The concept of cost of capital is one source of difference between accounting and economic profits. For instance, accounting profit only recognises cost of debt as interest charges without recourse to the associated implicit costs. Generally, an investor’s cost of capital is the minimum required rate of return on his investment (Vernimmen, Quiry, Dallocchio, LeFur, & Salvi, 2014; Pouraghajan et al, 2012). Put succinctly, an investee’s cost of capital is an investor’s return. It therefore implies that investees would seek to minimise their cost of capital by carefully selecting the most efficient mix of capital structure (referring to an optimal combination of debt and equity), whereas investors would carefully choose a portfolio of asset classes (bond, public equity, private debt, private equity) that minimises their regrets or risks while maximising their returns, always ensuring that their risks are adequately priced as reasonable return. As Dagogo and Imegi (2017) described it, every mention of capital structure has a corresponding mention of its implications for overall cost of capital or the weighted average cost of capital (WACC). Albeit, there are competing views about the reality of creating value only by altering the ratio of debt-to-equity in a manner that reduces WACC. Barring any dissenting views and letting the traditional view to prevail, an investee’s ability to create incremental business value is enhanced by reducing WACC (Casseli & Negri, 2018). Capitalisation of private firm of the mid-market range may attract mezzanine, venture capital, securitisation, bank lending, leveraged buyout, management buyout (or buy-in), and other private debt or equity investments. Markets for these alternative assets are less efficient than markets for public equity or bond investments because there are fewer organised markets for investors, and therefore deals are usually on an ‘ad hoc’ basis, and fundraising is transacted on a deal-by-deal basis. (Dagogo & Imegi, 2017; Slee 2011) Eventually, the financial costs attracted by mid-market firms reflect the perceived riskiness of their investment projects measured by the project’s propensity to generate expected cash flow, and perceived financial risk measured by the degree of financial leverage or their sensitivity to credit and liquidity risks. Therefore, to determine the private cost of capital, we consider a risk-free rate plus a risk premium, which does not only cover operating and financial risks but also includes firms reflect the perceived riskiness of their investment projects propensity to generate expected cash flow, and perceived financial ris The Internationalfinancial Journal of Banking and Finance, Vol. 16,sensitivity Number 1 (January) 2021, pp:and 1–20 liquidity risks. leverage or their to credit private cost of capital, we consider a risk-free rate plus a risk premium and financial also includes a premium for inefficien a premium operating for inefficient market, risks asset but intangibility, asymmetric andrisks the are likes. In other words, risks are pr information,asymmetric and the likes.information, In other words, priced by capital type, capital access point, so so thatthat the the models of cost point,and andinvestee investeetype, type, models of cost of capital of capital based on public equities and bonds such as the capital asset bonds such as the capital asset pricing model (CAPM) and dividend pricing model (CAPM) and dividend growth model are unsuitable for for determining the cost of capital in mid-market firms. (Slee, 2011). determining the cost of capital in mid-market firms. (Slee, 2011). For of debtcost is generally given as: given as: example, after-tax of debt is generally (𝑟𝑟𝑟𝑟 + 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠)𝑥𝑥(1 − 𝑡𝑡𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟)

Where rf equals risk-free rate. This measure helpful in understanding Where rf equals risk-free rate.is This measure is helpful in understandi the overall rate paid by a firm for a given type of debt capital. The firm for a given type of debt capital. The measure can also give invest measure can also give investors an idea of the firm’s risk level compared to others because riskier firms have higher cost of generally hav level compared to generally others because riskier firms debt (Van Binsbergen, Graham&&Yang, Yang, 2010). 2010). Graham

Meanwhile, cost of equity represents a firm’s hurdle rate, marginal cost a firm’s rate of efficiency of Meanwhile, capital invested or of theequity internalrepresents rate of return. It hurdle rate, is often used as the invested capital budgeting threshold or the internal ratefor of required return. Itreturn. is often used as the c A firm’s cost of equity represents the compensation the market return. A firm’s cost of equity represents the compensation demands in exchange for owning the asset and bearing the risk of owning the asset and bearing the risk of ownership. CAPM ownership. CAPM seems to be more suitable for a wider spectrum of spectrumpaying of firms (including non-dividend firms (including non-dividend firms) than the dividend growthpaying firms) t thebehind theoryCAPM behind is moreascomplex, as it is ba model. That said, thesaid, theory is CAPM more complex, it is based on an equity stock’s volatility and the level of individual stock’s the level of individual stock’s risk exposure relative to the m risk exposure relative to the market risk. It is given as:

Where Ke equals costWhere of equity capital, cost rf equals risk-free rate or the Ke equals of equity capital, rf equals risk-free ra rate of return paid on risk-free investments such bills, as treasury bills, the beta or m investments such as treasury and β equals and β equals the beta or measure of risk calculated as a regression on the firms’ stock price. The higher the volatility, the hi on the firms’ stock price. The higher the volatility, the higher the of return 2009; Vernimmen e beta and relative risk.equals And mmarket equals rate market rate of (McLaney, return (McLaney, study Since is on external retained 2009; Vernimmen et this al., 2014). the centralfinancing, theme of this study earnings are as is on external financing, retainedBesides, earningsthe are effect assumed and earnings not literature. of away retained on lower discussed in this literature. Besides, the effect of retained earnings on doubt. Unfortunately, the resulting overall cost of capital lowering the overall cost of capital is not in doubt. Unfortunately, the data(kcannot be applied derive the private cost of cap resulting overall costreturn of capital ) formulated with to large enterprises o discount modelthe to determine of return required return data cannot be applied rate to derive private costthe of rate capital. Slee (2011) recommends a private discount rate model to determine following form:

Besides, theretained effect ofearnings retained earni thislowering study is the on external financing, this study isliterature. on external financing, retained earnings aras al., Since central theme of the 2014). effect of retained earnings on the overall cost ofof capital isaway not in this study isthe onUnfortunately, external financing, retained earnings are assumed and notare discu literature. Besides, effect of retained earnings on lowering the doubt. the resulting overall cost capital (k ) formulated w o doubt. Unfortunately, the resulting overall co Besides, the effect oflarge retained earnings on(k lower literature. Besides, the effect of retained earnings on low umed away andBesides, not discussed inliterature. this ly, the resulting overall cost of capital (k with enterprises o) formulated literature. the effect of retained earnings on lowering the overall cost of cap doubt. Unfortunately, the resulting overall cost of capital ) fo o return data cannot beofapplied toFinance, derive private cost of2021, capital. Slee (2011) re return data cannot be applied derive the priva The International Journal Banking and Vol.the 16, Number 1 resulting (January) pp:to 1–20 doubt. Unfortunately, the resulting overall cost of capital doubt. Unfortunately, the overall cost of large capi ge applied thedoubt. overall cost of capital iscost not incannot be to Unfortunately, derive the private of capital. Slee (2011) recommends a private the resulting overall cost of capital (k ) formulated with o return data be applied to derive the private cost of capital. Sl discount rate model to determinediscount the rate rate of return required by private capital model to determine thecost rate ofcap re return data cannot becapital applied tocapital. derive the private of return data cannot becost applied to of derive the private cost of ) formulated with large enterprises loto determine the rate of return required by private investors. It is of the return data cannot be applied to derive the private of Slee (2011) recommen discount rate model to determine the rate return required by priv following form: following form: discount model to determine theprivate of return required discount rate torequired determine the of return requi the model rate offollowing required by private capital investors. Itrate israte of capital the al. Slee (2011)rate recommends a private discount toreturn determine therate rate ofmodel return by investor form: following form: following form: following form: 𝑀𝑀𝑀𝑀 y private capitalform: investors.𝑁𝑁 It is of the 𝑀𝑀𝑀𝑀𝑖𝑖 following 𝑖𝑖 ∑]𝑁𝑁 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑀𝑀𝑀𝑀 =𝑖𝑖 ∑𝑖𝑖=1[(𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 ) 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 × ∑𝑁𝑁 =𝑀𝑀𝑀𝑀 𝑖𝑖 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 ) × ∑𝑁𝑁 𝑀𝑀𝑀𝑀 ] 𝑖𝑖=1[(𝐶𝐶𝐶𝐶𝐶𝐶 𝑀𝑀𝑀𝑀 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖=1 𝑖𝑖=1 𝐶𝐶𝐶𝐶𝑖𝑖 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 ) × ∑ ] (3) = ∑𝑁𝑁 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑁𝑁 𝑀𝑀𝑀𝑀 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑖𝑖 𝑁𝑁 𝑖𝑖 ) × ∑𝑁𝑁 𝑀𝑀𝑀𝑀 ] 𝑀𝑀𝑀𝑀𝑖𝑖𝑀𝑀𝑀𝑀𝑖𝑖(3) 𝑖𝑖=1[(𝐶𝐶𝐶𝐶𝐶𝐶 𝑁𝑁 𝑖𝑖 𝑖𝑖=1 𝑖𝑖 × 𝑖𝑖=1 ∑𝑀𝑀𝑀𝑀 ∑𝑖𝑖[(𝐶𝐶𝐶𝐶𝐶𝐶 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = = 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 ] 𝑖𝑖 +𝑖𝑖 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖 ) × 𝑁𝑁 ∑𝑀𝑀𝑀𝑀 𝑖𝑖 ) ∑ 𝑖𝑖=1 𝑁𝑁 ] 𝑖𝑖=1[(𝐶𝐶𝐶𝐶𝐶𝐶 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = ∑𝑁𝑁 𝑖𝑖=1[(𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 ) × ∑𝑁𝑁 𝑀𝑀𝑀𝑀 ] 𝑖𝑖=1 𝑖𝑖=1𝑖𝑖𝑀𝑀𝑀𝑀𝑖𝑖 𝑖𝑖 𝑖𝑖=1 Where: N equals number ofmedian sources expected of capital;r Where: number of 𝐶𝐶𝐶𝐶𝐶𝐶 equals median Where:NNequals equals numberofofsources sources ofcapital; capital; (3) 𝑖𝑖 equals umber of sources of capital; 𝐶𝐶𝐶𝐶𝐶𝐶 equals median expected return for capital type Where: N equals number of sources of capital; 𝐶𝐶𝐶𝐶𝐶𝐶 equals median 𝑖𝑖 returnspecific for capital type 𝑖𝑖; adjustment 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 equals risk adjustment 𝑖𝑖𝑖𝑖; while 𝑖𝑖; expected 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 equals CAP risk forspecific capital CAP type risk 𝑀𝑀𝑀𝑀𝑖𝑖 eq Where: N equals number of sources of capital; 𝐶𝐶𝐶𝐶𝐶𝐶 equals Where: N equals number of sources of capital; 𝐶𝐶𝐶𝐶𝐶𝐶 equ 𝑖𝑖 𝑖𝑖 𝑖𝑖; adjustment for capital type equals market value of ecificWhere: CAP risk adjustment for of capital type 𝑖𝑖; while 𝑀𝑀𝑀𝑀 value of Noutstanding equals number sources of capital; 𝐶𝐶𝐶𝐶𝐶𝐶 equals median expected return for 𝑖𝑖;securities. 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 equals specific CAP risk adjustment for capital type w 𝑖𝑖 outstanding Theinvolves process of determin 𝑖𝑖 securities. 𝑖𝑖 The process of determining PCOC four steps. 𝑖𝑖; 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 equals specific CAP risk adjustment for capital ty 𝑖𝑖; 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 equals specific CAP risk adjustment for capital outstanding securities. The process of determining PCOC involves edian expected return for capital type 𝑖𝑖adjustment 𝑖𝑖 ies. The process of determining PCOC involves four steps. First is the to determine 𝑖𝑖; 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 equals specific CAP risk capital type 𝑖𝑖;types; while 𝑀𝑀𝑀𝑀 mar the appropriate second isinvolves dete outstanding securities. The process ofcapital determining PCOC 𝑖𝑖the 𝑖𝑖 equals appropriate capital second isfor to determine market value oftoeach four steps. First is outstanding totypes; determine the appropriate capital types; outstanding securities. The process of determining PCOC securities. The process of determining PCOC 𝑖𝑖; while 𝑀𝑀𝑀𝑀 equals market value of ital types; second is toadetermine the market value of each capital type; is 𝑖𝑖 outstanding securities. The process of determining four steps. First is totypes; apply aPCOC specific type (SCAP) risk a the capital second isinvolves tocapital determine the market vain tosecond apply capital type (SCAP) risk adjustment tothird the selected med isspecific to appropriate determine the market value of each capital type; the appropriate capital types; second is to determine the ma the appropriate capital types; second is to determine the olves four steps. First is to determine capital (SCAP) adjustment to the selected median capital type; andeach fourth iscapital to calculate the percentage of capital thetype appropriate types; second iscapital to determine thestructure market value of apply specific capital (SCAP) risk adjustment tocapital theacce set third iscapital isrisk totocalculate apply a aspecific type (SCAP) risk adjustment fourth to the percentage oftype for each capital to apply a specific capital type (SCAP) risk adjustment toc to apply a specific capital type (SCAP) risk adjustmen et value of each capital type; third is te thetopercentage of capital structure for eachtype; capital point and add themedian individual percentages to derive PCOC. Acco applyindividual atospecific capital (SCAP) risk adjustment the selected capit the selected median capital andaccess fourth istoto calculate the fourth istype to calculate the percentage of capital structure forbetween each percentages to derive PCOC. Accordingly, the difference fourth is to thecapital percentage ofpoint capital structure for fourth is tocapital calculate the percentage of capital he median capital type; and of capital structure for access andaccess gesselected to derive PCOC. Accordingly, the between WACC and PCOC lies in the manner of risk weighting. While fourth islies topercentage calculate the percentage ofcalculate structure for each capital point individual toeach derive PCOC. Accordingly, thestructure differen in the manner of percentages riskdifference weighting. While the overall cost of capital for add point the individual percentages to derive for PCOC. Accordingly, the individual to PCOC. the individual percentages to derive PCOC. Accordingly, ach capital access and add weighted in derive the known broad classifications of risk weighting. While the overall costpercentages of capital public firms isAccordingly, riskindividual percentages to derive PCOC. Accordingly, the difference between WACC in thethe manner of risk weighting. While the overall cost of (th cd weighted inlies the known broad classifications such as debt(𝑘𝑘 ), preference 𝑑𝑑risk difference between WACC and PCOC lies in the manner of lies in the manner of risk weighting. While the overall cfi lies the manner of risk While the overal ference between WACCweighted and own broad such asin debt(𝑘𝑘 ),While preference (𝑘𝑘 ),weighting. ordinary (𝑘𝑘 )capital and retained earnings(𝑘𝑘 accou lies in classifications the manner ofWhile riskPCOC weighting. the overall cost of such capital for public thein known classifications asPCOC debt(𝑘𝑘 𝑑𝑑cost 𝑝𝑝public 𝑟𝑟 ), 𝑑𝑑 ), p the overall of𝑒𝑒broad for firms isequity risk(𝑘𝑘weighting. earnings(𝑘𝑘 ), PCOC accounts for risk properties of each 𝑒𝑒 ) and retained 𝑟𝑟 weighted in the known broad classifications such as),debt( weighted inwithin the known broad classifications such as de t of capital forPCOC public firms is riskthe debt and equity classifications. weighted in thethe known broad classifications asPCOC debt(𝑘𝑘 ), preference, preference earnings(𝑘𝑘 accounts for risk properties ofsuch each specific type weighted in known broad classifications such debt 𝑑𝑑capital 𝑝𝑝 ord (𝑘𝑘the and retained earnings(𝑘𝑘 accounts for risk(𝑘𝑘properti 𝑟𝑟 ),within 𝑒𝑒 )and 𝑟𝑟 ), as debt equity classifications. (𝑘𝑘𝑒𝑒(𝑘𝑘 )),and retained earnings(𝑘𝑘 accounts riskrisp ), preference (𝑘𝑘𝑝𝑝 ), ordinary ordinary equity )and and retained earnings(𝑘𝑘 ), PCOC for equity retained earnings PCOC accounts 𝑟𝑟 ),properties 𝑑𝑑equity 𝑒𝑒 PCOC 𝑟𝑟PCOC (𝑘𝑘𝑒𝑒classifications. ) and retained earnings(𝑘𝑘 accounts for risk ofaccounts eachfor specific within the and equity classifications. 𝑟𝑟debt for risk properties ofwithin each capital typeclassifications. within theValue debt and thespecific debtdebt and equity the and equity classifications. Incremental Business operties of each specific capitalwithin type within the debt and equity classifications. Incremental Business Value equity classifications. ness Value Incremental Business Value Incremental Business Value Incremental Business Value Incremental Business Value Incremental Business Value

Incremental business value (IBV) is the result of generating a return in5 excess of the corresponding cost of capital. Recently, IBV has become the most preferred value 5 metric because it applies economic definition ofIncremental value rather than value accounting value. is dynamic, business (IBV) isbook the result of Secondly, generating aitreturn in excess of the corres applicable for continuous valuehas determination processes, and value is multicost of capital. Recently, IBV become the most preferred metric because it purposed such as for appraising whole entity, strategic business units, economic definition of value rather than accounting book value. Secondly, it is dynamic, ap product lines, value specific projects or employees. is also useful foras for appraisin for continuous determination processes, and isItmulti-purposed such capital decisions and in measuring the efficiency capital It is also us entity, allocation strategic business units, product lines, specific projects orofemployees. employed. For private firms, expected the cash flow is of discounted by capital allocation decisions andthe in measuring efficiency capital employed. For priva athe rateexpected derivedcash upon the use of private cost of capital model given in flow is discounted by a rate derived upon the use of private cost of capita Equation 3. It is given given in Equation 3. It isthus: given thus:

Where incremental business value; and EBITDA is earnings Where IBV IBV isisincremental business value; and EBITDA is earnings before interest, tax depr before interest, tax depreciation which isexpenses adjusted and amortisation, which is adjustedand for amortisation, owners’ discretionary and one-time en for owners’ discretionary expenses and one-time enterprise expenses. expenses. For investment, the relevant figure is the greater of the total amount of capital inv or the financial market value. A positive IBV reflects shareholders’ value accretion and a n IBV indicates depletion. While Drucker (1998) alluded that IBV is the genuine econom that arises from generating revenue beyond the corresponding economic cost, Mi Modigliani (1958; 1963) found it as the key indicator of the required rate of market return

For investment, the relevant figure is the greater of the total amount of capital investment or the financial market value. A positive IBV reflects shareholders’ value accretion and a negative IBV indicates depletion. While Drucker (1998) alluded that IBV is the genuine economic profit that arises from generating revenue beyond the corresponding economic cost, Miller and Modigliani (1958; 1963) found it as the key indicator of the required rate of market return, which is sufficient to compensate for risk and economic income. It was this understanding that laid the foundation for discounted cash flow methodology, and the use of net present value (NPV). Another metric that became more acceptable in the industry than NPV is economic value added (EVA), patented by Stern Stewart & Co. The difference between IBV and EVA lies in the nature of expected cash flow to be discounted. In EVA the expected cash flow is the same as operating profit, whereas in IBV it is EBITDA (Slee, 2011; Copeland et al., 1996; and Kanabali & Kashinath, 2015). Garcio and Aguilera (2014) made a novel departure in the process of identifying and measuring value by redefining the taxonomy of value creation and appropriation. They proposed incremental value creation and appropriation, which in their view, attended to the difficulty in estimating customers and suppliers value-added. They assumed that the sum of value created should not always be equal to the sum of value appropriated, leading to a zero-sum game at all times. In other words, their position is that value creation and appropriation could be positive, negative or zero sum. Also, their research notes asserted that value is not necessarily created by fund providers alone but by multiple stakeholders. To that extent, they classified previously used accounting measures of value as too absolute. While we acknowledge the wider scope and application of coverage of incremental value, as noted above, this work is concerned only with private capital elasticities of incremental business value. Middle Market Finance Theory Middle market finance theory finds a dividing line between small and medium enterprises finance theories on the one hand, and between medium and large enterprises on the other hand. In doing so, the midmarket finance theory identifies a triadic integration of valuation, capitalisation and transfer of ownership interests that explains capital market decisions of medium-sized firms. Slee (2011) argues that valuation forms the base of the interconnection, and maintains a balance with the other two components, without which much of it will be done in isolation from the market and will amount to vanity. Absence of this balance will be a source of disequilibrium in the case of private securities that do not have access to the active trading market. They must rely on point-in-time appraisal or discrete rather than continuous pricing and value determination. Next, capitalisation rests on valuation to enable private capital markets price and allocate capital, given their risk-return sensitivity. Although every round of capitalization requires pre- and post-money valuation, the inefficiency of the private capital markets, arising from informational asymmetry, distorts the valuation results. (Leleux, Swaay & Megally, 2015). Finally, transfer of ownership interest relies on capitalisation and valuation. It takes different forms and involves parties from within and without the firm. The shortcomings of the earlier pillars inhibit the ease of transfers of ownership of private firms. Lack of liquidity hinders diversification, which increases the riskiness of private transfer markets. It is the coherent importance of all three concepts at any given time in exhibiting the same weakness of mid-market firms that inspired the thinking of constituting the midmarket finance theory.

Ex-post facto research design was adopted since the research relied on historical data generated from annual reports of mid-market enterprises. Besides, because of our limited ability to subject the study environment to an acceptable degree of control and our normative processes in the inclusion of independent variables, the study fits the quasi-experimental research design. The population consists of all enterprises in Nigeria whose total assets (excluding land and building) are above one hundred and fifty thousand USD but not exceeding one million five hundred thousand USD and with a total workforce above fifty employees but not exceeding one hundred and ninety-nine employees (SMEDAN, 2013). A sample of ten such enterprises listed under the Nigerian Association of Stock Dealers (NASD), an overthe-counter securities market, was selected not without the following considerations: maintenance of balanced panel, ease of data collection, reliability of data, and completeness of requisite data. (CSCS, 2018).

Ten years data were collected covering 2009 to 2018 from a cross section of the sample. It may seem at first that the time series data for ten years were insufficient, nevertheless, the study remains robust on account of the cross sectional data, which sufficiently compensated for the limited time series, leaving us with a panel dataset of 100 data points.

We employed a two-stage analytical process. The first stage examined the panel regression models, showing two alternate models: fixed effect and random effect models. The second stage tested for serial autocorrelation to show if there is a serial relationship between a variable in time t and the same variable in time t - l. Two models were formulated withautocorrelation IBV as dependent in each costrelationship bet serial to variables show if there is case: a serial serial autocorrelation to show if there is a serial relationship of debt (represented by the one most important source of debt capital same variable in time t - l. Two models were formulated with IB variable time t - l. Two modelsby were in the firms’ capitalsame structure) and in cost of equity (represented oneformulated with case: cost of debt (represented by the one most important source o case:of cost of debt (represented by the one most most important source equity in the firms’ capital structure) as important sour structure) and cost of equity (represented by one most importa independent variables for one and model; an independent structure) costand of PCOC, equity as(represented by one most impo structure) as independent variables one model; and P variable for thecapital other model. It is the proxy for overall cost of for capital capital structure) as independent variables for one model; an for private firms. The first model examined the individual effects for the other model. It is the proxy for overall cost of capital f for thefor other model. It is firm the while proxy forsecond overall cost of capit of each source examined of capital the the mid-market individual effects of each the source of capital for the m examined the individual effects of forret autocorrelation to each show if thereofiscapital a serial looked at the global effect. serial Both models were isolated tosource avoid looked at the global effect. Both models were isolated to avoid mu looked at the global models were isolated to avoid multicollinearity. The model follows theeffect. classical same variable in Both timelinear t - l. regression Two models were form the classical linear regression equation of the following form: equation of the following form: the classical linear regression equation of the following form: case: cost of debt (represented by the one most impo structure) and cost of equity (represented by one (4) 𝑖𝑖  0  capital 1  𝑃𝑃𝑃𝑃𝑃𝑃 2  𝑃𝑃𝑃𝑃𝑃𝑃 𝐼𝐼𝐼𝐼𝐼𝐼 = + + + variables for one 𝑖𝑖 1 structure) 2 as independent for the other model. It is the proxy for overall co

𝐼𝐼𝐼𝐼𝐼𝐼 = 0 +  𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 +the 𝑖𝑖 individual effects of(5) 1  𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 each source of c 𝐼𝐼𝐼𝐼𝐼𝐼 = 0examined + + 𝑖𝑖 looked at the global effect. Both models were isolate Where IBV equals incremental business value, business calculatedvalue, for each Where IBV equals incremental calculated each theequals classical linear regression equation of theforfollow Where IBV incremental business enterprise as defined in equation 3; PCD equals private cost ofvalue, debt; calculated for 3; PCD equals private cost of debt; PCE equals private cost of eq PCD equalsPCOC private costprivate of debt; PCE equals private cost of PCE equals private 3; cost of equity; equals cost of capital; capital;  equals constant term;   equal of pred 𝐼𝐼𝐼𝐼𝐼𝐼 =  +  𝑃𝑃𝑃𝑃𝑃𝑃 +  𝑃𝑃𝑃𝑃𝑃𝑃 +coefficients 𝑖𝑖 coefficients b0 equals constant term; b1 - b02 equals equal coefficients of predictors; and capital; constant 12- 2 equal of p 1 term; stochastic variable representing thethe uncontrolled country-specific equals error term orstochastic stochastic variable representing uncontrolled variable representing the uncontrolled country-spec country-specificbusiness factors cycle, such aslabour demand volatility, business cycle, etc. (Pourghajan et al., 2012). 𝐼𝐼𝐼𝐼𝐼𝐼 = market, + 1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 + (Pourghajan 𝑖𝑖 cycle, labour etc. et al., 2012). 0 market, labour market, etc. business (Pourghajan et al., 2012).

Two panel regression models were specified in line with fixe

Twomodels panelWhere regression models were specified line with Two panel regression were specified in line with fixed and in value, IBV equals incremental business calcf respectively (See results presented in Table 2). random effectsprocedures, estimation procedures, respectively (See results procedures, results in Table 3; respectively PCD equals (See private cost presented of debt; PCE equals2). priv presented in Table 2).

capital; 0 equals constant term; 1 - 2 equal coeff a. Fixed effects: a. Fixed effects: stochastic variable representing the uncontrolled cou business cycle, labour market, etc. (Pourghajan et a This focuses on whether there are differences by using a fixed This focuses on whether there are differences by using a fi cross-sectional structures, as the difference may be due to sp business cycle, labour market, etc. (Pourghajan et al., 2012).

Two panel regression models were specified in Number line with fixed2021, andpp: random effects estim The International Journal of Banking and Finance, Vol. 16, 1 (January) 1–20 procedures, respectively (See results presented in Table 2).

Fixed Effect a. Fixed effects:

This focuses on whether there are differences by using a fixed intercept This focuses on whether there are differences by using a fixed intercept for each of the di for each of the different cross-sectional structures, as the difference cross-sectional as the difference may be due tofirm. special may be due tostructures, special features of each mid-market It isfeatures given of each mid-m firm. thus:It is given thus: 𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = 1 + 2 D2𝑖𝑖 + 3 D3𝑖𝑖 + 4 D4𝑖𝑖 + 5 D5𝑖𝑖 + ⋯ 10 D10𝑖𝑖 + (6)

relation to show if there is a serial relationship between a variable in time t and the = β1were + β21formulated 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝐼𝐼𝐼𝐼models 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = + D2𝑖𝑖𝑖𝑖 +𝑖𝑖𝑖𝑖β+ D𝑖𝑖𝑖𝑖with + βIBV D + β5 D5𝑖𝑖𝑖𝑖 + ⋯ βvariables D + (7) βin in time t - l. Two𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 each 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + µ(7 11 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑖𝑖𝑖𝑖 3 3𝑖𝑖𝑖𝑖 4 4𝑖𝑖𝑖𝑖 as dependent 10 10𝑖𝑖𝑖𝑖 ebt most important source of debt capital in the firms’ capital + β(represented D + β4 D4𝑖𝑖𝑖𝑖by +the β5 Done 5𝑖𝑖𝑖𝑖 + ⋯ β10 D10𝑖𝑖𝑖𝑖 + β11 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + µ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 3 3𝑖𝑖𝑖𝑖 Because the fixed account for both cross-sectional and time-series data, the inc cost of equity (represented byeffects one most important source of equity in the firms’ covariance caused by individual-firm differences is eliminated, thereby increasi re) as independent variables for effects one model; andforPCOC, as an independent variable Because the fixed cross-sectional andoftimeestimation-result efficiency. account In this case, both equals intercept value firm 1; 2 – 10 rep model. It is theseries proxydata, for overall cost of covariance capital for caused private by firms. The first model the increased individual-firm the differences of the other firms from firm 1, measured by their differences o differences is eliminated, thereby its firm estimation-result individual effects of each source of capital for theincreasing mid-market while the second coefficients; D2𝑖𝑖 − D10𝑖𝑖 represent 9 dummy variables representing 9 firms; 11 - 12 efficiency. In this case, b equals intercept value of firm 1; b2 - bfollows 1 avoid multicollinearity. The model lobal effect. models were isolated to les that are Both merely explained away.and i = 1, 2, …. mid-market firms; and t = 1, 2, coefficients of predictors; equals error from term or stochastic variable represent the differences of the other firms firm 1, measured by representing near regressiontheir equation of the following form: differences of their coefficients; D2i - D10i represent 9 dummy variables that are merely variables representing 9 firms; b11 - b12 equal coefficients of predictors; years. 𝑃𝑃𝑃𝑃𝑃𝑃 +  𝑃𝑃𝑃𝑃𝑃𝑃 +  equals (4) firms and error or stochastic variable variables thatterm are merely explained away. irepresenting = 1, 2, ….10other mid-market

1 effects:2 m variables that are merely explained away. i explained = 1, 2, …. variables that aremerely merely variables that are away.mid-market i = 1, 2, …. ….10 mid-market mid-marke merely 1,1,2,2,…. firms; and merelyexplained explainedaway. away.ii==years. ….10mid-market mid-market firms; andtt=explained =1,1,2,2,...... ...... b. Random effects: firms; and t = 1, 2, ...... years. years. years.

ocuses on the relationship with the study sample as a whole, thus the samp b. Random effects: Random Effects Randomeffects: effects: b.b. Random This total focusessample on the regr rela mly opposed to calculated using the entire population. The uals selected, incrementalasbusiness value, for each enterprise as defined in equation randomly selected, as This on focuses on the relationship with the studyassample as a whole, op th This focuses the relationship with the study sample a whole, onprivate of thecost random can beThis expressed as:equity; This focuses on the relationship relationship study sample focuses the with theprivate sample as aa wh wh of debt;effect PCE equals private cost on of PCOC equals cost ofas function of the random ef randomly asselected, opposed toopposed using the entire population. The total he as thus samples thusthe the study samples areselected, randomly as toare using the entire the relationship relationship with with study sample sample as aa whole, whole, thusasthe the samples are randomly selected, selected, as opposed to using the entire population. Th randomly opposed population. Th uals constant term;   equal coefficients of predictors; and  equals error term or 2entire , as entire The total sample population. The population. total sample regression function of the random effect of the random effect can be regression expressed as: d, as opposed opposed toto using using1the thefunction population. The sample functionof ofthe thetotal random effectregression can be be expressed expressed as: function random effect can as: can be expressed as: =  +  𝑃𝑃𝑃𝑃𝑃𝑃 +  𝑃𝑃𝑃𝑃𝑃𝑃 + w able representing the uncontrolled country-specific factors such as demand dom effect can be expressed as: ndom effect can 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 =volatility, 0 + 1 𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖(8 + 1 be expressed 2as: 𝑖𝑖𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = 0et𝐼𝐼𝐼𝐼𝐼𝐼 +al., 1 𝑃𝑃𝑃𝑃𝑃𝑃 +  𝑃𝑃𝑃𝑃𝑃𝑃 + w , labour market, etc. (Pourghajan 2012). 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 + w𝑖𝑖𝑖𝑖 + 𝑃𝑃𝑃𝑃𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃 +  𝑃𝑃𝑃𝑃𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃 𝑖𝑖𝑖𝑖 + 𝐼𝐼𝐼𝐼𝐼𝐼 𝑖𝑖𝑖𝑖 ==  𝑖𝑖𝑖𝑖+ +w

𝑃𝑃𝑃𝑃𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖++22𝑃𝑃𝑃𝑃𝑃𝑃 𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖++ww𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = 0 + 1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖 = 0 + 1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖 + w𝑖𝑖𝑖𝑖 (9 𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = 0in +  𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 + w 𝐼𝐼𝐼𝐼𝐼𝐼 =  +  𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 + w gression models were specified line with fixed and random effects estimation 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 + w 𝑖𝑖𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖1= 00+𝑖𝑖𝑖𝑖11𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃

(9) 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (9) 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 +ww𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (9) 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖+ spectively (See results presented in Table 2). If this is represented with r thiswith represented with random variables, then  ++ i,i,the which in s represented Ifwith then 0𝑖𝑖with = variables, random + i, which diff If is variables, represented with random then indicates =  + i, =i, which indicate thisrandom is this represented random variables, then IfIfthis isisrepresented variables, =that which i 0𝑖𝑖 then +0𝑖𝑖 0𝑖𝑖 occurs randomly, and the ddwith random variables, then  =  + i, which indicates that the difference with random variables, then  =  + i, which indicates that the difference occurs randomly, and the expected value of  is  , and w = ε which indicates that the difference occurs randomly, and the expected 0𝑖𝑖 𝑖𝑖𝑖𝑖+= occurs0𝑖𝑖randomly, and valuewof𝑖𝑖𝑖𝑖value = isof0+ and = εw𝑖𝑖𝑖𝑖is, με𝑖𝑖 0 , that occurs andthe expected ,0𝑖𝑖 is w and cts: randomly, and the expected valuerandomly, of the 0𝑖𝑖 expected is , and 𝑖𝑖𝑖𝑖 th 0𝑖𝑖 ε𝑖𝑖𝑖𝑖 0𝑖𝑖μ𝑖𝑖𝑖𝑖 ,0 𝑖𝑖𝑖𝑖 0 two nd the 0𝑖𝑖 isis 0 , ,and and ww𝑖𝑖𝑖𝑖w = ε + μ , that is, the error term w is a sum of tw value ofterm that is, the error a and theexpected expected value valueof and = ε + μ , that is, the error 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 term is a sum of parts: ε is the combined time serie 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 0𝑖𝑖 w𝑖𝑖𝑖𝑖0 is a sum of atwo is theε𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖combined series term w𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is sumparts: of twoε𝑖𝑖𝑖𝑖parts: is the combined time and serie w sum εsum two parts: is the combined time series andtime cross-section ofthe two parts: εcomponent, is the combined timecross-section series and cross-section 𝑖𝑖𝑖𝑖 component, and μ is the two combined time series and cross-section error and μ is the or firm-specific error compo m𝑖𝑖𝑖𝑖of of is twoa parts: parts: εof is the combined time series and cross-section error 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is component, and μ𝑖𝑖𝑖𝑖 is intercept the cross-section or firm-specific error compo and cross-section orfor firm-specific component. ses on whether there arecomponent, differences byμusing a fixed each of theerror different 𝑖𝑖𝑖𝑖 is the or error component. nent, and μ𝑖𝑖𝑖𝑖 is the cross-section firm-specific error component. isthe thecross-section cross-section orfirm-specific firm-specific erroror component. 𝑖𝑖is ional structures, as the difference be due of each mid-market c. may Hausman test:to special features c. Hausman test: c. test: Hausman test: c. Hausman given thus:

man test:

This is the most commonly used method for evaluating fixed and the most commonly used for method fixed and Thisfor is evaluating thefixed mostand common This is theThis mostis commonly used method evaluating rando

If this is represented with random variables, then 0𝑖𝑖 =  + i, which indica The International Journal of Banking Finance, Vol. 16, Number 1 (January) occurs randomly, andandthe expected value of 0𝑖𝑖2021, is pp: 01–20 , and w𝑖𝑖𝑖𝑖 = ε𝑖𝑖𝑖𝑖 + μ term w𝑖𝑖𝑖𝑖 is a sum of two parts: ε𝑖𝑖𝑖𝑖 is the combined time series and errorcomponent, component, and and μ𝑖𝑖𝑖𝑖 isisthe the cross-section cross-section or error orfirm-specific firm-specific error component. component.

Hausman test:

Hausman Test

ThisThis is theismost commonly used method for evaluating fixed andfixed and rand the most commonly used method for evaluating random effects. It shows whether sectional unit differences fixed It has an as whether sectional unit differences are fixed or are random. or random. It has an asymptotic chi-square distribution with degrees distribution with degrees of freedom equal to the number of independ of freedom equal to the number of independent variables. If the variables are significantly then theestimation fixed effects estimatio variables are significantly correlated,correlated, then the fixed effects efficient, should adopted. On theOnother hand, if the variables is consistent andand efficient, andbeshould be adopted. the other hand, if the variables then are not significantly correlated, thenis the randomand efficient, an correlated, the random effects estimation consistent effects estimation is consistent and efficient, and should be adopted. (Gujurati, Porter, & Gunasakar, 2013). (Gujurati, Porter & Gunasakar, 2013).

Serial autocorrelation was also conducted test the existence Serial autocorrelation was also conducted to test thetoexistence of serial of serial rela variablebetween in time one t and the same variable in same time variable t - 1. The relationship variable in time t and the in result shows timeamongst t - 1. The the result shows no serial correlation amongst the variables. variables.

The following tables explain the dynamic relationship between the dependent and independent variables. First is the analysis of Hausman test. The null hypothesis states that the fixed effect does not have a better model fit than random effect. Here, null hypothesis was rejected, implying that the fixed effect model does have a better model fit than the random effect model. The above result now leaves us with the analysis of fixed effect model alone.

F-Test

DW