Emerging markets, especially the BRICs, have attracted a lot of attention from international inves- tors. Brazil, the B of the acronym, has been the country with the largest share of the fundraising in Latin America: 83.8% in 2011 and 64.9% in 2012. (See Exhibit 1). Although there are many opportunities for international investors in emerging markets, there are also significant risks. The objective of this article is to analyze Brazilian PE and VC returns over the last 20 years and to see if Brazil was worthwhile for international investors.
Analyzing returns of private equity (PE) and venture capital (VC) funds is a challenging task, due to data availability and reporting biases. PE and VC are not obligated to disclose their investments and cash f lows. Therefore, even in countries with a long his- tory of PE and VC industries, there are no comprehensive databases for fund returns. The main data sources are obtained through PE and VC firms and limited partners (LPs) that voluntarily provide information to commercial data providers or directly to researches. In emerging markets like Brazil, which have a shorter history for the PE and VC industries, the challenge of access to data is even more significant.
Our analysis is based on the Spectra- Insper database. This private dataset has been built through a partnership between Spectra Investments, a Brazilian PE and VC fund of funds, and the Insper Institute of Education and Research, a leading Brazilian Business and Economics School. It contains informa- tion about firms, funds, and deals in Latin America. In order to protect the identities of the firms, funds, and deals, Spectra Invest- ments sanitizes the data before loading them into the database. Main sources of informa- tion are private placement memoranda (PPM) in Spectra’s possession, although some data were hand-collected by the CVM (Comissao de Valores Mobiliarios—Brazilian Security Exchange Commission), and other sources like Thomson TRAA.
In this article, we focus on fund-level information. Our sample contains data on 172 funds from 78 domestic and international organizations that were raised between 1990 and 2013. We were able to analyze perfor- mance of only 46 funds, with vintage year from 1990 to 2008, since we use only funds that exited or had liquidity events in 60% or more of their investment portfolios.
Our conclusions are that although the Brazilian PE and VC industries are young, the players are slowly maturing, with 72% of the firms operating for 5 years or more. The industry performed well over the last 20 years, compensating international LPs for investing in Brazil. The average gross return for the sample is 22% (with a median of 23%). The average multiple of money (MoM) is 3.6x its invested capital (with a median of 2.5x). We estimated the average holding period as 6.2 years (with a median 4.6 years). On average, Brazilian funds with vintage years from 1998 to 2008 outperformed U.S. funds in the equivalent vintages; this is a reversal from the first seven years of our sample.
This article contributes to the private equity litera- ture in emerging markets, since it is the first study that analyzes reliable PE and VC fund return data from the last two decades in Brazil.
The remainder of the article contains a literature review section, where we describe the main datasets and their limitations in the U.S. and Brazil. We include the main findings about fund performance, a data and analysis section, where we describe and analyze fund performance, and finally we discuss the results and the conclusions.
Exhibit 1. Evolution of Fundraising in Latin America
Source: LAVCA [2013].
Data are a major constraint in conducting research on PE and VC return, because firms are not required to disclose information about funds, investments, and cash f lows.
According to Higson and Stucke [2012], many researchers collect cash f low data directly through single or very small numbers of limited partners. This is a more reliable source of information than collecting data directly from a fund’s general partners (GPs), because LPs have strong incentives to maintain and supply accu- rate data. However, it is difficult to generalize from such studies. The fund portfolios differ greatly from one LP to the next, as there are wide differences in GP skill and performance. Commercial databases, such as Thomson One (VentureXpert), Preqin, and Cambridge Associates have been widely used by researchers. They cover a broader range of funds and have a long history of data, but information about cash f lows is missing in some cases, producing a systematic downward bias in the measurements.
The return from a PE or VC fund is only measured with certainty when the fund is fully liquidated and all cash f lows are completed. Net asset value (NAV) may be reported at “fair value” or cost, and it is very difficult to judge the quality of fair value in illiquid markets. Stucke [2011] shows that a large number of funds in VentureX- pert stopped being updated during their active lifetime, so the data record is truncated, while the NAV remains constant, distorting calculations of fund performance. For some funds, data are not updated and returns are underestimated. Yet other funds may be reluctant in disclosing “living dead investments” representing write- offs or those with a NAV close to zero. There are many discrepancies among commercial data sources in terms of fund size, vintage year, and fund type. None of the data providers is complete with respect to the PE and VC universe.
Harris et al. [2014] compared various commercial data providers. The Burgiss dataset is provided exclu- sively by LPs and includes all funds and cash f lows from those providers. The information is accurate, since LPs use the record for monitoring fund investments, and it is also up-to-date, given the requirement of quarterly reporting by main investors. However, each database has a potential bias. Burgiss has a selection bias, because it only contains data from the LPs that are reporting. Venture Economics is sourced by LPs and GPs and has problems with updating NAVs. Preqin is dependent on public filings by pension funds and therefore may be missing some of large funds that do not have public pension funds as LPs. It contains internal rate of return (IRR) and MoM information, but no cash f low infor- mation. Cambridge Associates has access to information from LPs and GPs who have raised or are trying to raise capital, and it may have a bias toward funds that are high performers.
Usually commercial datasets report net returns, or the returns actually received by LPs, so they are net of all management and performance fees. Gross returns represent the gains generated by the investments in the portfolio companies and do not account for fees received by the GP. They are usually reported on the deal level of return.
There are three common measures for a fund’s performance: i) LP annualized net IRR, calculated from the cash f lows generated through fund contributions and distributions to LPs; ii) multiple of money, where the numerator is the sum of all fund distributions and the value of unrealized investments, and the denominator is the sum of all fund contributions by LPs, including the payment of fees to GPs; iii) public market equivalent (PME), introduced by Kaplan and Schoar [2005], which compares an investment in a private equity fund to an equivalently timed investment in a public market. “The PME calculation discounts (or invests) all cash distribution and residual value to the fund at the public market total return and divides the resulting value by the value of all cash contributions discounted at the public market total return” (Harris et al. [2014, p. 12]). The PME calculation requires cash f low information at the fund level and represents a market-adjusted multiple of invested capital. IRR is widely used by practitioners and academic studies, but it has several shortcomings, such as reinvesting the dividends at the IRR itself and different measures for aggregated and disaggregated cash f lows from the investments (Da Rin et al. [2011]).
There is evidence that actual PE and VC returns are not as high as those reported by the industry (Da Rin et al. [2011]). Kaplan and Schoar [2005] and Phalippou and Gottschalg [2009] used the VentureXpert dataset and found that PE funds underperform in relation to the S&P 500 Index. In addition, Stucke [2011] and Phalippou [2012] found evidence that a significant percentage of the sample had missing values in cashflows and this could have caused a downward bias in the return measurement.
Harris et al. [2014] found that mean net IRRs are not significantly different among Burgiss, Preqin, and Cambridge Associates for 1990s and 2000s vintages, but VentureXpert underestimated returns, according to other commercial providers. They observed that the Burgiss PME of buyout funds consistently and signifi- cantly exceeds 1.0, with an average of 1.11 and a median of 1.2. That is to say, on average, buyouts outperformed the S&P 500, but PME does not account for liquidity risk. They also found evidence that VC investments of equivalent vintages underperformed the S&P 500. According to the Burgiss data, except for 2005, none of the PME values for venture capital funds with vintage years between 1999 and 2008 exceeded 1.0.
Higson and Stucke [2012] used fund cash f lows information from Cambridge Associates and extended their sample by collecting data from LPs. Their results indicate that funds with vintage years from 1980 to 2008 outperformed the S&P 500 by over 500 basis points per annum, and if they excluded young funds, the excess return increased to 800 basis points per annum. They documented an extreme cyclicality in returns to U.S. buyout funds through time. Similar to Kaplan and Schoar [2005], they also observed a considerable cross- sectional variation in performance. Around 40% of the funds underperformed the S&P 500. As the mean is higher than the median, returns above the S&P 500 are mainly driven by outliers. They also estimated gross returns of the funds; their results imply that gross returns can be 60% to 80% higher than net returns.
Overall, there is evidence of some return persistence in the industry and a concave relationship between fund size and performance (Kaplan and Schoar [2005]).
Selection bias and small samples are more dra- matic problems in emerging markets. EMPEA pub- lishes industry statistics comparing the performance of different emerging markets for 1-, 3-, 5- and 10-year periods. Performance statistics are based on the Cam- bridge Associates index, with pooled end-to-end net returns. In December 2012, the entire sample con- tained 437 PE and VC funds that were formed between 1986 and 2012. (See Exhibit 2 for a distribution of the sample across regions). We observe that many regions are not well represented. It was not possible to calculate an index for the Middle East, because of insufficient sample size; the whole Latin America and Caribbean region has only 47 funds (in contrast, our sample for estimating performance for Brazil only has 46 funds), and the Africa index consists of 43 funds (Missankov et al. [2008] built a sample of 11 private equity funds over a 13-year period with cash f low information only for South Africa).
In Brazil, the main PE and VC data provider has been GVcepe, with the launch of more than one census.
Exhibit 2. Number of PE and VC Funds that Comprise Cambridge Associates LLC Proprietary Index for Emerging Markets PE and VC per Region
Source: EMPEA (2012) Industry Statistics: Fundraising G Investment Analytics. Published 29 April 2013.
The census information is obtained from PE and VC firms with offices in the country.
More recently, ABVCAP (Associação Brasileira de Venture Capital), with the collaboration of KPMG, has built a historical and systematic dataset about the PE and VC industry. The effort is based on information reported by GP members, as required by self-regulation, and contains information about the VC/PE landscape in Brazil since 2011.
Although the census data are the source of many articles about the Brazilian PE and VC industry and provide an important picture of the ecosystem, they do not contain reliable information about fund returns. Information about fund performance provided by GPs, if it exists, may have reporting bias. Siqueira et al. [2011] investigated the determinants of fund performance at the deal level using census information, but in the absence of return information, they use exits through IPO or trade sale as a proxy for success.
Unlike the situation in the U.S., Brazilian PE and VC funds registered at CVM as FIP (Fundo de Investi- mento de Participações) and FMIEE (Fundo Mutuo de Investimentos em Empresas Emergentes) are required to report quarterly financial information. Although CVM is a very valuable source of information, it has been a complex task to extract cash f low information from PE and VC funds and deals. Besides the PE and VC funds, FIPs comprise vehicles of offshore funds investing in Brazilian portfolio companies, real estate funds, and other holdings. Funds are sometimes registered using different names and some may constitute more than one vehicle: FIPs are geared for raising money with Brazilian investors and serve as offshore vehicles for raising money with international LPs. Most of the information at the deal level are provided in explanatory notes, which requires collection by hand and individual judgment.
The Spectra-Insper database is private and has been built through a partnership between Spectra Investments, a Brazilian PE and VC fund of funds, and the Insper Institute of Education and Research, a leading Brazilian Business and Economics School. It contains information about firms, funds, and deals in Latin America.
This article is based on a sample with information on 172 funds of 78 domestic and international orga- nizations, comprising PE (buyout and growth funds), VC, mezzanine, and infrastructure funds, with vintage years from 1990 to 2013 (September). Our sample does not contain the whole universe of firms and funds in Brazil, but it is representative of the industry. ABVCAP and KPMG [2014] mapped 187 firms and 483 funds in January 2013; they included seed and real estate funds that are not in the Spectra-Insper database.
Although the Brazilian PE and VC industry is young, the players are slowly maturing. Exhibits 3, 4, and 5 draw a picture of the Brazilian PE and VC industry maturity. We observe that
Exhibit 3. Number of Funds Raised by Organization
Exhibit 4. Vintage of First Fund Raised by the Organization
Exhibit 5. Number of Funds Raised by Organization vs. Vintage of First Fund Raised by the Organization
The dataset contains information about gross return and MoM at fund level, but it does not con- tain cash f low information. Only five funds report net returns and net MoM. Most funds report information in U.S. dollars and eight of them report IRR and MoM also in Brazilian reais. Therefore, we are limited to analyzing gross IRR and gross MoM of funds in U.S. dol- lars and we cannot estimate PME and compare the PE and VC performance with the public stock market.
We include in our sample only funds that exited or had liquidity events in 60% or more of their investment portfolios. We have excluded one fund that is a clear outlier: with an IRR equal to 502%, and MoM equal to 3.4x. Our final sample contains IRR and MoM of 46 funds, vintage years 1990 to 2008. As the Brazilian industry is very young, we believe it provides a repre- sentative picture of fund performance.
PE accounts for 69.6% of our sample of funds with IRR information (22 growth and 10 buyout), VC accounts for 21.7% (10 VC funds), and infrastructure and mezzanine for the remaining 8.7% (three infrastructure funds and one mezzanine fund).
The mean IRR of the entire sample is 22%, the median is 23.5%, the standard deviation is 31.1%, and 19.6% of the funds have negative IRR. Even if the Spectra-Insper database may be subject to some selection bias, the dataset does not contain a severe survivorship bias, and it seems to represent the Brazilian PE and VC fund universe well, with bad, average, and good per- formers. By comparison, Missankov et al. [2008] build a sample of 11 PE funds over a 13-year period for South Africa, with cash f low information. Their sample con- tains only funds that have been operating for at least three years. The net returns ranged from 14.3% per annum (excluding one IRR of –100%) to 65%, with an arithmetic average of 35.7%. The worst performance in our sample (also excluding one IRR of –100%) is –19%.
Similar to Kaplan and Schoar [2005] and Higson and Stucke [2012], we find significant dispersion in fund performance. The difference between top and bottom quartiles in terms of mean IRR is around 72% and in median IRR is around 60%. Exhibit 6 illustrates the differ- ence in performance.
Exhibit 6. Comparison between Top and Bottom Quartiles
Exhibit 7 contains descriptive analysis between the four quartiles. We treated the highest performance fund, that yielded an IRR of 111%, and the lowest performance fund, that yielded an IRR of -100% as outliers, and we calculated the mean and standard deviation of the top and worst quartiles without these outliers.
Exhibit 7. Descriptive Analysis among the Four Quartiles
The average gross IRR is 22% and the median gross IRR is 23%. On a multiple basis, private equity funds have returned 3.6x its invested capital on average, with a median of 2.5x. (Exhibit 8). This translates to a median holding period of 4.6 years and an average holding period of 6.2 years.
Exhibit 8. Comparison between PE and VC Funds MoM
Funds that operate exclusively in Brazil have a slightly higher IRR in comparison to geographically diversified funds (Exhibit 9), although the difference between groups is not significant. The sample comprises approximately 50%–50% of each group. Buyout funds outperformed growth funds in the period analyzed (Exhibit 10), but the difference in performance is not statistically significant in our sample.
We compare Brazilian PE and VC fund returns with the performance of the industry in the U.S. in the equivalent period. In order to do so, we used the tables in Harris et al. [2014], which report net return of VC and PE organized by vintage year from different providers: VentureXpert, Cambridge Associates, Preqin and Burgiss.
As we have only gross return information, we had to estimate the average percentage of the deduc- tion due to fee payments to GPs. In Brazil, management fees range between 1.5% and 2%, with a fee reduction after the investment period, and typically 20% carried interest on the profits above a hurdle rate. The carried interest is higher for top performer funds, but as there is practically no catch-up for performance fees in Brazil, it should be lower than in the U.S. We conducted informal interviews with GPs and estimated that 30% of fee deductions would be a conservative estimate for the average in Brazil. Therefore, we made a simplifying approach of multiplying gross return by 70% to obtain net returns. We observed that in the five funds where we had information about gross and net returns, net returns corresponded to 79% of gross returns on average, which is in accordance with our conservative estimates of net returns.
Exhibit 9. Difference in Performance between Geographically Diversified Funds and Brazil-Focused Funds
Exhibit 10. Difference in Performance between Buyout and Growth PE Fund
On average, Brazilian PE (Buyout + Growth) funds outperformed U.S. funds with vintage years 1990 to 2008. This is true for the sample of all providers of U.S. PE and VC fund performance, as seen in Exhibit 11. However, Brazilian VC funds outperformed only Bur- giss and Preqin and underperformed VentureXpert and Cambridge Associates.
Exhibit 11. Average Net IRR of PE and VC Funds with Vintage between 1990 and 2008 in Brazil and U.S.
As the sample is small, we aggregated Brazilian PE and VC in the vintage year analysis. Exhibits 12 and 13 compare the average IRR of Brazilian funds by vintage year with the average IRR of U.S. PE from different providers. As we can observe, Brazilian funds underperform the U.S. funds from 1990 until 1997, and after that, with the exception of 2002, they invert the situation, outperforming the U.S. funds. As in Higson and Stucke [2012], we observe that Brazilian PE and VC performance is cyclical. It appears that the cyclicality is higher in Brazil, but this may be due to the small sample, rather than the emerging market effect.
Exhibit 12. Comparison between Brazilian Fund Performance by Vintage with U.S. PE and VC Fund Performance Reported by Different Providers
Exhibit 13. Comparison of PE Return in the U.S. and Brazil by Vintage Year
On average, Brazilian PE and VC funds outper- formed U.S. funds with vintage years between 1990 and 2008. Therefore, international investors were remu- nerated for assuming country risk in Brazil. There are three main explanations for this good performance: 1) the Brazilian economic boom between 2004 and 2012; 2) the limited competition for deals in Brazil at that time; and 3) the fact that Brazilian PE and VC managers are becoming more experienced, thus driving better performance.
When we analyze Brazilian fund IRR by vintage year, we observe that the cyclicality is higher than in the U.S. It is not clear if this is due to the emerging market effect, since the sample is still small.
Brazilian PE and VC funds with vintage years between 1990 and 1997 underperform the U.S. funds on average, and those with vintage year between 1998 and 2008 outperform. One possible explanation for this fact is learning curve. The Brazilian PE industry started to become a significant player after 1990 (Klonowski [2011]), with the Brazilian privatization wave. The institutional environment had experienced significant improvements since 2000: BOVESPA created differen- tiated stock listing segments for companies with better governance practices in 2000, the new bankruptcy law (similar to Chapter 11) was launched in 2004, and the Brazilian Supreme Court (STJ) has recognized conflict resolution by Arbitrage Chambers since 2005. In early 2000, ABVCAP (the Brazilian Association of Ven- ture Capital and Private Equity) was founded and the INOVAR program was launched by FINEP, educating entrepreneurs, managers, and Brazilian investors about private equity and venture capital. Many international funds that operated in Brazil from abroad opened offices in the country around that period of time.
Healthy returns and good prospects have pro- pelled a growing number of PE and VC organizations in Brazil. At the same time, the industry will undergo a natural selection process stimulated by the huge differ- ence between top and bottom performers. In that regard, 11% of the organizations have not raised a new fund in five years and, most probably, will cease to exist soon.
The growing number of PE and VC organizations looking for investment opportunities leads to a higher number of Brazilian companies improving their corpo- rate governance practices, reducing their labor and legal contingent liabilities, and strengthening their profession- alization process so that they can become candidates for IPOs and targets for strategic industry players. Based on CVM data, we estimate that more than 40% of Brazilian IPOs were backed by private equity f irms. Minardi et al. [2013] found evidence that PE-backed IPOs per- form better than non-PE-backed IPOs in Brazil. More candidates for IPOs and strategic acquisitions generate better exit opportunities, encouraging new managers to emerge. A greater number of experienced managers should lead to more LPs looking to invest. A virtuous circle may emerge that will drive the growth of the industry in the years to come.
Da Rin, M., T. Hellmann, and M. Puri. “A Survey of Ven- ture Capital Research.” (2011) Available at http://ssrn.com/ abstract=1942821.
Higson, C., and R. Stucke, “The Performance of Private Equity.” (2012)
Harris, R.S., T. Jerkinson, and S.N. Kaplan. “Private Equity Performance: What Do We Know?” Journal of Finance, forth- coming 2014. Available at http://onlinelibrary.wiley.com/ doi/10.1111/jofi.12154/pdf.
Kaplan, S., and A. Schoar. “Private Equity Performance: Returns, Persistence, and Capital Flows.” Journal of Finance, Vol. 60, No. 4, (2005), pp. 1791-1823.
Klonowski, D. “Private Equity in Emerging Markets: Stacking Up the BRICS.” The Journal of Private Equity, Summer 2011, pp. 24-37.
KPMG and ABVCAP. “Consolidação de Dados da Indústria de Private Equity e Venture Capital no Brasil 2011/2012/2013.” Research, 2014. Available at http://www.abvcap.com.br/ Download/Estudos/2716.pdf.
LAVCA. “2013 LAVCA Industry Data & Analysis: Update on Latin America Private Equity & Venture Capital.”
Latin America Private Equity and Venture Capital Asso- ciation (LAVCA), 2013. Available at lavca.org/wp-content/ uploads/2013/07/LAVCA-2013-Industry-Data-and-Anal- ysis-Sample.pdf.
Minardi, A.M.A.F., G.L. Ferrari, and P.C. Tavares. “Perfor- mance of Brazilian IPOs backed by Private Equity.” Journal of Business Research, Vol. 66, No. 3 (2013), pp. 448-455.
Missankov, I., R. Van Dyk, A. Van Biljon, M. Hayes, and W.Van der Veen. “Private Equity Returns in Emerging Mar- kets: The South African Case.” The Journal of Private Equity, (Fall 2008), pp. 42-52.
Phalippou, L. “A Comment on Recent Evidence on Private Equity Performance.” Working paper, 2012.
Phalippou, L., and O. Gottschalg. “The Performance of Pri- vate Equity Funds.” Review of Financial Studies, Vol. 22, No. 4 (2009), pp. 1747-1776.
Siqueira, E.M.R., A.G. De Carvalho, and H. Gallucci-Netto. “Determinantes do Sucesso dos Investimentos de Private Equity e Venture.” Revista Brasileira de Finanças, 9, 2011, pp. 189-208.
Stucke, R. “Updating History.” Working paper, Oxford Uni- versity, 2011.