Critical factors for insolvency prediction: towards a theoretical model for the construction industry

Many construction industry insolvency prediction model (CI-IPM) studies have arbitrarily employed or simply adopted from previous studies different insolvency factors, without justification, leading to poorly performing CI-IPMs. This is due to the absence of a framework for selection of relevant factors. To identify the most important insolvency factors for a high-performance CI-IPM, this study used three approaches. Firstly, systematic review was used to identify all existing factors. Secondly, frequency of factor use and accuracy of models in the reviewed studies were analysed to establish the important factors. Finally, using a questionnaire survey of CI professionals, the importance levels of factors were validated using the Cronbach's alpha reliability coefficient and significant index ranking. The findings show that the important quantitative factors are profitability, liquidity, leverage, management efficiency and cash flow. While important qualitative factors are management/owner characteristics, internal strategy, management decision making, macroeconomic firm characteristics and sustainability. These factors, which align with existing insolvency-related theories, including Porter's five competitive forces and Mintzberg's 5Ps (plan, ploy, pattern, position and perspective) of strategy, were used to develop a theoretical framework. This study contributes to the debate on the need to amalgamate qualitative and quantitative factors to develop a valid CI-IPM.


Introduction
As much as owners and major stakeholders do not like to hear it, the prospect of construction business insolvency in any case is a real one. The negative impact of such insolvencies on the economy and society in general has led to the development of many insolvency prediction models. However, the effectiveness of an insolvency prediction model (IPM) is dependent on, amongst other elements, the variables that are chosen to develop it. These variables are used to measure various factors that may affect the insolvency of a construction firm. Many construction industry (CI) studies have employed different variables in their work, chosen either arbitrarily (Chen 2012), by statistical analysis (Abidali & Harris 1995;Ng et al. 2011;Bal et al. 2013) or by adoption from previous studies, which is more common with non-construction studies (Wilson & Sharda 1994;Boritz & Kennedy 1995). This is because there was, and is still, no clear theoretical framework for choosing insolvency factors or variables (Du Jardin 2012); a defect that is restraining scientific advances towards a highly effective insolvency prediction for the CI (Balcaen & Ooghe 2006). At this early stage, it is imperative to distinguish between variables and factors as referred to in this study.
Variables: A variable is a measurable quantity that represents a certain characteristic of a firm, usually in the form of a numeric value. Financial ratios are the most common variables in IPM research. Variables can also be obtained through a Likert scale questionnaire. Example variables include current ratio, quick ratio, age of firm, turnover (size) of firm, etc. Factor: This is the characteristic being measured by a variable. There are always many variables that can be used to measure a particular factor. Variables that measure the same factor belong to the same group; the group here is what is termed a factor. The aforementioned current ratio and quick ratio belong to the 'liquidity' factor while age and turnover (size) of firm belong to the 'firm characteristics' factor.
Pioneering prediction studies normally employed a number of factors with a large number of quantitative variables to measure them, usually in the form of financial ratios, based on authors' experience and presence of ratios in financial statement of sample firms, before using statistical analysis to select a limited number of ratios for the prediction model (Beaver 1966;Altman 1968). 'A financial ratio is a quotient of two numbers, where both numbers consist of financial statement items' (Beaver 1966, pp. 71-72). Older and more recent construction industry insolvency prediction models (CI-IPM) studies (e.g. Mason & Harris 1979;Abidali & Harris 1995;Ng et al. 2011;Bal et al. 2013;Horta & Camanho 2013) have erroneously simply copied the methods of early IPM studies. This is because the CI-IPM literature has not provided any real coherent theory underpinning the use of financial ratios along with the insolvency factors they measure (Du Jardin 2012). Factors chosen because of their presence in financial statements of sample firms as done by virtually all IPM studies are generally sample specific (Balcaen & Ooghe 2006;Hafiz et al. 2015), thus making them unfit for generalization and consequently inappropriate for adoption.
Although their exclusive use is common with an overriding percentage of existing CI-IPMs due to blind copying of past methods (Ng et al. 2011;Huang 2009;Chen 2012;Bal et al. 2013 among others), using quantitative factors alone to develop a prediction model for the CI is insufficient since financial distress only tends to be noticeable when the failure is almost complete (Abidali & Harris 1995). Though many failure-related theories are finance centred, there are as many non-financial failure-related theories which are known to be viable. These (non-financial theories) include Michael Porter's five forces of competitive position model and Mintzberg's five Ps of strategy, which are employed in this study, among others. As Argenti (1976, p. 138) rightly said: 'while these (financial) ratios may show that there is something wrong. I doubt whether one would dare to predict collapse or failure on the evidence of these ratios alone'. In fact, it is adverse managerial actions, poor company strategy, etc. (qualitative factors) that normally lead to poor financial standing of construction businesses and in turn cause insolvency. Hence, to achieve early prediction, which is required in any robust prediction model to allow enough time for remedy, the use of qualitative factors is important and has been strongly encouraged (Arditi et al. 2000;Koksal & Arditi 2004;Horta & Camanho 2013;Alaka et al. 2015 among others). However, the use of qualitative factors in developing CI-IPMs have been hampered by their being unreadily available and the absence of a theoretical framework which encompasses qualitative and quantitative insolvency factors for construction firms.
Unlike simply finding the causes of failure of construction firms as done by Holt (2013), this study seeks to establish the CI insolvency factors that can help create more valid CI-IPMs. The main goal is to create a comprehensive theoretical framework that will form the platform for selection of the most important CI insolvency factors and explain the relative importance of each in relation to the solvency of construction businesses. To achieve this aim, the study has following objectives: 1. To identify the CI insolvency factors that largely influence the performance of CI-IPMs through a systematic review of literature. 2. To analyse the summary of findings of the systematic review and rank the identified factors in order to establish the most important ones. 3. To validate the importance level of the factors by using statistical analysis of questionnaire data from experienced CI professionals to triangulate the review analysis.
This study will contribute to knowledge by presenting and justifying the most important CI insolvency factors required to build a high performance CI-IPM, as omission of such factors can easily lead to a poor CI-IPM. This study will also eliminate the problem of analysing variables under all existing factors in order to identify the important ones before building a CI-IPM. Only variables under the identified factors will need to be analysed. This will greatly improve the efficiency of the CI-IPM development process. The scope of this work is limited to identifying and verifying the most important factors for developing CI-IPMs. The validation is via questionnaire data with responses from experienced professionals in the construction industry. Developing a CI-IPM by collecting numerous construction firms' historic data and checking its accuracy is outside the scope of this study as it does not seek to build a CI-IPM. The next section describes the methodology by first explaining the systematic review of the quantitative and qualitative factors before describing the questionnaire methods for both factor types. The data analysis section then follows, describing step-by-step analysis of data from the systematic reviews and questionnaires. This is followed by the results of the analysis. The discussion and proposed framework section discusses how the results relate to existing theories and construction world hierarchy, while the conclusion rounds up the work.

Methodology
The philosophical paradigm adopted in this study is pragmatism. This is because it advocates using a combination of any set of methods that best answers the research questions or best achieves the research objectives rather than rigidly dictating specific methods (Johnson & Onwuegbuzie 2004). It allows the researcher to 'study what interests you and is of value to you, study in the different ways in which you deem appropriate, and use the results in ways that can bring about positive consequences within your value system' (Tashakkori & Teddlie 1998, p. 30).
This study uses a mixed method approach to identify the important qualitative and quantitative factors required to develop a high performance CI-IPM. In each case, the factors are initially aggregated using a systematic literature review and ranked based on the frequency of usage. For quantitative factors, the accuracy of CI-IPMs that have used them are also considered in the ranking. For validation purposes, the Cronbach's alpha reliability coefficient and significant index ranking of questionnaire data from construction industry professionals were used to triangulate the results from the systematic review. Triangulation is defined as the 'use of two or more independent sources of data or data collection methods to corroborate research findings within a study' (Saunders & Paul 2013, p. 154).
Systematic review for quantitative factors 'A systematic review is a summary of the research literature that is focused on a single question. It is conducted in a manner that tries to identify, select, appraise and synthesize all high quality research evidence relevant to that question' (Bettany-Saltikov 2012, p. 5). The systematic literature review method calls for a broad search of the literature (Smith et al. 2011) using unambiguous exclusion andinclusion criteria (Nicol as &Toval 2009). Systematic review is renowned for yielding valid and repeatable/reliable results because it reduces bias to a minimum; hence its high level of recognition and frequent use in the all-important medical research world (Tranfield et al. 2003;Schlosser 2007) and its use in other research areas like IPM (Appiah et al. 2015). The general review and synthesis of various existing knowledge is also a recognized method which contributes immensely to the progression and expansion of knowledge (Aveyard 2007;Fink 2010). This is the reason it has been widely employed as methodology in various research areas including insolvency prediction (Adnan Aziz & Dar 2006;Balcaen & Ooghe 2006) and construction business failures (Edum-Fotwe et al. 1996;Mahamid 2012).
The single research question on which the systematic review of this study focuses is 'which are the most important insolvency prediction factors (quantitative and qualitative) for construction firms?' Since results from peer reviewed journals are generally considered to be of high quality and validity (Schlosser 2007), this systematic review employs only peer reviewed journals. This will ensure a high validity of the review results.
The databases searched for this review include Google Scholar (GS); Wiley Interscience (WI); Science Direct (SD); Web of Science UK (WoS); and Business Source Complete (BSC). This is done in tandem with the latest published systematic review article on IPM (i.e. Appiah et al. 2015). Observations revealed that GS, WoS and BSC contained all the journal articles provided in Wiley and Science Direct since the latter are publishers while the former are general databases. To further broaden the search, the Engineering Village (EV) database was added to the GS, WoS and BSC databases for the final search.
Pilot searches revealed that studies use bankruptcy, insolvency and financial distress interchangeably to depict the failure of firms. A search structure which included all these words was subsequently designed with the following defined string ('Forecasting' OR 'Prediction' OR 'Predicting') AND ('Bankruptcy' OR 'Insolvency' OR 'Distress' OR 'Default' OR 'Failure') AND ('Construction' OR 'Contractor'). A process flow of the systematic review methodology for quantitative factors is presented in Figure 1.
To avoid database bias, ensure high repeatability and consistency of this study, and consequently high reliability and quality, all the relevant studies that emerged from searching the databases were employed in the review (Schlosser 2007). Since the databases host studies from around the globe, geographic bias was readily averted. Considering that the first set of IPM studies emerged in the 1960s (Beaver 1966;Altman 1968), the period 1960-2015 was used for the search One of the inclusion criteria was for the IPM study to focus solely, or mainly, on the CI. Another is that the study must employ quantitative factors (i.e. financial ratios as variables). The titles and abstracts of the studies that the search returned were typically adequate to decide the ones qualified for use in this study; otherwise articles' introductions and/or conclusions were read to determine their suitability. The extent of reading was dependent on the information from initial readings. In exceptional cases, the whole article was read. At the end, GS produced 31 results, EV 14, BSC 11 and WoS 7. Most of the articles returned in searching EV, BSC and WoS were present in the GS search results. In fact, all EV results were present in the GS result, while BSC and WoS were only able to produce four and one unique articles respectively The exclusion criteria included, among others, articles that were not written in English. Although language constraint is not favoured in systematic review, it is unavoidable and thus acceptable when there is a lack of funds to pay for interpretation services (Smith et al. 2011). An example of a study excluded based on language is Wedzki (2005), which is written in Polish. Review studies were not considered as they contained only factors taken from other studies. After removing unsuitable articles with titles like 'default prediction for surety bonding' (e.g. Awad & Fayek 2012) and 'contractor default prediction prior to contract award', which fixate on a contractor's capability to successfully execute a specific kind of project (e.g. Russell & Jaselskis 1992), only 28 studies were left. Note that 'contractor default prediction prior to contract award' articles that fixated on insolvency probability as the main/only judging criteria were not excluded as the studies effectively built a form of CI-IPM.
In the final 28 articles reviewed in this study, where multiple accuracy results are presented for multiple CI-IPMs, only the accuracy result of the technique proposed in the article is presented in this study. Where no particular technique is proposed, the highest accuracy result is presented here. Where the results for training and validation samples are given, the validation result is used here, otherwise the training result is adopted. Where error types are calculated independent of accuracy values and the Receiver Operating Characteristic (ROC) curve is used to determine performance, the area under the curve (AUC) value as a percentage is taken as the accuracy result. Where accuracy results of multiple years are given, the result of the first year is adopted to allow fair comparison since the first-year result is the most commonly presented result in IPM studies. As required for systematic review, a meta-analysis was done with data synthesized through the use of 'Summary of Findings' tables, statistical methods and charts (Higgins 2008;Smith et al. 2011) (see analysis of data section)

Systematic review for qualitative factors
The systematic review for the qualitative factors was quite similar to that of the quantitative factors except for a few differences which are explained here. Pilot searches revealed that there are very few studies that used qualitative factors for their CI-IPM. The quantitative factors review search structure already revealed three studies (i.e. Abidali & Harris 1995;Koksal & Arditi 2004;Horta & Camanho 2013) with qualitative factors, two of which combined quantitative with qualitative factors, and are thus present in this review. Since it is clear that studies using qualitative factors for CI-IPM are scarce, a new search structure was developed to identify studies that provide factors leading to the insolvency of construction firms. A search structure with the following defined string was designed: ('Business' OR 'Firm' OR 'Company') AND ('Bankruptcy' OR 'Insolvency' OR 'Distress' OR 'Default' OR 'Failure') AND ('Construction' OR 'Contractor'). A process flow of the systematic review methodology for qualitative factors is presented in Figure 2.
After various pilot searches and the use of the structured search led to very few and usually unsuitable results in the databases except GS, the search was limited to GS. After a thorough inspection of more than 500 articles only eight suitable articles were found, in addition to the three previously identified. The result was improved by checking review articles and checking through their citations/references. Three more studies were added using this method (Jannadi 1997;Robinson & Maguire 2001;Arslan et al. 2006). With no resulting article identifying the role of environmental, social and governance (ESG) factors in the failure of construction firms, the search words 'sustainability practices and failure of construction companies' were used on Google and the first suitable article (i.e. Siew et al. 2013) was selected. As a result, a total of 15 primary studies was reviewed.

Questionnaire data for quantitative and qualitative factors
The factors identified from the analysis of the systematic review were used to formulate a very simple preliminary questionnaire to determine how important each identified factor is in terms of predicting failure/survival of a construction firm. A Likert scale of one to five was used, where five represents 'most important' and one represents 'least important'. Example qualitative variables were given in brackets for each qualitative factor. The preliminary questionnaire served as a pilot study with the aim of evaluating the relevance, complexity, length and layout of the questionnaire.
Since the quantitative factors represent accounting ratios, the target respondents were insolvency practitioners (normally accountants), who specialize in dealing with construction firms that go into administration or file for bankruptcy. Using the UK government insolvency practitioner online directory, 500 insolvency practitioners were randomly selected and sent the final questionnaire via email. The questionnaires stated clearly that only practitioners with vast experience in dealing with construction firms' insolvency should complete them. Following numerous reminder emails, 106 usable questionnaires were obtained. This was considered a good response rate and was probably down to the simplicity and brevity of the questionnaire.
For the qualitative factors, the target respondents were managerial-level staff of insolvent and existing construction firms. Contacts for insolvent construction firms were obtained in two main ways. The first was to use the FAME Bureau Van Dijk UK financial database to identify directors of failed construction firms, and subsequently identify existing firms where those directors currently worked. Questionnaires were then posted to those directors at the addresses of their current firms, if such information was available. The second was to liaise with college lecturers teaching on construction apprenticeship programmes to allow sharing the questionnaires with the students. Some students were, in themselves, suitable respondents while others volunteered to pass them to their colleagues and/or bosses at work who had worked in a now defunct construction firm. The contacts for existing firms were obtained from the FAME Bureau Van Dijk UK financial database and questionnaires were sent out through post and by email. This method of sampling is known as convenience sampling and has been used in a number of construction studies (e.g. Li et al. 2005;Spillane et al. 2011a;Oyedele 2013). This sampling method became necessary because of the difficulty involved in finding employees of failed firms. Overall, over 500 questionnaires were distributed. Following numerous reminder emails, only 76 usable questionnaires were returned. The demographics of the survey respondents for the quantitative and qualitative factors questionnaires are presented in Table 1.

Analysis of data
In an effort to achieve the main study objectives, which included identifying the most important CI insolvency factors in order to create a comprehensive theoretical framework that will form the platform for selecting important CI insolvency factors, a rigorous statistical process was employed. First the factors were ranked based on frequency of usage in CI-IPM studies, and accuracy of models that used each factor in the case of quantitative variables. This was done using information from the summary of findings of the systematic reviews. CI-IPM studies normally use sample construction firm data and various statistical techniques to identify the best factors (and variables) for their models, and the selected factors are usually susceptible to sample specificity (Balcaen & Ooghe 2006;Agarwal & Taffler 2008;Jackson & Wood 2013 among others). This implies that the most used factors are definitely those that have been consistently selected using different samples and statistical methods; the most used factors are thus suitable for most sample data and are consequently the most important factors.
The questionnaire responses were analysed using reliability tests and then calculating the significance index (SGI). The SGI was then used to rank the factors in terms of level of importance and validate the result from the review analysis by triangulation. The rankings helped to identify the most important factors and the least important factors. The questionnaire responses are a measure of the importance of using experienced practitioners that deal with these factors on almost daily basis and can make very reliable judgements of their importance.

Analysis of quantitative factors
The summary of findings from the systematic review of quantitative factors is presented in Table 2. The quantitative variables (i.e. financial ratios) used in all the primary (i.e. systematically reviewed) studies are presented as well as the factors/ categories the variables fall under. The factors were taken directly from the studies where available; otherwise, the Profitability -2. Mason and Harris (1979) Profit before interest and tax/opening balance sheet net assets; profit before interest and tax/opening sheet net capital employed; debtors/creditors; current liabilities/current assets; log 10 (days debtors); creditors trend measurement Profitability, working capital position (liquidity), leverage, quick assets position, trend 87 3. Kangari et al. (1992) Current ratio, total liabilities to net worth, total assets to revenues, revenues to net working capital, return on total assets, and return on net worth Profitability, Efficiency, Liquidity -4. Langford et al. (1993) Liquid asset ratio; acid test ratio (quick ratio); net worth to fixed assets; working capital to total assets; net profit to net worth; net worth to total liabilities Short-term solvency, solvency, liquidity, profitability 63.33 5. Abidali and Harris (1995) Ratio of earnings after tax and interest charge to net capital employed; ratio of current assets to net assets; ratio of turnover to net assets; ratio of short-term loans to earnings before tax and interest charge; tax trend; earnings after tax trend; short-term loan trend Profitability, leverage, activity/net asset, turnover, liquidity, trend measurement 70.3 6. Russell and Zhai, (1996) Prime interest rate; new construction value in-place; new construction value in place; net worth/total asset; trend-gross profit/total asset; and net working capital/ total asset Trend, future position, volatility 78.3 7. Singh and Tiong (2006) Working capital to current liabilities (WC/CL); cash flow to current liabilities (CF/CL); net worth to total liabilities (NW/TL); return on total assets (EBIT/TA); revenue to total assets (REV/TA) 16. Chen (2012) Profit margin; return on assets; after-tax rate of return; operating profit to paid-in capital; pre-tax net profit to paid-in capital; earnings per share; operating margin; operating profit; growth rate; after-tax net profit growth rate; revenue growth rate; growth rate of total assets; growth in the total return on assets; equity ratio; debt to assets; long-term funds to fixed assets; dependence on borrowing; inventory turnover ratio; receivable turnover ratio; total assets turnover ratio; fixed assets turnover ratio; net worth turnover ratio; current ratio; acid-test ratio  (2015) Quick ratio; net working capital to total assets; current assets to net assets; total liabilities to net worth; retained earnings to sales; debt ratio; times interest earned; revenues to net working capital; accounts receivable turnover; accounts payable; sales to net worth; quality of inventory; fixed assets to net worth; turnover of total assets; revenues to fixed assets; return on assets (roa); return on equity (roe); return on sales (ros); profits to net working capital Liquidity, leverage, activity, and profitability 96.0 28. Tserng et al. (2015) Return on equity; return on sales; profitability including return on assets; profits to net working capital; debt ratio;, net working capital to total assets; retained earnings to sales; current assets to net assets Profitability, leverage, leverage group.

84.8
Notes: Management efficiency include asset utilization, activity ratio, working capital utilization. Solvency ratio is the same as leverage.
Growth ratios are a form of trend ratios.
variables were correctly categorized by the co-authors with accounting background who made use of the accounting literature. The frequency of use of each factor by study is plotted on the chart in Figure 3. Figure 4 presents the 'average accuracy by factor' plot of the CI-IPM of studies that employed each factor. The cash flow and interest coverage had too little data to give fair comparative results. For example, only two studies used the cash flow factor and only one of these two provided its accuracy result which was 96.9%; using this figure will clearly lead to unfair advantage for the factor. The factors are ranked in Table 3 according to the charts in Figures 3 and 4. In both cases, profitability, liquidity, leverage, management efficiency and trend factors occupy the first five positons but in alternating ways. This gives an indication  that these factors are important based on usage frequency and accuracy of CI-IPMs. However, the use of the trend factor is well below 50% (Figure 3) among CI-IPM studies; hence its level of importance is doubtful on the frequency scale.
To validate the results from the analysis of the systematic review, the questionnaire responses were analysed. As advised by numerous social scientist (Spector 1992 The main aim of the test is to check whether the factors and their associated Likert scale are actually measuring the construct they were intended to measure, which is the level of importance of the factors in relation to the insolvency of construction firms, by checking the consistency of the data. The value of Cronbach's alpha coefficient ranges from 0 to 1 and, as a rule of thumb, George and Mallery (2003) suggested 0.7 as the lowest score and 0.8 as an indication of good internal consistency. The SPSS (Statistical Package for Social Sciences) computer package was used to calculate the Cronbach's alpha coefficient. The results are presented in Table 4. A score of 0.149 was achieved, depicting a very low consistency and reliability of the questionnaire responses. To examine the data and establish if there are some factors in particular that led to the poor result, the third column of Table 4 titled 'Cronbach's alpha if item deleted' was inspected. According to Field (2005), if a factor is reducing/worsening the overall reliability and consistency of the data, and therefore is not a good measure of the construct, its associated Cronbach's alpha coefficient would be higher than the overall coefficient (0.149). From Table 4, trend, interest coverage and turnover factors have higher associated Cronbach's alpha coefficients. What this implies is that there is no consistency in the responses given to these factors in the questionnaires. Simply put, the respondents are far from a consensus on whether these factors contribute greatly to the insolvency of  construction firms or not. This portends that these factors are not important in measuring this construct and should be removed. After removing these three factors, the Cronbach's alpha coefficient jumped to 0.874. This means data for the remaining factors have a high consistency and reliability and do actually measure the construct. None of the remaining factors had an associated Cronbach's alpha coefficient greater than the overall Cronbach's alpha coefficient (0.874).
In order to measure the respondents' perceptions of the importance of each factor in predicting the insolvency of construction firms, a significance index score was calculated using the formula below. The equation was derived from similar formulae computed in previous construction studies (e.g. Kometa et al. 1994;Spillane et al. 2011b;Oyedele 2013). The significance index is SGI D X N n D 1 S n NS 2 4 3 5 £100% (2) Where the S in S n represents the significance/importance rating from 1 to 5 given by the n th respondent; n D 1, 2, 3, 4, 5… N; N is the total number of respondents for that particular factor; and S is the highest possible significance/importance rating, which is 5. The sixth column in Table 4 shows the SGI values for each factor while the last column shows the ranking of the factor based on the SGI values.
The three factors with unreliable questionnaire data (i.e. Trend, Interest coverage and Turnover), also happen to have the least significance and are thus not very important for use in CI-IPMs. This validates the review analysis result that Interest coverage and Turnover are not important but discredits the idea that the trend factor is important. The case of the trend factor is not too surprising as it has already been shown under the review analysis that its importance is doubtful because its frequency of use is below 50%. The profitability, liquidity, leverage and management efficiency are confirmed to be important as they are in the top ranks of 1 to 5 with over 60% SGI value each. The main surprise factor here is the cash flow. It ranks second with a reliable data and SGI value of 77%; this means many practitioners agree that this is a very important factor that influences the insolvency of a construction firm even though it has not been frequently employed by CI-IPM developers. This will be discussed further in the results section.

Analysis of qualitative factors
The summary of findings from the systematic review of qualitative factors is presented in Table 5. The qualitative variables used in all the primary (i.e. systematically reviewed) studies are presented as well as the factors/categories the variables fall under. The factors were correctly categorized according to what is popular in construction management literature. The frequency of use of each factor by study is presented in Figure 5. Since most of the primary studies did not build a CI-IPM, an average accuracy chart was not provided here. None of the factors was present in up to 50% of the studies; hence the chart in Figure 5 was plotted according to the actual frequency (first bars in the chart), and based on the most used factor being considered as 100% frequency of use (second bars in the chart). The discussion here is based on the second bars in the chart.
The factors are ranked in Table 6 according to the chart ( Figure 5). Based on the second bars in the chart, only the first six factors out of ten had above 50% frequency of use and are considered to be the most important according to this simple analysis. They are management decision making, firm characteristics, management/owner characteristics, internal strategy, macroeconomic and skill of workforce factors, in that order. Of the six, only the skill of workforce (57.1%) had a percentage below 70%. It should also be noted that of the remaining four factors, only the external strategy factor (42.9%) is quite close to the 50% mark. Also, it was obvious that the sustainability and health and safety factors would achieve a low frequency rating right from the methodology stage since an extra effort had to be made to find just one study that used them. They are thus excluded from the ranking in Table 6. The questionnaire response analysis presented next can however shed more light on their importance.
To validate the results from the analysis of the systematic review, the responses to the qualitative factors questionnaire were analysed in the same way as the quantitative factor responses. The results of the analysis are presented in Table 7. The Cronbach's alpha coefficient of the data was -0.946, indicating very inconsistent and highly unreliable data. Since many factors had an associated Cronbach's alpha that was higher than the overall coefficient, factors were removed one at a time until an acceptable or good Cronbach's alpha was achieved. The factor with the highest associated Cronbach's alpha was removed in each case and the analysis was rerun.
By the time the Cronbach's alpha coefficient reached the acceptable figure of 0.755, skill of workforce, health and safety, motivation and external strategy factors had been removed. The skill of workforce factor, which was noted in the review analysis to be the only factor categorized as being important but yet with a frequency of use value below 70%, had  the least reliable data here in questionnaire analysis. It also ranked nine out of ten factors with an SGI value below 50%; hence its 'important' status was inaccurate. The result for the other three factors is just a validation of their 'unimportant' status as realized from the review analysis. At the acceptable Cronbach's alpha coefficient of 0.755, the sustainability factor (0.862) still possessed a higher associated Cronbach's alpha coefficient. This means the sustainability factor is reliable, but only 'just', and does not contribute to the overall reliability (Field 2005); its removal led to a better Cronbach's alpha coefficient of 0.862, which can be considered good. At this point, only internal strategy (0.863) had a higher associated Cronbach's alpha; however the difference was negligible (0.863 À 0.862 D 0.001) and the data for all other factors (inclusive of internal strategy factor) were very consistent and reliable.
The management/owner characteristics, internal strategy, management decision making, firm characteristics and macroeconomic factors are confirmed in this result as being very important as they rank first to fifth, in that order, and all have an SGI score above 75%. Although the external strategy factor has an SGI score above 50% and ranked sixth, next to the aforementioned factors, its data is not reliable. The case of the sustainability factor, which similarly has a SGI score above 50% but with a contentious data reliability, will be discussed further in the results section.

Result of analysis of the quantitative factors
From the two major analyses done, it is clear that the profitability, liquidity, leverage and management efficiency factors are very important to predicting the insolvency of construction firms. However, as against the review analysis, the  questionnaire data analysis shows the cash flow factor to be very important as it ranked second with an SGI score of 77% using reliable data. This is a result from industry experts who have dealt with the accounts of multiple insolvent construction firms, especially during the period they are going into administration and hence has high validity. The verdict here is that CI-IPM studies need to consider the cash flow factor if they are to build a sound model. The importance of cash flow management in ensuring the survival of construction firms has been highlighted by many construction management non-CI-IPM studies (Robinson & Maguire 2001;Arslan et al. 2006;Holt 2013 among others). Recall that the only primary study that used a cash flow factor and presented its result had a CI-IPM with an accuracy result of 96.9%. One reason the cash flow factor has not been commonly used is because cash flow variables (i.e. financial ratios) are not very common in financial statements of firms. The additional task that CI-IPM developers will need to take on is to calculate cash flow ratios from available ratios in the financial statements. The five important quantitative factors are briefly explained below.

Liquidity factor
Liquidity is an important factor which interests a lot of construction firms' stakeholders such as material suppliers, site employees and staff in general since it shows to what extent a firm can meet its commitments without 'liquidating the nonliquid assets' (Ng et al. 2011;Horta et al. 2012;Horta & Camanho 2013); inability to cover such liabilities generally leads to insolvency. The more liquid a construction firm is, the healthier it is likely to be (Edum-Fotwe et al. 1996). Liquidity might be poor for early warning systems (Bilderbeek 1977, cited by Altman 1984) but is very good for near-immediate and immediate predictions. A fairly high liquidity level is very important for construction firms as cash availability is vital for execution of construction projects.

Cash flow
A construction firm is substantially reliant upon the success of its construction projects; hence for a construction firm to be more solvent, a reasonable size of the firm's cash flow should be employed in operations with a reduced cash flow in investment (Arditi et al. 2000;Enshassi et al. 2006;Chen 2012). This is because of the cash flow conditions of firms in the CI where: Client only pays for completed work that has been financed by the firm, usually on a monthly basis. A percentage (normally 10%) of payment is held back by the client for potential omissions and/or defects.
It is thus almost impossible for firms to recover expenses, not to mention make a profit, before completion of projects. A robust cash flow plan for operations is thus necessary to avoid extreme leverage, being cash strapped or having a negative cash flow, all of which risk the survival of the construction firm (Kale & Arditi 1999). The challenge is to achieve a positive cash flow from project(s), since a negative cash flow increases the risk to its survival. Notes: Higher associated Cronbach's alpha in italics. Ã D yes, but only just. Such reliability can be challenged (Lance et al. 2006).

Management efficiency
The management efficiency factor, measured by asset utilization, activity ratio, working capital utilization ratio, etc. are used to check how efficient a management team is in using a firm's assets and leverage (Edum-Fotwe et al. 1996;Ng et al. 2011;Bal et al. 2013). The CI is characterized by heavy operating expenses which become especially problematic as firms 'need to shrink and expand in cycle with the job market and competitive conditions' (Arditi et al. 2000); improper management in this situation can lead to insolvency. Activity ratios are more concerned with management's ability to turn a firm's assets into cash (Ng et al. 2011). This can help to reduce the possibility of insolvency that could result from liquidity problems.

Leverage
As opposed to liquidity, leverage ratios measure long term solvency and thus contribute greatly to early warning systems for the CI (Horta et al. 2012). Because construction works are normally paid for only when specific segments have been completed, usually on a monthly basis or even after longer periods when works are delayed, construction contractors are exposed to high debt (leverage) as they are typically required to pay subcontractors and suppliers; these debts make construction firms more susceptible to failure (Arditi et al. 2000).

Profitability
According to Arditi et al. (2000), the single most common budgetary factor that leads to the failure of construction firms is insufficient profit. This is because of extremely aggressive bidding, with far from accurate estimates and the one-off and custom-made production systems that are synonymous with the CI. Ideally, the higher the profitability of a construction firm the more solvent it is taken to be. However, developers using the multi-discriminant analysis (MDA) technique to develop CI-IPM need to be careful as the technique sometimes wrongly assigns a negative sign to the profitability ratio (see Mason & Harris 1979;Abidali & Harris 1995). This problem is commonly known as the counterintuitive sign problem.

Result of analysis of the quantitative factors
The verified most important factors from the two analyses are management/owner characteristics, internal strategy, management decision making, firm characteristics and macroeconomic factors. From the results of the analysis, the labelling of the sustainability factor as important or not breeds controversy, with questionnaire data of 'acceptable' reliability and an average SGI score of 54%. Tan et al.'s (2011, p. 229) 'comprehensive review of studies on the relationship between sustainability performance and business competitiveness finds that there is no unique relationship between the two variables'. This, according to Wagner and Schaltegger (2003), is due to lack of data. However competitiveness and (in)solvency are not even exactly the same thing, although they are highly correlated. The verdict here is that sustainability is an important factor to consider for CI-IPM developers but not as important as the aforementioned factors. The external strategic factor has an even higher SGI score compared to sustainability, albeit with unreliable data. The unreliability makes it hard to consider it a very important factor. The identified important qualitative factors are briefly explained below.

Management/owner characteristics (MOC)
Certain MOCs of a construction firm have negative effects on its solvency. These include inertia, unfounded optimism, taking inappropriate risks with relatively large construction projects, autocracy of managers/CEO/president, a person holding multiple executive positions, an executive with too much power, etc. (Abidali & Harris 1995). Autocracy leads the race in this factor and is synonymous with an executive with too much power or a person holding multiple executive positions, all of which cause failure of construction firms. A very powerful dual-position CEO/chairman, nullifying the all-important managerial power of the chairman being able to sack a deficient CEO, is a common feature of failed construction firms (Hall 1994). In contrast, a balanced board which efficiently controls managers' actions help improve a company's solvency. The inertia of a construction company's owner/management leads to a lack of awareness of the opportunities and threats to the business (Gilbert 2005). When business is slow, a construction firm specialized in pile foundation installation, for example, should be able to identify opportunities of excavation projects and make use of its excavators.

Internal strategic factors
The inclusion of internal strategic factors for developing CI-IPM is vital if a robust CI-IPM is to be achieved (Henricsson et al. 2004;Dangerfield et al. 2010). Key strategic factors, according to Arditi et al. (2000), include sales/bids, competitiveness, planning etc., all which are based on the adaptability of a firm. The more successful bids a construction firm makes, the more it grows and the more solvent it becomes; lack of successful bids is tantamount to failure (Bal et al. 2013). Bidding in an area of expertise ensures a competitive bid; thus a firm must have an, or identify its, area of strength where it has an advantage over competitors. The importance of competitiveness cannot be overemphasized and efforts have been made to measure it in the CI (Henricsson et al. 2004;Dangerfield et al. 2010) in order to establish the state of solvency of a firm. Having the correct knowledge of itself and competitors can help a construction firm in designing the right strategy.

Management decision making
This factor is usually a result of MOC and directly influences internal strategy. However, the resources at the firm's disposal and some other elements do affect it. Decisions on project should be based on what is best for the firm rather than on ego, friendship etc. For example, project selection should be based on what the firm is comfortable with and should be, as much as possible, limited to a familiar geographic area to keep detrimental surprises to a minimum. Taking on a project a long distance away can lead to managing from a distance, procuring and engaging unfamiliar subcontractors of unknown quality and running into unexpected geological conditions (Denyer & Tranfield 2006). Generally, construction firm managers that go through the firm's financial statement carefully before making decisions have been more successful (Hall 1994).

Firm characteristics
Firm characteristics such as size, age, experience, maturity, flexibility, etc. can have a reasonable effect on a firm's solvency (Ng et al. 2011;Bal et al. 2013). Age is the most important of these because it has been proven that a lot of young firms fail due to their newness (Kale & Arditi 1999). The possibility of a construction firm piling up business knowledge and skills through organizational learning is largely dependent on its age (Arditi et al. 2000). Such learning over time, and the resulting knowledge and skills, help a construction firm to identify favourable markets, create a positive image, establish important partnerships with construction materials suppliers and subcontractors, build positive relationships with financial institutions and potential clients, and adapt easily to the latest technologies (March 1991), among others; the combined absence of these factors can lead to a firm's failure. The ease of measuring the age of sample firms in months or years enables a CI-IPM developer to include this factor.

Macroeconomic
Macroeconomic factors include the amount of construction activities by existing firms, the number of available construction contracts in a country at any time, interest rate, industry weakness, and the threat of new entrants, and are considered among the most important insolvency factors for developing IPM for the CI (Arditi et al. 2000;Sang et al. 2013). Construction firms are highly susceptible to macroeconomic effects. However, the susceptibility level of each construction firm differs (Ng et al. 2011;Sang et al. 2013). Industry weakness is not important when only one industry is being considered, as in the case of CI-IPMs. Kangari (1988) suggested the 'construction-contract valuation index by F.W. Dodge' as a measure for construction activity in the US, while 'The Construction Index' can be used to measure the number of new businesses in the industry in the UK.

Sustainability
The effect of sustainability on the solvency of construction firms is largely dependent on government legislation and environmental standards as these can help to bring about innovations that lower cost and improve value. This will make a firm more competitive. Practising sustainable construction will also improve the image of a firm and qualify it to bid for contracts with strict sustainability requirements. However there are only a few such projects and many, especially less wealthy, owners will put cost before sustainability. This is probably why it is not too directly linked to the insolvency of construction firms.

Discussion and framework
The five most important quantitative factors (i.e. profitability, liquidity, leverage, management efficiency and cash flow), one way or the other À all deal with sufficient availability of cash (for projects). This is not surprising since the CI is operations based and construction firms generally tend to take on projects involving costs that exceed their financial worth or equity. Firms will thus need all the money they can get to keep a project(s) running before the client pays for the completed portion according to the contract terms. Without enough cash to run projects, a construction firm can easily become insolvent. This agrees with the literature; Chen (2012) identified that construction firms must allocate more cash to operations than securing assets to avoid project failure because a single project failure can result in insolvency.
The most important insolvency factors measured by qualitative variables include management/owner characteristics (MOC), internal strategy, management decision making (MDM), firm characteristics, macroeconomic and sustainability factors. The high importance of managerial factors is evidenced in the emergence of two management-related factors in the result. This, along with the internal strategic factor ranking, supports Jennings and Beaver's (1995) assertion that the major cause of company failure is almost invariably poor management or lack of management attention to strategic issues. Together with the macroeconomic (external) factors, they corroborate Mahamid's (2012) findings as the most important agents for the survival of construction contractors.
Without inclusion of any of these factors, important circumstances that lead to failure of construction businesses, as in Porter's five forces and Mintzberg's five Ps of strategy (among other failure-related theories), cannot be measured/represented in a CI-IPM, making such CI-IPMs suboptimal. This fact is in line with many studies (Arditi et al. 2000;Koksal & Arditi 2004;Horta & Camanho 2013;Alaka et al. 2015, among others). Further, these factors cut well across all the levels in the construction world hierarchy (Figure 6), making them more exhaustive.
On the industry level of the construction world hierarchy the threat of new entrants (macroeconomic factor), as in Porter's theory, is a big problem in the CI because there is almost no requirement for new entrants. This normally results in influx and fierce competition, leading to high firms-to-contract ratios and consequently a high failure rate for firms. It is well known that the older or more established a firm is, the less susceptible it becomes to threats from new entrants (Hill et al. 2014). The 'age' element of the firm characteristics factor and 'managers' experience' element of the MOC can be used to take care of this area in a CI-IPM.
On the organizational level, the construction material suppliers' power (Porter's theory) is quite low in the construction industry because of the aggressiveness in the suppliers market (Muya et al. 1997), resulting in low prices for materials. However, the high level of competition in supplier selection is starting to be seen as a driver for negative effects on established supply chain relationships. Good relationships are known to improve prices, delivery time, supply preference etc. for the construction firm because of the opportunity of repeat business. The level of 'business knowledge', which is an element of MOC, and the internal strategy, are known to affect supply chain relationships and can thus be used to represent this area. A poor strategy would be to consistently buy randomly from the cheapest supplier rather than to cultivate better relationships which can, for example, lead to the supplier giving the firm preferential treatment during materials scarcities.
The strategic pattern (Mintzberg's strategy theory) which results from managers' experience can only be measured with qualitative variables like 'construction managers experience' (MDM). Other Mintzberg's Ps of strategy which are known to be key to the survival of firms include plan, ploy, position and perspective. Strategy as a position is a matter of where in a market a construction firm concentrates (e.g. new build, homes, pavement construction, renovations etc.) (Mintzberg 2003) and can be represented by 'company's main activity' (firm characteristic). Strategy as a plan and perspective can be represented with elements like 'emphasis on innovation, and geographic location of headquarters' under internal strategy and firm characteristics respectively (Mintzberg 2003).
On the project level, employing 'skilful workers' and 'highly experienced foremen' (MDM), can affect the duration and cost of projects, which are both major factors in deciding the Porter's competitive rivalry level of construction firms (Shash 1993). Also, 'emphasis on innovation (internal strategy)' can measure how flexible a construction firm has been to adopting/creating innovating techniques for executing construction projects. For instance, modular construction is what currently reigns in London and any firm not adopting this method faces a high threat of substitution from clients, as in Porter's theory. 'General construction experience of owner/CEO', 'level of managerial experience in the CI' and 'education level of owner/CEO' (MOC) all represent the individual level in the construction world hierarchy Basically, it is clear that the factors given in this study cut right across the construction world hierarchy and address most of the areas highlighted in business failure/survival-related theories. Although CI-IPMs built solely on financial or quantitative variables do work, they do not really predict/foresee the failure of construction firms. Rather they only reveal a company that is already failingwhich might not leave enough time for remedy. It is the factors measured with qualitative variables (strategic, MOC, management, etc.) that can actually predict the potential failure of a construction firm even when it is healthy, since they consider the actions and characteristics of a firm; in fact financial variables are only the result of the actions of a construction firm, such as MOC, strategic, management, as well proven in the literature (Arditi et al. 2000;Koksal & Arditi 2004;Horta et al. 2012;Horta & Camanho 2013;Alaka et al. 2015 among others). Both sets of factors are thus key to developing a robust CI-IPM. This implies that considering all the important factors provided in this study's framework (Figure 7) in developing a CI-IPM will definitely result in a more accurate, more reliable and especially more valid early holistic prediction model as virtually all key areas that can lead to the failure of construction firms would have been considered. This framework (Figure 7) will benefit future CI-IPM researchers by providing an initial platform from which the important factors and variables affecting construction firms' insolvency can be selected, as omission of such important factors can result in a poor CI-IPM. The practical implication of this study is that, as it makes the most important quantitative and qualitative factors for CI-IPM readily available, researchers in the CI-IPM area of study will increase the use of qualitative factors in tandem with quantitative factors in order to build much better CI-IPMs, having recognized that early insolvency predictions cannot be achieved without them. This is because the challenge of qualitative factors and variables not being readily available for CI-IPM is solved in this study. The study will also reduce the time spent on the statistical analysis of very many factor variables for the purpose of selecting the best ones since the search can be narrowed down to the variables of the important factors presented in the framework. Further, this study will ensure that no important factor (e.g. the frequently unconsidered/unused cash flow) is left out when building a CI-IPM.

Conclusion
Many IPMs have been developed for the CI but most of them have used solely quantitative (financial) insolvency factors simply because they are readily available. Unfortunately, these have led to non-robust models as they miss out some important CI insolvency factors that cannot be measured with financial/quantitative variables. In fact, the positions of financial factors are only a result of qualitative factors (e.g. managerial, strategic, macroeconomic, etc.) and hence early prediction of the insolvency of construction firms largely depends on these factors. This study set out to create a comprehensive theoretical framework that will form the platform for selection of vital CI insolvency factors and explain their relative importance in relation to the solvency of construction businesses.
The study used the systematic literature review research strategy, triangulated with questionnaire data to create the theoretical framework. The framework highlighted the most important quantitative and qualitative factors. Results showed profitability, liquidity, leverage, management efficiency and cash flow to be the most important quantitative factors. Though not common in the reviewed studies, cash flow is of vital importance to the survival of construction firms and must be adequately represented on its own in any valuable CI-IPM. Results also showed management/owner characteristics, internal strategy, management decision making, firm characteristics and macroeconomic factors along with sustainability to be the most important qualitative factors.
The study clearly showed that the highlighted factors cut across the entire construction world hierarchy and are in line with firm insolvency/failure-related theories like Porter's five forces and Mintzberg's five Ps of strategy, making them more significant to developing credible and valid holistic CI-IPM. That is in addition to their effect on early insolvency prediction, which will allow time for implementation of remedies. Overall, this study proposes the use of qualitative factors, alongside quantitative factors, having shown their (i.e. qualitative factors) acute necessity and having partly solved their not being readily available challenge.
One limitation of this study is that the best variables for measuring the highlighted factors could not be established because virtually every past study pointed to different variables as being the best representative of a factor. Future studies should thus focus on establishing these best variables. Future studies should also make an effort to identify more qualitative variables so that the problem of lack of availability could be resolved. This will benefit researchers who prefer to have a pool of variables to analyse statistically rather than accept established best variables. Further, for assessment purposes, future studies should attempt to implement the use of highlighted factors in developing their CI-IPMs.

Disclosure statement
No potential conflict of interest was reported by the authors.