The limited literature on firms deciding to convert into the REIT structure seems most relevant for this investigation. Mendell, Mishra and Sydor (2008) and Piao, Mei and Zhang (2017) used event study methods to investigate the conversion of four forest industry firms into REITS and found that the conversion adds value to shareholders, which they ascribe to tax and liquidity benefits. Ling et al. (2023) investigated the reasons why some firms choose ‘REITing’ and others ‘De-REITing’, arguing that tax and dividend implications are driving the choice. The literature review will progress by presenting three main theories for understanding capital structure choice, starting with trade-off theory that incorporates tax effects, shown previously as important for listed property.

In 1958, Modigliani and Miller introduced the irrelevance theorem, suggesting that a firm’s value is independent of its leverage ratio provided considerations such as taxes, bankruptcy and transaction costs are ignored. This theorem was subsequently updated by the introduction of the tax-shield benefit concept resulting from the reduction of tax liability because of tax-deductible interest on debt (Modigliani & Miller 1963). Thus, as a firm’s leverage increases, its market value should increase by the present value of the debt tax-shield benefit – motivating management to maximise debt within the capital structure. However, Solomon (1963) pointed out that extreme debt levels increase the cost of capital as capital markets demand higher rates of return, and Baxter (1967) suggested that excessive debt increases the probability of default and hence bankruptcy. The tax-saving benefit of more debt has a cost.

These arguments led to the trade-off theory of capital structure (Kraus & Litzenberger 1973), which suggests that capital structures trade-off reduced taxes because of the tax-shield benefit of interest and financial distress costs. Thus, firms with higher marginal tax rates benefit more and may be inclined to increase their leverage (Miglo 2010). Mackie-Mason (1990) found that firms with higher marginal tax rates are indeed more leveraged, and Barclay, Heitzman and Smith (2013) found that taxable US real estate firms use more debt than their non-taxable counterparts (Calvoso 2020). This finding lends credence to the idea of a lack of a tax-shield effect for non-taxable real estate firms. Further, because of financial distress costs, less risky firms can leverage more than riskier ones (Myers 1984).

Pecking order theory (Myers & Maljuf 1984) posits a cost-based preference order for financing sources, resulting from the pricing of asymmetry of information between insiders and financiers. Equity is considered to have the highest information asymmetry cost, followed by debt, and then retained earnings with no cost. Because of regulatory capping of debt levels and mandatory profit payouts in most REIT markets, the pecking order theory is only partly applicable for listed property. Thus, Ott, Riddiough and Yi (2005) find that US REITs placed little reliance on retained earnings, and Feng, Ghosh and Sirmans (2007), considering US REITs between 1991 and 2003, concluded a debt preference whenever the cost of discounted equity exceeds financial distress costs.

Lastly, market timing theory claims that firms attempt to time debt or equity issuances with favourable market conditions. The dynamic asymmetric information version of market timing (Myers & Maljuf 1984) assumes that firms can create their own market timing by publicly releasing positive firm information before equity issuance. Further, market participant irrationality may cause temporary stock mispricing, which managers of firms can use to issue (repurchase) equity when shares are over(under)priced, regardless of actual financing need (Baker & Wurgler 2002). In the United States, market timing theory is supported by the REIT studies of Ooi, Ong and Li (2010) and Boudry, Kallberg and Liu (2010).

Giacomini, Ling and Naranjo et al (2017) present the following stylised facts regarding REIT leverage. REITs are highly levered, and leverage is persistent. They then interpret their results as showing evidence consistent with trade-off theory. It is unlikely that regulations, or trade-off theory, will fully explain the capital structure choices of listed property firms in South Africa that chose to convert to the REIT legal form because the South African regulations are less restrictive than those in other jurisdictions (see Table 1-A1 in Appendix 1) leaving more room for flexibility. We will, therefore, investigate the determinants of leverage for these South African property firms, without restricting the choice of determinants. Table 2 summarises the expected relationship between various determinants and leverage in terms of the three capital structure theories presented.

The literature and arguments presented up to this point create the expectation that the 2013 regulatory change in South Africa will probably lead to a move away from trade-off theory-based determinants of leverage. However, this cannot be assumed as the listed property literature reviewed above shows listed property sensitive to determinants across all three main types of capital structure theories. The impact of the change will have to be empirically determined.

Our sample comprises the 39 South African listed property companies (PLSs, PUTs and REITs) listed on the JSE for at least 2 years between 2005 and 2019. This represents an unbalanced panel of 314 firm years, consisting of 137 REIT observations post-2013, and 53 PUT and 124 PLS observations, nearly all up to 2013. Dual-listed firms and firms domiciled outside South Africa were excluded. To avoid survivorship bias, delisted, renamed, taken over or merged firms were included in the sample. Data were obtained from Bloomberg or annual financial statements. We rely on panel regressions for our models, similar to Giacomini et al. (2017) and Ling et al. (2023). Panel regressions are particularly relevant for REITs where leverage is ‘persistent’ (Ling et al. 2023). The demeaning process removes the persistent component of leverage. Because of the small sample size, and the fact that the sample is unbalanced, we could not use the Generalised Method of Moments (GMM) approach.

South Africa’s regulatory REIT debt limit is based on the book value of total assets, and hence the book value ratio of total interest-bearing debt-to-total assets was used as our leverage proxy. As managers generally cannot influence the constantly changing market values of firms, book leverage is more appropriate to research actual capital structure decision making, in line with Versmissen and Zietz (2017) and Breuer et al. (2018). Internal or intercompany financing arrangements were excluded from total interest-bearing debt as they do not represent true third-party financing (and hence external financing risk) and often are merely structured for tax purposes and do not reflect true capital structure decisions. The linked debenture financing instruments characteristic of PLSs were treated as equity, given their quasi-equity nature. The independent variables used are shown in Table 3.

F-tests for poolability indicated that two-way fixed effects were valid across all models, but both ‘between’ (pooled regression) and ‘within’ (fixed effects regression) models were run. In the pooled regression, cross-sectional effects would dominate as the data panel is wider than deep over time, thus called ‘between’ as the difference between firms is emphasised. In the fixed effects regression, time-series effects would dominate as the cross-sectional effects are absorbed, thus called ‘within’. Note that the fixed effects model incorporates time effects, which displaced the impact of macro-economic variables such as inflation and interest rates (inflation and lagged interest rates were statistically significant in models excluding time fixed effects). These variables, as well as the fixed sectoral diversification dummy, were therefore not included in the fixed effects model. As robustness check for possible endogenous variables, the fixed effects model was regressed with lagged growth, lagged share price volatility and lagged share performance. Model specifications are as follows:

Pooled regression model:

Fixed time effects model:

Lagged fixed time effects model:

All variables are as defined in Table 3. µ and ג are the intercept terms of each firm and time-specific effects, respectively, for the fixed effects regressions; α is the common intercept, and ε is the error term.

Both variance inflation factor (VIF) statistics and correlations (see Table 4 with no correlation coefficients larger than 0.6) indicated the probable absence of multicollinearity. However, it is to be expected that inflation and short-term policy interest rates would be collinear under an inflation targeting monetary policy regime. This is not the case here because our interest rate variable is a long-term market determined rate. Heteroscedasticity was addressed by using White two-way clustered standard errors and covariances across all models.

Table 4 presents descriptive statistics. The mean leverage ratio for South African listed property over the sample period was 28.2% (PLSs: 29.8%, PUTs: 16.8% and REITs: 31.2%). It is surprising that the mean for REITSs is higher than the mean for PLSs as REIT leverage is limited by regulation while that for PLSs is not. Most observations lie between 20% and 40% – well below the 60% REIT regulatory limit. This may be because firms wish to maintain buffers to the limits in case of external shocks, market downturns or liquidity issues (Calvoso 2020).

Figure 1 depicts average leverage ratios over the sample period. Understandably, because of debt restrictions, PUTs exhibited lower leverage than PLSs prior to the REIT regime. After REIT introduction in South Africa in 2013, PUTs that converted to REITs took advantage of the additional debt allowance by increasing leverage ratios, as did PLSs, but to a lesser degree (Calvoso 2020). It seems that a long-term market-equilibrium leverage ratio is around 30-35% debt to gross assets with clear convergence after 2013. Interestingly, Wesson and Carstens (2019) find that the introduction of REITs in South Africa seems to have decreased leverage on average, although their sample ends in 2016 and hence only includes 3 years of post-REIT data.

Table 5 shows correlations between the variables with probabilities less than 5% in bold. Firm size, share price volatility and PLS structure show statistically significant positive correlations with leverage. Profitability, share performance, interest coverage ratio, PUT structure, sectoral diversification and inflation show statistically significant negative correlations with leverage. The coefficients for size and share price performance support trade-off theory, while the coefficients on the interest coverage ratio and inflation can be seen as opposing trade-off theory. The coefficients on profitability and share price volatility support pecking order theory, while the coefficient on share performance supports market timing theory. Thus, univariate correlations seem to lean towards pecking order and market timing theory but require confirmation in a multivariate setting.

Table 5 presents univariate correlation coefficients at the top of each block with the statistical significance of the coefficient indicated at the bottom of each block in italics. Statistical significance is bold when it is below 0.05, indicating that the correlation coefficient is likely different from zero. The data used comprise an unbalanced panel from 39 listed property firms in South Africa for the years 2005 to 2019.

Table 6 shows the regression results. In the ‘between’ firm model, where cross-sectional effects dominate, statistically significant positive correlations are found between leverage and tangibility of assets, log of size and share price volatility, and statistically significant negative correlations between leverage and interest coverage ratio, PUT structure, diversified property portfolios and inflation. All signs correspond to the correlation table, and interpretations are therefore the same, except to note that the ‘between’ model will emphasise differences between firms rather than changes over time.

Log of firm size, share price volatility, interest coverage ratio and the PUT legal structure retain statistical significance and the same sign across all three models. Further, the PLS structure variable is highly statistically significant in both the ‘within’ firm models, implying that PLS structure firms have higher leverage, explained by the unlimited debt capacity available to PLSs compared to the 30% regulatory cap on PUTS and 60% on REITs, resulting in PLSs utilising greater leverage compared to the other structures, and PUTs less. This is consistent with Dogan et al.’s (2019) multi-country findings that regulatory restrictions affect listed property leverage ratios.

Although the firm size finding is consistent with trade-off theory, the highly statistically significant negative coefficients for inflation and the interest coverage ratio provide strong evidence against this theory in explaining the capital structures. The information asymmetry of a larger firm should be less because of a greater analyst following. Thus, the positive coefficient can also support pecking order theory. The significant positive coefficient on growth, in the fixed effects model without lags, supports trade-off theory. The strongly significant positive coefficient on share price volatility across all models is in line with pecking order theory and with prior work on REITs (Versmissen and Zietz 2017). Further, although the negative sign on share price performance is consistent with timing theory, it is not statistically significant. Lastly, the validity of time fixed effects and its displacement of macro-economic variables such as inflation, interest rates and GDP is indicative of market timing theory.

The strongest evidence indicating the validity of trade-off theory explanations for PLSs, and not for PUTs or REITs, is found when interacting the indicator variable for the PLS legal structure with the other independent variables. We found a statistically significant positive interaction between the PLS firm type and profitability (coefficient of 0.429 and statistical significance of 0.023), which leaves the total correlation between leverage and profitability positive, as would be expected under trade-off theory. We also found a statistically significant positive interaction between the PLS firm type and the interest coverage ratio (coefficient of 0.015 and statistical significance of 0.054), again indicating that trade-off theory predictions are more valid during the PLS regime.

In combination, the above evidence supports both market timing theory and pecking order theory, with evidence against trade-off theory during the REIT regime. Unsurprisingly, market timing theory and pecking order theory overlap as both rely on information asymmetry. The findings, therefore, align with our expectations. It is likely that, for the regulated listed property industry, the lack of tax-shield benefits resulting from REITs’ non-taxable status and, less so, the regulatory leverage constraints eliminate trade-off as a major capital structure consideration. This supports Barclay et al’s (2013) finding that taxable US listed property companies are more leveraged than tax exempt ones.

As per Figure 1, after the introduction of REIT structures in South Africa, converted PUTs converged to the mean PLS leverage range of around 30% to 35%. This supports previous findings that regulatory restrictions affect capital structure. It is also interesting that the mean PUT historical leverage was for most of the pre-REIT period well below the regulated maximum of 30%, while PLS similarly maintained low ratios relative to their theoretically uncapped maxima. Thus, the 60% maximum leverage requirement for REITs seems to be higher than the natural leverage level in the industry and, therefore, is unlikely to play an important role in the leverage decisions of South African listed property companies.

In conclusion, listed property capital structure in South Africa is influenced by regulation but not determined by it. It seems as if leverage levels can find their natural market level within the regulatory limits set for REITs (30%–35%). We expected trade-off theory to be of little importance during the REIT regime because the main benefit of the tax deductibility of interest is of little importance. We also expected the pecking order theory to be of reduced importance because fewer profits are retained by REITs. It was difficult to distinguish between pecking order results and market timing results because of both theories relying on information asymmetry as foundation and tests that overlap. However, we argue that leverage in South African listed property companies is mostly determined by market timing considerations for the following reasons: (1) firms do not appear to use maximum leverage levels, and so retain some optionality for dealing with economic fluctuations, (2) the relationship between inflation and leverage is negative when considering that rental increases are mostly fixed (thus, an increase in inflation is related to a decrease in leverage, all else equal, which can be understood as responding to an expected economic downturn), (3) time fixed effects are valid predictors of leverage, and (4) leverage is consistently sensitive to the volatility of listed property equity.

South Africa’s listed property industry is large and well-developed by developing world standards, with some unusual features that could affect listed property capital structure decisions, such as a fairly recent regulatory framework change and underlying rental agreements with fixed escalations. This makes South Africa an interesting emerging market context in which to test the empirical evidence against capital structure theories as applied to the listed property industry.

Our study applied ordinary least squares and panel regressions to an unbalanced panel of all property companies listed on the JSE between 2005 and 2019. Because South African listed property after 2013 is subject to regulatory leverage restrictions and the tax deductibility of interest is no longer important, we expected the trade-off theory to be least important in listed property capital structure decisions. This is supported by our findings. In contrast, as expected, evidence supporting both the market timing and pecking order theories are found. These results are in line with the current REIT regulatory requirements in South Africa, which, compared to most other larger REIT markets, are less restrictive regarding earnings payout requirements, thus leaving more scope for the pecking order theory, but more stringent in terms of leverage limits, thus (together with REITs’ tax-free status) reducing the relevance of the trade-off theory.

We conclude that in South Africa, as an example of an emerging market, capital structure decisions of listed property companies seem mainly based on information asymmetry considerations, which is the common foundation of both the market timing and the pecking order theories. Further, we indirectly observe some evidence of the impact of regulatory restrictions on capital structure decisions, but particularly the leverage restriction of 60% of gross assets does not seem to be a constraint on South African listed property companies, whose leverage ratios are largely well below the maximum allowed and seems to be determined by non-regulatory considerations.

The reasons for capital structure decisions and their interaction with regulatory constraints remain of great interest to researchers and regulators. Our study sheds additional light on these questions, specifically within the under-researched context of listed property, which has peculiarities that impose additional constraints on this decision, as well as emerging markets, which similarly have received limited research attention.