removed from the portfolio and the model is re-run. So there are ways to deal with these higher transaction costs which are to take them into account and design models with a lower tolerance for turnover.
Charlwood: That’s quite encouraging because if you read research then you often get a little note in the small print at the end that this doesn’t take into account transactions costs.
Laplénie: All our models take into account transaction costs because I know a lot of factors which can perform without transaction costs. Factors which perform after transaction costs are much, much rarer.
Peach: You take either your regional or your global bond factors and then you say you control that for liquidity. Have you got a feel for how much of the sweetie shop is left by the time you’ve done that?
Laplénie: What I can tell you is after all the filters we apply, including the mapping, we are left with approximately 70% to 80% of bonds in the corporate investment-grade space. It’s more around 50% for high yield bonds, but that’s still a significant amount because there are thousands of issues.
Hollis: Even active managers in some of the high-yield space, where bonds trade once every three months if you’re lucky, won’t touch them or they will buy them and never sell them.
PI: Are investors playing a role in developing new factors?
Laplénie: Not so much new factors because it’s quite hard to find new factors. Investors are becoming aware of their implicit factor exposure. It is well known in the equity world, but in the fixed income world investors are starting to become aware. They are starting to run analysis on their global portfolio, including all their active managers. We can help these investors with this analysis if they wish to, but most investors do not like to share their data with us. Generally what we will give them is a detailed breakdown of the performance of our strategies and a detailed factor attribution, so that they can use this data to analyse the factorial exposures of their portfolios themselves.
Hollis: When you start screening portfolios you need to be a little bit thoughtful about how you are doing it and if you want to tilt your portfolio one way it may implicitly mean that you are tilting it a different way for different sets of factors or a different universe factors.
Laplénie: Factor-based performance attribution is a huge quant topic. That’s probably the most sophisticated part of the investment process. You really need advanced statistical techniques. We have PhDs working on that.
Charlwood: One issue that’s developed in factor investing in equities is the concept of crowding. Too many investors are trying to get into the same factors at the same time and it’s arbitraged away. I suppose if you’re a bit early into the game of factor investing in fixed income crowding might be less of an issue.
Laplénie: Today it is not an issue because it’s really early. We’ve seen some investors actually switch in Asia and in Europe, but it’s limited. There is a large capacity for this strategy because not everyone agrees on how to construct factors. Different managers have different factor portfolios so they are not buying the same bonds. Also, the factors we are talking about – quality, low risk, momentum and things like that – have been used implicitly by active managers for decades without causing excessive market disruption. Crowding risk might be higher in some less liquid areas, like high yield, but it’s quite limited for now.
Charlwood: How do you deal with things like excessive sector biases when screening for factors?
Laplénie: We pay close attention to building what we call pure factors. The factors should have no beta bias or sector bias against the benchmark. All our indicators are computed at the sector level and so we aim to have only low sector divergent from the benchmark.
Hollis: But you can control things in the same way as you would in anything else. A couple of years ago when energy was really cheap there were caps on sector exposures and so you would have been maybe max weight in energy but not 100% of the portfolio.
PI: If you look at the surveys, reports and press releases in my inbox you would think that the world was only interested in ESG. Is ESG playing a larger role in factor investing?
Hollis: I’ve seen emerging real market debt products screened for ESG and using it as a quality factor. Not so much in credit though. They’re not quite the same thing. Again, it’s another of the differences of fixed income and equities that ESG in credit needs to be expressed differently than the way it is expressed in equities.
Laplénie: For us ESG is a big topic. BNP Paribas Asset Management has invested a lot in building our internal ESG team, which gives us scores at the issuer level. The way we use the ESG score is not as an alpha factor. We use it as a constraint on the portfolio construction side, meaning we have scored all our bonds and we are ready to select investable portfolios. It is at this step that we include the ESG aspect.
The way we do this is a mix between excluding the worst-rated issuers and selecting more better-rated issuers, which improves the global average ESG score of the portfolio. Also something to note is that we use a best in-class approach in the way we score issuers. They are scored at the sector level in order to minimise the sectorial biases which could easily arise within ESG.
Charlwood: We’ve been looking at it for equity investing. There are two ways of doing it. Narrow the universe down, for example, using the MSCI World Socially Responsible Investing index. Then you are automatically screening out the worst and having a bias in favour of the better companies. The other thing is to apply a tilt. There is always a question there of whether you exclude carbon companies totally or just tilt away from certain ones in favour of others. There are funds that have been developed, such LGIM’s Future World Fund, which is used by HSBC’s pension scheme, which is a multi-factor fund with a carbon tilt on top.
Hollis: But again, you need to be thoughtful about it because in the credit universe there are issuers who are not companies. So, for example, you can look at EDF. It is the largest nuclear operator in Europe and also has lots of wind farms, so where do you put that on your climate change spectrum? All sorts of government-guaranteed issuers would not have a climate bias at all, like German sovereign-owned agency KfW. Again, you need be a bit more thoughtful in credit.
Charlwood: You can screen them out entirely, it will probably be significantly underweight the oil and gas sector or you can apply it as a tilt. Underweight say BP and Shell but then overweight clean energy companies. Hollis: Or you can apply a different tilt and overweight the most improving companies.
PI: Will awareness of factors in fixed income improve when the data improves?
Wardle: Olivier, I was interested to hear what you had to say about the effort you’ve put into the data side.
Laplénie: There is no public database available, but at some point there will be one. No public database means there are more opportunities. If you have the technology you can grab these opportunities, if you don’t have it then too bad. As an investor you might benefit from this situation.
Wardle: That’s true, but there’s an element there if someone comes to you and it’s a trust me situation whereas there is more research in equities then there’s more funds out there with a track record.
