in focus
Residual Value in Mature Private Equity Funds
By Pierre Garnier, Associate, Nik Morandi, Partner and
Dr. Ian Roberts, Senior Research Associate
November 2014
NOVEMBER 2014
introduction
Long-term value creation remains at the heart of private equity’s enduring appeal, with funds following a lifecycle
that is very often in excess of 10 years. As a result, private equity has long been seen as a buy-and-hold investment,
attractive to investors with longer time horizons and the patience to reap the rewards that top-performing private
equity managers have historically generated for investors.
The development and ever-increasing scale and depth of the secondary market has changed the mindset of many
LPs and, consequently, how they approach the asset class. Private equity investors can now use secondary sales as an
effective portfolio management tool, to liquidate all – or a portion of – their portfolios at short notice. This suggests
that private equity investors are today better-positioned to manage the lifecycle of their fund investments. Moreover,
the secondary market has now grown to accommodate an extremely broad range of transaction profiles, including
very mature portfolios and tail-end vehicles. These have often attracted significant interest from certain secondary
purchasers, keen to take advantage of perceived near-term distributions.
This raises the question of whether there is a break-point after which it is not necessarily rational to hold on to a
portfolio of private equity funds, taking into account an investor’s cost of capital for the asset class as well as the
investment opportunities that may be available elsewhere. This can, of course, include the option of reinvesting sale
proceeds back into private equity via less mature portfolios that offer the prospect of greater future upside.
key findings
>> This study has found that, based on the median fund in our dataset, value creation and cash generation were
limited after year nine in the life of a private equity fund, relative to total returns.
>> Judged in terms of Total Value to Paid In (“TVPI”), the median fund generated the vast majority of value for
Limited Partners (“LPs”) by years eight to nine.
>> Similarly, there was strong evidence to suggest that a large number of funds in the universe under consideration would not have generated, from years eight to nine onwards, a net internal rate of return (IRR) much in
excess of 5% for existing LPs.
>> Depending on an individual LP’s target rate of return, retaining mature portfolios until they self-liquidate may
therefore not always be a rational decision.
>> Lower quartile funds appear to be “late developers”: for example, total value creation for third quartile funds
was significantly more back-ended compared to first and second quartile funds.
>> Whilst selling a mature PE portfolio might require accepting a discount on residual Net Asset Value (“NAV”),
this discount may have a limited impact on the portfolio’s overall returns from inception to sale.
>> A subset of funds still yielded a significant proportion of total distributions subsequent to year nine, reaffirming
the need for effective and detailed due diligence as a key component of any decision to sell older portfolios.
>> The conclusions from this study may be relevant for investors seeking to manage “zombie” fund exposures, or
LPs seeking to actively manage their private equity portfolios on a more tactical basis.
1
Past performance is no guarantee of future results. Future returns are not guaranteed and a loss of principal may occur
3
Pantheon has conducted a quantitative study to understand, firstly, whether there is an inflexion point in the life of
a private equity fund where expected value creation and distributions may become negligible relative to total value
already generated. In addition, we set out to understand whether there is an identifiable point in time after which,
irrespective of the historic performance of a private equity fund, it is logical for an investor to sell. These two questions
are closely related to one another. The former will be impacted by how funds generate value through their lifecycle,
whilst the latter will also be influenced by individual investor preferences concerning expected or target returns from
the asset class, as well as the price that a mature portfolio can potentially secure in the secondary market.
Taken together, the answers to the above two questions are likely to be highly relevant to any long-term private equity
investor seeking to maximize the total return potential from a private equity allocation.
Value creation over time
We collected performance data on almost 700 private equity funds (buyout and venture), and looked at how fast
these funds created value as a proportion of their total performance2.
Figure 1: TVPI progression over time
Median
120%
25th percentile
75th percentile
100%
80%
60%
Median {(TVPI i,T – TVPI i,t)/(TVPI i,T – TVPI i,0)}
with 1 ≤ i ≤ 692
0 ≤ t ≤ 80
t being number of quarters
56%
43%
40%
19%
20%
28%
12%
20%
4%
2%
%
9
10
0%
-20%
0
1
2
3
4
5
6
7
8
11
12
13
14
15
16
17
18
19
# of years
We show in Figure 1 the proportion of total TVPI that remained to be generated, at any given point in time, for our
median fund as well as the break-points for the 25th and 75th percentiles within our dataset. The analysis clearly shows
that starting from years eight to nine the median fund had already distributed, or had accounted for in its valuation,
the vast majority of its ultimate gain. In fact, by year nine the median fund had already generated 96% of its total
gain when measured by TVPI3.
This poses an interesting question for investors looking to crystallize their returns to date. The results of our analysis
suggest that, historically speaking, holding on to a median or “average” portfolio beyond year nine may not have
been an entirely rational economic decision. Freeing up and recommitting that capital elsewhere may have enabled
the holder of this portfolio to generate stronger returns, or at a minimum provided the opportunity to access greater
upside, through re-committing the proceeds to a less mature portfolio.
Interestingly, the results provided no statistically significant evidence that venture funds followed a different pattern
to that of buyout funds4, nor that the TVPI progression of European funds differed from that of U.S. funds. There was
The methodology of the quantitative study is described in detail on page 11
Similarly, when looking at distributions, we found that the median fund had generated the significant majority (80%) of its distributions by the end of year
nine, increasing to 88% by year 10. As at the end of year 12, only 2% of aggregate distributions remained outstanding for the median fund
4
Nevertheless, one interesting conclusion of the study is that the degree of dispersion in the TVPI profiles (relative to the median) was greater for venture
funds than for buyout. As a result, the need for effective due diligence in mature portfolios is arguably even more important for venture than for buyout,
whilst remaining critical for both
2
3
4
NOVEMBER 2014
insufficient data to conduct a reliable statistical test comparing the relative progression of Asia/Rest of World (“RoW”)
funds: however, the data suggested that these funds had a slower and more volatile TVPI progression than the rest of
our dataset. Our hypothesis is that these differences may reflect value creation persisting over longer periods, and GP
input being applied over a longer period of time, for RoW funds compared to those in the U.S. and Europe. We also
found no significant vintage dependency in the results, implying the results were not skewed by any specific vintage
or associated set of market conditions5.
However, as can also be seen from Figure 1, the results for the median fund masked a very high level of dispersion within
the overall dataset: 25% of funds (represented by the “75th percentile” sub-sample) had yet to deliver a significant
proportion of their aggregate performance at an advanced stage of their life. Indeed, by year nine, this subset had yet
to generate almost 30% of total proceeds (over 40% as at year eight). Conversely, a different subset – also constituting
25% of funds in the dataset – was exhibiting a higher TVPI between years six and 12 compared to final performance
at maturity, implying an investor was actually losing value in these funds from approximately year six onwards.
The high degree of dispersion displayed by the dataset undermines the significance that should be derived from the
median fund findings. It also reaffirms the need for a detailed, fund-specific understanding of the residual value left
in more mature funds. The data shows that this in-depth knowledge should be a key component of any decision
to sell. Unless an investor has formed an in-depth understanding of the potential for future value creation beyond
a certain date, on a fund by fund basis, the decision to sell can expose the investor to significant risk of foregoing
material future upside.
The key conclusion one can draw from the above analysis is that a holder of the median portfolio should have seriously
considered exiting funds once they reached nine years of age. However, significant caution was warranted for the
broader dataset: the holder of 75th percentile funds in our analysis would, for example, have lost out on substantial
remaining value through a sale at that point.
The relevance of fund quartile rankings
Estimating or anticipating which funds are likely to retain significant upside is at the core of what experienced secondary investors seek to understand, and what sellers of private equity portfolios should also aim to achieve from active
portfolio management. However, in order to identify funds likely to possess significant residual value – i.e. those that
fall into the 75th percentile category in our example above – can we identify any potentially useful indicators?
We examined whether fund quartile rankings at any given point in time could have provided a useful, high-level
indicator of potential remaining upside. Differentiating funds by TVPI quartiles at the year of observation, it appears
that third quartile funds were in fact “late developers”: total value creation, whether in the form of distributions or
NAV growth, was more back-ended compared to first and second quartile funds (see Figure 2)6.
These results may seem unsurprising given we would expect top quartile funds (by TVPI) at any given point in time to
be those with a faster TVPI progression up until that date. But our analysis measures the relative development of TVPI
during a fund’s lifetime, not the absolute amount of the gain a fund generated7. So whilst this was our intuition, the
late development of third quartile funds was by no means a foregone conclusion.
There were only a few instances where this did not hold true and these exceptions systematically concerned third quartile funds
The TVPI profile of fourth quartile funds was also more back-ended compared to first and second quartile funds. However, overall the available performance
data for fourth quartile funds was more limited and less complete than for first to third quartile funds, which is why we excluded these results from Figure 2
7
The percentiles referred to in Figure 1 should therefore not be confused with performance quartile rankings. The percentiles are based purely on how
quickly each fund in our dataset reached its final TVPI outcome. As a result, the actual TVPI multiple of each fund at any given point in time was not relevant
in determining its percentile ranking
5
6
5
The third quartile analysis throws up other counter-intuitive implications. Within our dataset, at any given point in time
from year three onwards a seller of third quartile funds would have foregone a higher proportion of total TVPI than
holders of first and second quartile funds. The proportion of the total gain foregone from selling early (from year three
onwards) was therefore higher, and often significantly so, for the more poorly-performing funds within our universe.
The large spread between the TVPI progression of first / second quartile versus third quartile funds persisted all the
way through until year nine, as Figure 2 clearly demonstrates.
Figure 2: TVPI progression over time, by quartiles
1st quartile
120%
2nd quartile
3rd quartile
100%
80%
Median {(TVPI i,T – TVPI i,t)/(TVPI i,T – TVPI i,0)}
with 1 ≤ i ≤ 692
0 ≤ t ≤ 80
t being the number of quarters
55%
60%
37%
40%
27%
20%
10-11%
3-8%
0%
2-3%
2
3%
-20%
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
# of years
Figure 2 does not address the size of the gain foregone, and what this actually represents in terms of a performance
metric such as IRR or TVPI. We return to this question later.
We also carried out the same analysis as that summarized in Figures 1 and 2 on the distribution profiles of the funds
within our dataset, looking at how fast the funds reached their ultimate level of distributions (as opposed to TVPI). The
findings were consistent with the conclusions we reached for the TVPI-based analysis, with minor differences only. For
the median fund, 88% of total distributions had been generated by year 10 (as compared to year eight using the TVPI
analysis). There was again no statistical evidence supporting the hypothesis that venture funds followed a significantly
different trend to that of buyout funds, and limited evidence8 that European funds followed a significantly different
trend to that of U.S. funds. Lastly, as was the case for the TVPI-based analysis, the distribution pace of third quartile
funds was significantly more back-ended compared to first and second quartile funds.
The dilemma between holding and selling
While the results and conclusions presented above can help an investor understand when key inflexion points may
occur during the course of a private equity portfolio’s life, they do not by themselves answer the key question: should
an investor sell? And if so, when? At any given point in time, this trade-off between holding on for more value
growth versus monetizing existing residual portfolio value is governed by three variables:
(i)What are the returns I believe the residual value left in my portfolio is likely to generate going forward?
(ii)What pricing do I believe I could achieve for my remaining portfolio if I were to sell now via the secondary market?
(iii)What are the returns I am targeting for my private equity program overall?
The first consideration addresses the future cash flows the portfolio’s residual value should generate going forward.
We have analyzed the distribution profile of our dataset in order to provide a possible benchmark. Clearly, ex-ante an
Based on our dataset, the distribution profile of U.S. funds appeared slightly more back-ended compared to European funds. In fact, the difference in the
distribution profile of the median fund in these two geographies, as at years seven and eight, passed a test of statistical significance using the Wilcoxon ranksum test. Outside of these two years, the distribution profiles were not statistically significantly different. See the methodology on page 11 for further details
8
6
NOVEMBER 2014
investor will need to form its own opinion on what it believes may be the future potential of its specific portfolio, but
the results based on our historic dataset could prove helpful in anchoring expectations and in providing a reference
point for investors considering a portfolio sale.
In our analysis, we assumed an investor committed equally-weighted amounts in each of the 692 funds in our universe
on a primary basis and considered selling the entire portfolio on the secondary market at the end of year eight. Based
on a par sale, the median IRR the investor would have foregone when selling would only marginally have exceeded
5% (see Figure 3 below). Starting from the middle of year 10 onwards, this same seller would never have foregone
more than a 5% median IRR on the value left, again assuming a par-based transaction.
The corollary of this is that a decision to hold onto the portfolio from approximately year eight onwards would only
have generated an aggregate IRR of c.5% until final portfolio maturity, relative to NAV. There would therefore have
been a clear rationale for the holder of this median portfolio (or a holder of the median fund) to consider a sale, if
this future performance would have been dilutive to the overall performance of an investor’s private equity program.
Figure 3: IRR from residual portfolio NAV, over time
Median
35%
25th Percentile Median IRR
75th Percentile Median IRR
25%
Median {IRR t->T,i}
with 1 ≤ i ≤ 692
28 ≤ t ≤ 56
t being number of quarters
15%
5%
-5%
-15%
7
8
9
10
11
Sale / Purchase year
12
13
14
We found the above results informative, suggesting that some investor portfolios may simply “run out of steam”
by year eight. However, the dispersion in returns based on this analysis is again significant, as shown by the spread
between the 25th and 75th percentiles’ median IRRs in Figure 3. In economic terms, the sale at par of the 75th percentile portfolio at any point from year seven onwards would have translated into an IRR for the buyer that generally
exceeded 20%, implying the vendor left a significant amount of upside on the table at the point of sale. As noted
earlier, we believe there is no substitute for detailed knowledge of a portfolio in order to ensure any decision to sell
fully considers remaining upside potential.
RELEVANT CHARACTERISTICS
We list below examples of characteristics that we believe are relevant in assessing the potential for remaining
upside in older funds:
>> Is the GP in carry? Is the GP otherwise incentivized to continue to maximize portfolio value for LPs?
>> Has the GP since raised another fund? If so, is the GP likely to become distracted as a result?
>> What are the stage weightings within the portfolio10?
>> Is there significant public exposure left?
>> What is the average maturity of the underlying assets, rather than the age of the fund?
>> What is the GP’s valuation policy and does the GP have a track record of generating positive surprises at exit?
In an earlier study (Pantheon InFocus: “Valuation Exit Analysis”, March 2013) Pantheon identified stage-based differences (e.g. growth equity vs. buyout) in
average uplifts to latest holding NAV achieved by managers
10
7
This level of dispersion is also apparent when sorting our portfolio funds into performance quartiles, as set out in Figure
411. Assuming a sale at par in year nine, the IRR generated by the buyer of an equally-weighted portfolio composed
of the first quartile funds would have been 13%, versus only 5% for the acquisition of the second or third quartile
funds. Whilst the quartile-based return profiles shown in Figure 4 are somewhat “noisy”, in general top quartile funds
delivered greater tail-end performance than the rest of the sample set, and significantly outperformed third quartile
funds at every point within our observation period.
Unlike the seemingly counter-intuitive outcome we referred to under Section 3, these results suggest that quality
does partly drive secondary returns, assuming par pricing. Based on our dataset, this appears to hold true even at an
advanced stage of a fund’s life and despite top quartile funds reaching their ultimate level of performance at a faster
pace than the rest of the universe of funds.
Figure 4: Returns on remaining NAV, by quartiles
1st Quartile
25%
Median {IRR t->T,i}
with 1 ≤ i ≤ 692
28 ≤ t ≤ 56
t being number of quarters
20%
15%
2nd Quartile
3rd Quartile
13%
10%
5%
5%
0%
7
8
9
10
11
12
13
14
Sale / Purchase year
The impact of discounts: from a buyer’s perspective
We assumed in the above analysis that a buyer would have been willing to pay par. However, given the returns our
median portfolio would have generated after sale, it is certainly debatable whether buyers would have been prepared
to pay this full a price. In our opinion, and based on our own experience of the secondary market, buyers would more
typically have considered a purchase at a discount – providing that their views at the time of due diligence did not
significantly overestimate the future performance potential of the underlying funds.
Figure 5: Buyer’s median IRR, by discount
15% discount
5% discount
5% premium
25%
20%
15%
10% discount
Par
Buyer's straight cash target returns
Median {IRR t->T,i}
with 1 ≤ i ≤ 692
28 ≤ t ≤ 48
t being number of quarters
10%
5%
0%
-5%
7
8
9
10
Sale / Purchase year
11
Again TVPI-based
8
11
12
NOVEMBER 2014
In order to quantify the impact that a discount to NAV would have on our analysis, we calculated the implied IRR that
a buyer of the median portfolio would have generated by buying the residual portfolio at various discounts to NAV.
The results are summarized in Figure 5.
For example, in order to achieve a minimum IRR of at least 13% by acquiring the median fund in our dataset at any
point between years seven and 10, a buyer could not have paid more than a 15% discount to NAV. The required
discount to hit this minimum IRR threshold would have fallen from year 10 onwards to approximately a 10% discount.
Any other price scenario would have led to underperformance relative to the return target.
The main conclusion we draw from this analysis of our dataset is that, on a historical basis, a seller may have had
to accept a discount to NAV in the low double digits in order to dispose of a median portfolio to a well-informed
secondary purchaser. In turn, this could easily influence an investor’s decision to hold onto residual portfolio value,
versus selling. But how much difference should it have made in practice?
The impact of discounts: from a seller’s perspective
It is important to note that a discount upon sale, even if relatively large, may have a limited impact on the overall
performance a portfolio generates from inception to sale. The older the fund, the higher the probability that the residual
value represents a relatively small proportion of the portfolio’s overall size, and therefore the lower the probability this
discount will have a significant impact on overall returns.
Figure 6: Seller’s foregone gain, by discount
15% discount
10% discount
5% discount
Par
5% premium
12.0%
Median {1-[(Distrib.i,t + NAVi,t*purchase
price)/(Callsi,t)]/(TVPI i,T)}
with 1 ≤ i ≤ 692
28 ≤ t ≤ 48
t being the number of quarters
10.0%
8.0%
6.0%
4.0%
2.0%
0.0%
-2.0%
7
8
9
10
Sale / Purchase year
11
12
Figure 6 summarizes the percentage of the total portfolio gain that an investor would have foregone by selling the
median fund in our dataset, at various discounts, from year seven onwards. As can be seen from the chart, from year
eight onwards a holder of the median fund would not have given up more than 6% of total gains even when selling
at discounts of up to 10%. The sensitivity of foregone returns to the discount also weakens over time.
There are therefore two distinct return considerations for sellers of mature portfolios. The first is the implied IRR foregone
by selling early. As seen in Figure 5, discounts of up to 10% between years eight to 10 would have limited the foregone
IRR to a maximum of 13% for the median fund. The second consideration is the size of the gain foregone. Measured
relative to the total gains that would otherwise have been achieved, discounts upon sale of up to 10% would seem
tolerable, assuming an investor did not want to lose out on more than 5% - 6% of total portfolio returns.
9
In our view, the foregone IRR should be the more relevant consideration for most investors. In deciding to allocate capital
to various investment opportunities, the expected IRR can then be compared to the other investment opportunities
available at that particular point in time. The lower the expected returns available elsewhere, the lower the opportunity
cost of holding mature private equity portfolios until maturity.
Conclusions
We believe the findings from this study have applications that can help investors manage their private equity fund
portfolios more actively, and make better informed decisions on partial or full liquidation processes. Pantheon continually
monitors more mature portfolios on behalf of investors, selectively utilizing the secondary market if we believe doing
so will be in the best interests of our clients. The findings from this study will provide another datapoint for decisions
we take in this regard. Some of the findings may also act as useful benchmarks when considering secondary purchases
of more mature portfolios, or in determining how to manage so-called “zombie” funds.
There is strong evidence that a large number of funds in the universe under consideration for this study would not have
generated, starting from a relatively advanced stage of their life, annualized returns much in excess of 5%. However,
one key conclusion is the degree of dispersion found within our dataset. If this is representative of what could potentially
happen going forwards, it implies there may be large subsets of funds that can still yield very strong returns (> 20%
p.a.) on their residual value from year eight onwards. We believe an in-depth knowledge of underlying fund positions
is therefore essential, in order to allow an investor to develop a well-informed view on the performance their residual
value may generate, and to benchmark these potential upcoming returns with an overall target for the asset class.
The continued development of the secondary market has expanded the range of transaction opportunities available
on the buyside, but it has also provided new opportunities for portfolio management. The results and conclusions
summarized in this paper provide one more component of a sophisticated, pro-active investor’s toolkit.
10
NOVEMBER 2014
Appendix: methodology
Pantheon collected data on 692 funds with vintage years ranging from 1983 to 2004. Given the nature of the
study, we selected funds for which at least 10 years of performance data was available, and we therefore excluded
any funds whose vintage year was after 2004. When we refer to the quartile (or percentile) ranking of a fund or
subset in our results, this is based on quartile (or percentile) rankings at a particular point in time. Therefore, the
subset for top quartile (or percentile) funds as at period t is not necessarily the same subset of funds as at period
t+1. Quartiles are based on TVPI. Percentiles are based on the relevant measure being analyzed. The vintage year
of a fund is defined as the year at which we observe the first capital call. Additionally, the liquidation of the fund
is either defined as (i) the year at which the NAV becomes permanently nil; (ii) the last available year for which
information is available; or (iii) the end of year 20. If (ii) or (iii), the residual NAV will be assumed as a final cash
flow in TVPI and IRR calculations. Both TVPI and IRR figures are net of underlying fund fees and expenses. Fund
stages and strategies are early and late-stage venture, growth capital, buyout, mezzanine, turnaround, special
situations, distressed debt and balanced. i designates the number of funds and therefore cannot be greater than
692 (I=692). t designates the time, at a quarterly frequency, and therefore cannot be greater than 80 (T=80).
IRRs are always annualized. Each fund is considered to have the same fund size when we conduct our analyses
(we assume an equal-weighting to funds in our sample), and capital calls, distributions and NAVs are calibrated
accordingly. Vintage effects are neutralized as we synchronized the funds in time; for example, a median TVPI at
the end of year 11 is the median of all the funds’ TVPI at the end of the 44th quarter after their first capital call.
Tests for statistical significance were run using the Wilcoxon rank-sum test (for differences in medians) and the
Brown-Forsyth test (for differences in dispersions). Tests for vintage dependency were run by removing individual
vintages from the dataset and evaluating whether the residual dataset generated significantly different results
compared to results based on the full sample size.
For more information, please contact:
Pierre Garnier
Associate
Nik Morandi
Partner
Dr. Ian Roberts
Senior Research Associate
Phone: +44 20 7484 6200
Email: [email protected]
Phone: +44 20 7484 6200
Email: [email protected]
Phone: +44 20 7484 6200
Email: [email protected]
11
IMPORTANT DISCLOSURE
This publication has been prepared solely for illustration, educational and or discussion purposes. It does not constitute independent
research and under no circumstances should this publication or the information contained in it be used or considered as an offer,
inducement, invitation, solicitation or recommendation to buy or sell any security or financial instrument or service or to pursue any
investment product or strategy or otherwise engage in any investment activity or as an expression of an opinion as to the present or
future value or price of any security or financial instrument. Nothing contained in this publication is intended to constitute legal, tax,
securities or investment advice.
This publication may include “forward-looking statements”. All projections, forecasts or related statements or expressions of opinion
are forward-looking statements. Although Pantheon believes that the expectations reflected in such forward-looking statements are
reasonable, it can give no assurance that such expectations will prove to be correct, and such forward-looking statements should not
be regarded as a guarantee, prediction or definitive statement of fact or probability.
Pantheon has taken reasonable care to ensure that the information contained in this document is accurate at the date of publication.
However, no warranty or guarantee (express or implied) is given by Pantheon as to the accuracy of the information in this document,
and to the extent permitted by applicable law, Pantheon specifically disclaims any liability for errors, inaccuracies or omissions in this
document and for any loss or damage resulting from its use. Unless stated otherwise, any opinions expressed herein are current as of
the date hereof and are subject to change at any time. Unless stated otherwise all views expressed herein represent Pantheon’s opinion.
This document is distributed by Pantheon which is comprised of operating entities principally based in San Francisco, New York,
London and Hong Kong. Pantheon Ventures Inc. and Pantheon Ventures (US) LP are registered as investment advisors with the U.S.
Securities and Exchange Commission. Pantheon Ventures (UK) LLP is authorised and regulated by the Financial Conduct Authority
(FCA) in the United Kingdom. Pantheon Ventures (HK) LLP is regulated by the Securities and Futures Commission in Hong Kong.
All materials published on the Site are protected by copyright, and are owned or controlled by Pantheon as the provider of the
materials. If you download any information or software from this Site, you agree that you will not copy it without the prior written
consent of Pantheon or remove or obscure any copyright or other notices or legends contained in any such information.
Copyright © Pantheon 2014. All rights reserved.
12
London
San Francisco
Hong Kong
New York
www.pantheon.com
Seoul
Bogotá