A Case for Real Estate Inclusion in Pension Funds Mixed-Asset Portfolios in Tanzania

Traditionally, pension funds invest heavily in fixed income securities. More recently, pension funds have also been increasing their exposure to non-conventional asset classes including real estate. Over the last two decades, pension funds in Tanzania have increased their real estate allocations to more than 18%, which is relatively higher compared to the international practice. This paper looks into investment practice by pension funds in Tanzania with a view to examining whether real estate inclusion contributes to the attainment of optimal portfolios. The study entails mean-variance optimisation analysis of pension funds’ portfolios, covering a period between 2009 and 2018. Findings show that real estate inclusion in pension funds’ portfolios has risk reduction potential. This paper enriches literature on pension funds’ investment practice, particularly those in the countries that are characterised by nascent capital markets institutions. The study also compares conventional literature on pension funds investment practice and the reality on the ground.


Pension Funds Investment Practice
Globally, pension schemes have been in existence for many centuries and they originally targeted people injured in wars or industry, widows and aged people but with time they evolved and started covering poor people by protecting them against suffering caused by loss of income. Today, pension schemes exist as large institutions playing a major role in the social security systems. The schemes are in two main categories, namely defined contribution plans and defined benefits plans. Whereas in defined contribution plans pension benefits are not specified other than that at retirement, in defined benefits plans benefits are specified and would normally depend on years of service and the level of wages or salary of a member (Sharpe et al., 1999). In the defined benefits plan system, the pension plan sponsor accepts liability for future pension payments.
The primary objective of pension funds is to accumulate sufficient assets, through contributions and investment income to satisfy pension obligations on a timely basis (Sharpe et al., 1999). Administrators of pension funds have a fiduciary responsibility to the insured persons, which forces them to be prudent in carrying out their investment duties (Tamagno, 2000). Pension funds pool resources from beneficiaries and employers and invest them in various investment media. Most of the pension funds' aggregate assets are invested for a long term (Ryck, 1998). However, the liability structure is one of the main factors considered when deciding on the investment strategy of a pension plan (Mensonides, 1998). Whereas pension funds with young members are more likely to invest in long-term investments because they have few short-term obligations, mature pension funds need to carefully monitor their solvency conditions and liquidity of their investments to be able to honour their obligations.
Traditionally, pension funds around the world have been investing largely in fixed income assets, traditional equities and cash. However, this conventional asset allocation practice focusing on low-yield fixed income assets presents potential difficulties in meeting future obligations (APREA, 2010). As a result, over the recent years, pension funds have been extending their investment horizon by increasing their exposure to alternative assets, mainly real estate, private equity, hedge funds, infrastructure and commodity (Andonov et al., 2014;Aubry et al., 2017). Aging population is one of the factors that are increasingly putting pension funds, especially those in developed countries, under pressure to achieve portfolio diversification in order to meet their significantly increasing liabilities in an effective risk-adjusted manner (Newell, 2010).

Pension Funds Exposure to Real Estate
Viewed from a neo-institutional theory perspective, pension fund institutions are highly susceptible to transactions costs. Transaction costs arise due to the cost of time and effort invested in researching, creating, implementing, administering, M. M. Kusiluka, S. M. Kongela DOI: 10.4236/cus.2020.83024 431 Current Urban Studies monitoring and enforcing policies (Shahab & Viallon, 2019). They include the costs of time spent on each transactional activity and the direct monetary expenses incurred in the policy design, implementation or participation processes (Shahab et al., 2018). Pursuit of real estate investment exacerbates the situation because the process entails many transaction costs generating activities. This is mainly due to the lengthy investment acquisition and management processes involved. Despite the inherent transaction costs, pension funds still find real estate to be a worthwhile investment mainly due to its income and capital stability, steady capital growth and low volatility of returns (Newell, 2010;Kusiluka, 2012).
Although real estate is less favourable compared to the traditional pension funds' investment vehicles, its increasing importance as an asset class among institutional investors has attracted pension funds, among other institutional investors, to include it in their portfolios. For instance, in the USA, since the 1950s allocation to real estate has been kept in the range of 0% to 17% (Worzala & Bajtelsmit, 1993). The situation has been more or less similar in the UK over the same period of time, albeit prior to 1980s some isolated cases of UK pension funds had allocated up to over 20% to real estate (IPF, 1993). Some recent studies show that real estate is the most important alternative asset commanding portfolio allocation of above 5% (Andonov et al., 2013;Andonov et al., 2014;Van Nieuwerburgh et al., 2015).  (OECD, 2017). However, in the wake of advancement in financial engineering, real estate is increasingly becoming a liquid investment, which should allay illiquidity fears. This is however more practical in developed countries, most of which have well established and functioning financial markets. Pension funds, among other institutional investors, in those countries are increasingly investing in indirect real estate assets, both listed and unlisted (Andonov et al., 2013;Van Nieuwerburgh et al., 2015). In some countries such as Australia, Belgium, Canada and Denmark pension funds are free to allocate up to 100% to real estate assets (OECD, 2018).
Some studies, based on asset return mean-variance optimization, suggest a range of between 15% and 25% to be the optimal allocation to real estate assets in mixed-asset portfolios (Hoesli & Hamelink, 2004). However, most of the pension fund managers, apart from looking at asset returns and variances, also consider changes in pension funds liabilities and their covariance with asset returns (Craft, 2005). Many other investors pursue their own asset allocation poli-  (Hoesli et al., 2002).

Very limited literature exists on real estate allocation in African pension funds'
and other institutional investors' mixed-assets portfolios. Kwaku (2007) observes that portfolios of institutional investors in Africa are composed of real estate (15.5%), government securities (25%), stock market securities (47%), private equity (0.4%) and others (12.1%). Newell et al. (2002) point out that insurance companies and pension funds in South Africa allocate an average of 8% of their portfolio to real estate. Over the last two decades, pension funds in Tanzania have on average been allocating between 20% and 40% to real estate (Kusiluka, 2012;Kongela, 2013).

Real Estate Sector in Tanzania
When Tanzania achieved its independence in 1961, the level of urbanization was only 4.8% compared to 31% in 2018 (Kusiluka et al., 2017). Soon after independence, some initiatives were taken to deal with urbanization related challenges. The focus was mainly on addressing land rights and housing problems in urban areas because interventions made during colonial period were inadequate, short-lived and covered a very small section of the population. One of the important steps taken by the government was to review land and real estate ownership related policies and legislation in order to promote equitable access to land.
In 1967, the government adopted socialist policies, popularly known as ujamaa. In the course of implementing ujamaa which was promulgated through Arusha Declaration of 1967, some pieces of legislation aimed at strengthening government role in direct participation in the real estate market were enacted.
For instance, in 1971, the Acquisition of Buildings Act was enacted. The legislation, among others, provided for the nationalisation of private rental properties whose market values were over TZS 100,000 (equivalent of £6,000 then) or whose rental values were over TZS 833.3 per month (Meredith, 2006). Public servants were also prohibited from owning rental buildings.
In 1962, the National Housing Act No. 45 was enacted. The Act provided for the establishment of National Housing Corporation (NHC) whose main functions were to lend, guarantee or provide finance to local authorities and individuals for the construction and improvement of buildings and approved housing schemes (Kironde et al., 2003). Another notable government initiative was the establishment of Tanzania Housing Bank (THB) in 1973. The main objectives of THB were to mobilize savings and resources for housing development, promote housing development, and provide technical and financial assistance for owner-occupied housing. It is estimated that about 36,000 housing units were built using THB loans (Kusiluka, 2012). Due to persistent operational problems, THB collapsed in 1995. While THB collapsed, NHC to date remains to be the largest real estate owner and a real estate 'market mover' in Tanzania. with an estimated population of more than 5 million.
Prior to the reforms, pension funds in Tanzania were largely investing in fixed-income securities, largely treasury financial instruments and cash. In the wake of liberalization, pension funds began spreading their investments across a broader spectrum of investments. In fact, the enactment of the Public Corporations Act No. 2 of 1992 paved way for pension funds, among other parastatal organizations, which formerly used to invest heavily in treasury securities, to introduce non-traditional investments in their portfolios (Kusiluka, 2012). The legislation, among other things, liberalized investment policies of state-owned enterprises (SOEs).
Over the last two decades pension funds in Tanzania have made noticeable exposure to real estate. With their direct real estate investment value estimated at about TZS 1.8 trillion in 2018, pension funds have become the second largest real estate investors in Tanzania, after NHC whose real estate investment portfolio was valued at TZS 4.1 trillion in the same year. NHC has maintained its leading role in the provision of housing and non-residential real estate in urban areas. Figure 1 shows the growth trend of real estate investment held by pension funds and NHC over the recent years.
As shown in Figure 1, over a period of 10 years, both pension funds and NHC recorded significant growth in real estate investment value. Whereas NHC real estate portfolio value average annual growth over the period was 17.2% that of pension funds was 23.9%. The growth trend is mainly attributed to a significant addition of a stock of buildings, real estate revaluation and upward rental reviews coupled with a booming market during the period. In 2017, NHC total  portfolio was valued at TZS 4.4 trillion, which is the highest value ever recorded.
Pension funds recorded their peak value of TZS 1.8 trillion in 2018.

Pension Funds Investment Practice in Tanzania
All major government sponsored pension funds in Tanzania are contributory pay-as-you-go defined benefit schemes (John et al., 2017 As shown in Figure

Methodology
A combined portfolio of state-sponsored pension funds in Tanzania  to amplify the error between the model and reality hence asset grouping was opted for. The following portfolio return (E(r p )) and portfolio risk σ p formulae were applied: where, E(r p ) is portfolio returns for the period, w Re is portfolio weighting for real estate, w Fi is portfolio weighting for fixed income assets, w Eq is portfolio weighting for equities, E(r Re ) is real estate returns, E(r Fi ) is fixed income assets returns, E(r Eq ) is equities returns, σ p is portfolio standard deviation (risk) for the period, σ Re is standard deviation of real estate returns, σ Fi is standard deviation of fixed income assets returns, σ Eq is standard deviation of equities returns, Cov(Re, Fi) is covariance between real estate and fixed income assets returns, Cov(Re, Eq) is covariance between real estate and equities returns, Cov(Fi, Eq) is covariance between fixed income assets and equities returns.
Data collection also involved interviews with 14 respondents of whom ten were investment officials of the five pension funds, two NHC managers and one Capital Markets and Securities Authorities (CMSA) official. Interviews were useful in explaining pension funds' investment decisions and trends. Other official reports, publications and websites formed a source of some secondary data.

Entry of Pension Funds in the Real Estate Market
It was noted from the interviews with investment managers of the five pension funds that one of the main challenges that pension funds in Tanzania   The entry of pension funds in the real estate investment sector in Tanzania has significantly changed the sector. They started investing in prime commercial properties but over time some of them, particularly NSSF and PPF, extended their investment to residential properties. NSSF and PPF have also been engaged in the construction and financing the construction of a wide range of affordable housing including halls of residence for higher learning institutions and staff housing for various government institutions. Figure 3 shows the trend of real estate allocations by the five pension funds over some recent years.
As shown in Figure 3, over the last decade, pension funds' direct real estate allocation has generally been increasing, averaging above 20%. The average allocation was well in line with the statutory ceiling of 30%. Over the entire period, NSSF had the highest allocation averaging at 28.8% while at the lower extreme  was GEPF whose real estate allocation was generally declining over time, recording the lowest allocation of 5.2% in 2017. It is also evident from Figure 3 that real estate allocations by pension funds in Tanzania are much higher than of those in their counterparts in developing countries. The relative higher real estate allocations are apparently reflective of the young age structure of the population of Tanzania, which imply a long-term liability structure of pension funds. Another common explanation for the high real estate allocation was noted to be existence of limited investment options suitable for pension funds, considering the intensity of regulation exercised on pension funds operations.

The Effect of Real Estate in a Mixed-Asset Portfolio
Until 2015, pension funds had much more flexible asset allocation policies. For instance, for some years NSSF investment policy provided for 10% allocation to real estate assets. PPF investment policy, on the other hand, provided for a much higher and flexible range of 32% -48% for real estate allocation (Kongela, 2005). Even NSSF investment policy was not watertight; it gave managers some freedom to take on any new investment vehicle provided there was a strong business justification (Kusiluka, 2012). Observations, however, show that some pension funds had at times had real estate allocations higher than the proportions spelt out in their respective investment policies. However, the situation changed when the Social Security Schemes Investment Guidelines of 2015 took effect. planning to allocate more than 30% to real estate assets must obtain permission from the central bank. The guidelines also require that returns on investments other than treasury investments be above returns on treasury bills and treasury bonds. The 30% real estate allocation ceiling was also noted to be mindful of the liquidity risk associated with direct real estate investment which forms the majority of pension funds' real estate investment portfolio. Table 1 shows asset weightings and returns for the five pension funds combined.
It is clear from Table 1 that the pension funds have been heavily investing in fixed income assets. Fixed income assets, accounting for 67.9% portfolio allocation, largely include government securities, certificates of deposits, commercial papers, corporate bonds and term loans. Historically, pension funds in Tanzania have been very important participants in the treasury securities market. However, over the period covered in this analysis, the trend of investment in this asset group was slightly declining in favour of real estate and equities. Average allocation to real estate for the period was 18.4% and to equities was 13.7%. It is clear from the analysis that real estate allocation is well below the 30% ceiling stipulated in the Social Security Schemes Investment Guidelines of 2015 and it is within the range recommended in some literature (e.g. Hoesli & Hamelink, 2004).
During the period, annual allocations to real estate and equities were generally increasing by 6.7% and 3.9% respectively. Some of the main reasons for the trend include liberalisation of public corporations' investment policies in 1992  Figure 4 shows the trend of returns on the three investment categories over a span of ten years.
As depicted in Figure 4, during the period, fixed income assets returns have relatively been high except for 2011 and 2016 when equities outperformed fixed income assets. Equities ranked second but it had the highest volatility of returns.
On the other hand, real estate investment had the lowest but the most stable returns. Therefore, measured by the riskiness, equities with a standard deviation of returns of 3.6% is at the bottom, followed by fixed income assets which has a standard deviation of returns of 2.0%. Real estate, with a standard deviation of returns of 1.2%, had the most stable returns. This is consistent with Newell (2010) and Kusiluka (2012) who underscore stability of real estate income. However, further analysis of returns behaviour reveals portfolio diversification potential. Table 2 presents the correlation matrix of asset returns for the three asset classes.  Several implications can be drawn from correlation analysis results presented in Table 2. For instance, it is evident that if included in mixed-asset portfolios, real estate has the potential of reducing risk. However, it is also possible that real estate could substantially reduce portfolio return. To avoid this, it would be advisable to have a portfolio with assets whose returns have low correlation coefficients. From Table 2 it is clear that real estate can form efficient portfolios with both fixed income assets and equities. This is because the correlation coefficients between real estate returns and fixed income assets and equities returns are negative i.e. −0.60 and −0.18 respectively. Similarly, the fact that the returns of fixed income assets and equities are not perfectly correlated suggests existence of risk reduction potential when the two assets are combined. Table 3 summarises the various feasible portfolios for pension funds, based on historical returns and asset weighting.
As presented in Table 3, the results of the analysis of the current portfolio consisting of 67.9% fixed income assets, 18.4% real estate and 13.7% equities and others show that the portfolio return is 9.37% and portfolio risk is 1.29%. Real estate has the potential of reducing portfolio risk but it also reduces returns. For instance, by allocating 60% of the portfolio to real estate, portfolio risk is reduced from 1.3% to 0.6% but portfolio returns is also reduced from 9.4% to 6.2%. However, the proportions of assets in this portfolio do not represent an efficient diversification of the portfolio. Two better options exist. It is possible to reduce portfolio risk without affecting returns and it is also possible to increase portfolio returns without necessarily increasing portfolio risk.
When the portfolio allocation is adjusted to comprise 70.0% fixed income assets, 21.0% real estate and 9.0% equities its return is not affected. With these asset proportions, portfolio risk is reduced from 1.29% to 1.27%, which is a 1.55% reduction in the portfolio risk. On the other hand, when the portfolio is adjusted to comprise 35.5% fixed income assets, 36.6% real estate and 27.9% equities its

Other Emerging Investment Fronts for Pension Funds
Beside their interest in real estate, it was noted that pension funds over the last Local Government Training Institute (Kusiluka, 2012). Through a special purpose vehicle, namely Pension Properties Ltd, pension funds financed the construction of the parliament building in Dodoma. In some rare occasions, some pension funds also issued term loans to private individuals.
The practice being experienced in Tanzania is not an exception. A trend is emerging for pension funds around the world to take on non-traditional investments. In recent years, investing in infrastructure as an alternative to the mainstream investment vehicles has become increasingly popular with pension funds as they strive to manage risk (Andonov et al., 2018;Inderst, 2010;Newell, 2010).

Limitations
It should be noted that the real estate allocations considered in this analysis are only those of direct real estate investment. More recently, pension funds in Tan However, for financial reporting purposes, such indirect investments fall under equities. As a result of this, real estate allocations and income for the purpose of portfolio mean-variance analysis were understated by the amount of funds invested in such indirect real estate investment vehicles and the corresponding income that accrued from the respective investments. It is thus clear that the actual real estate allocation by pension funds in Tanzania is slightly higher than what is normally reported under the asset class.

Conclusion
This paper demonstrates that it is worthwhile including real estate in mixed-asset portfolio due to its potential to reduce risk. Despite the risk reduction potential inherent in real estate assets, pension funds in Tanzania have adopted a random diversification strategy of their portfolio. Their combined portfolio does not reflect efficient diversification as there is still potential for achieving better returns or lowering risk further. Their portfolio consisting of 18.4% real estate, 67.9% fixed income assets, and 13.7% equities had a return and risk of 9.37% and 1.29% respectively. Pension funds could have minimised risk, without affecting returns, by adjusting their portfolios to comprise 21.0% real estate, 70.0% fixed income assets and 9.0% equities. This combination would have reduced risk by 1.6%. Alternatively, pension funds could have maximised returns, without increasing risk, by allocating 20.0% to real estate, 71.0% to fixed income assets and 9.0% to equities. The proposed portfolio adjustments would have increased returns by 5.1%. Both options imply that pension funds should increase their direct real estate investment. This is feasible because their current real estate allocation is still below the statutory ceiling of 30%. Inclusion of real estate in the pension funds' portfolios is further augmented by the scarcity of credible investment vehicles that are capable of preserving capital and guaranteeing a stable stream of income and real estate potential in reducing portfolio risk. However, in order to reduce transaction costs inherent in real estate and to make it a much better and safer investment for institutional investors, policy interventions are required to ensure that transparency in the real estate sector is enhanced. This can be achieved through promotion of capital market-based real estate investments, establishment of real estate information databank institutions and strengthening professional practice standards and ethics in the fields of real estate. The findings of this study are not only relevant to institutional investors in Tanzania but are also useful to similar institutions and policy makers in other developing countries facing similar constraints.