Paper submitted to the Journal of Financial Management of Property and Construction

Insolvency in the UK construction industry

Author

Abstract

This paper is an analytical and empirical study of insolvency in the UK construction industry. This is of interest because of the apparently disproportionate share of insolvency affecting UK construction. A number of hypotheses are generate and tested. The principle findings of the study are that cash flow factors appear to be largely responsible for the level of insolvency in construction insolvency. Notable variables appear to borrowing by construction firms, availability of credit and domino effects from general insolvency. Also, while profitability appears to have a measurable impact on insolvency, it is less important than cash flow factors. Fluctuations in construction output and the cost of credit appear to have no significant impact on insolvency.

A robust predictive regression model of construction insolvency is generated using quarterly empirical data for construction insolvency in England and Wales over the period 1969 to 1994 against a variety of independent variables.

Keywords

Insolvency, Company liquidations, Construction, Cash flow, Working capital.

Background

Insolvency statistics

Of all company liquidations and receiving orders for the self-employed, a disproportionate number stem from the construction industry. In the case of the self-employed, over the past twenty five years, the share of bankruptcy attributable to construction varied from 21% to 30%. Similarly for company liquidations, over the same period, between 12% and 22% of which were for construction firms. This comes from an industry producing between 5% and 7½% of Gross Domestic Product.

Presentation of data

Hillebrandt (1984) pointed out that construction insolvency statistics appear to be high because they are not subdivided. There are twenty categories in the official Department of Trade and Industry statistics including seven categories for manufacturing and four for distribution and transport. Construction, by contrast, appears as a single entry. It consequently stands out prominently in the official statistics.

The insolvency statistics for England and Wales as presented in Table No 1 for companies in receivership and Table No 2 for individual bankruptcy are aggregated into six broad categories to eliminate this perception. The same information is presented graphically in Figure No 1 and Figure No 2 respectively. These figures show that insolvency is construction is not far below that registered in the much larger manufacturing and distribution sectors.

Failure rate in construction

A fairer comparison might take into account the large number of firms operating in the construction sector. Clearly, the more firms involved in a particular sector, the greater the likelihood of high failure rates. In particular the number of small firms and self-employed operating in construction makes it particularly vulnerable.

By this criterion, the peak failure rate for the self-employed and companies in the construction sector appears to be twice as bad as for other industries. In the case of 1975, when the construction industry was suffering in a major recession, 3.4% of all construction firms were liquidated compared to an overall rate of 1.8%. In addition, 0.5% of the self-employed in construction had receiving orders administered as compared to an overall rate of 0.25% (Hillebrandt, 1977). It should be pointed out that these failure rates tended to fall towards the average as the economy and the industry recovered (Hillebrandt, 1984).

Motivation for the study

These figures in themselves are ample justification for this study. Not only does construction appear to have high absolute figures for insolvency but it also appears to have a share disproportionate with its position in the economy. While the position might not be quite as bad as initial observation would suggest, there is still a case to answer.

Objective of this paper

The objective of this paper is to identify the main factors responsible for the apparently high levels of insolvency in the construction industry in the UK. The analysis commences with a theoretical review to generate a range of hypotheses which are tested by an empirical study using available published data from official sources. In addition to the hypothesis testing, it is intended to generate a predictive regression model of construction insolvency.

The study concentrates on the construction sector and the reasons for the current level of insolvency. It is not intended as a comparative analysis of construction as against agriculture or manufacturing.

The process of insolvency

Unprofitability

The most obvious immediate cause of insolvency is a failure to achieve a desired rate of return on capital invested. This corresponds to ‘normal profit’ from economic theory. That is the minimum return necessary to keep the capital in place and satisfy the shareholders. Much of construction work is highly competitive given the large number of firms involved in the UK industry. There are few barriers to entry for new firms or firms diversifying from other areas. It is an established feature of a competitive economy for free entry to be matched by free exit. Hence, it should be no surprise if a number of inefficient firms who are unable to compete on price and/or quality fail, except when demand is exceptionally buoyant. This will be of concern only if generally efficient companies become insolvent through no fault of their own.

Cash flow problems

Cash flow problems and shortage of working capital will inevitably hasten the downfall of an unprofitable and inefficient outfit. They can also, in extreme circumstances, push generally efficient and profitable firms into insolvency. Construction projects are of long duration and, consequently, a system of interim payments has evolved. Such interim payments are usually lag considerably behind progress and are inevitably subject to retention. More cash can be tied up in unresolved claims and variations. Thus a contractor will require considerable working capital in the form of credit or overdraft facilities up to the point where a project becomes self-financing.

A combination of delays in payment and impatience from creditors can lead to insolvency if the bank is unwilling to bridge the gap. This will only apply if the contractor has inadequate working capital to overcome cash flow difficulties. Many construction firms are highly geared (reliant on borrowed capital). In particular, a number of small firms have poor credit ratings and also have little in the way of fixed capital assets to act as collateral for bank loans.

Domino theory

It is also possible for a firm to be dragged into insolvency by the failure of another firm. The domino theory may apply if a client becomes insolvent owing large sums of money to the contractor or if a main contractor fails owing cash to one or more regular subcontractors. The chance of this occurring will depend upon the amount owing and the availability of working capital. As a firm approaches insolvency, it is likely that it will delay payments to creditors. Thus, the amount owed by an insolvent firm will probably be much higher than the average deficit for firms in general. The payment system should place limits on the extent of debt from the client to the contractor. In the case of the main contractor and subcontractor relationship, delays in payment are more likely and the risk is higher.

Credit availability and cost

Cash flow problems can become critical if credit is difficult or expensive to provide. This will arise due to the monetary strategy employed by the government. In the past, UK governments have used credit squeeze as a mechanism to restrict monetary growth without the need for punitive interest rates. This was accomplished by general restrictions on credit or by the Bank of England instructing the clearing banks to make special deposits and hence to restrict lending to their customers according to a multiplier formula. This forced bank managers to call in loans and overdrafts as well as refusing new loans in order to comply.

It is possible that credit problems triggered the increase in insolvency experienced by construction in the mid-1970s. The difficulty of obtaining credit forced the builders’ merchants to abandon their traditional rôle as the builders’ banker. This approach was effectively abandoned by the Conservative government in 1980 when controls on the movements of capital were removed. This meant that foreign based banks would be able to undermine any attempt to restrict the availability of domestic credit other than via higher interest rates.

Accordingly, the availability of credit would only be an issue if the company was perceived by the bank as a poor risk. Because interest rates effectively became the only instrument of monetary policy, emphasis shifted to the cost instead of the availability of credit.

Analysis of insolvency in construction

Overview

The above mechanisms do not explain why construction insolvency appears to be proportionately higher than for other sectors of the economy. Indeed on the basis of the key indicators, insolvency ought to be lower than other industries.

• Capital productivity measured in terms of return on capital invested is much higher than for any other sector (Lowe, 1993). Over the period 1948 to 1994, return on capital invested in construction averaged around 24% as compared to 9.5% for manufacturing and 8.3% for agriculture. Even for 1991, the peak year for construction insolvency, the return was nearly 12.5% as compared to 7% for distribution & transport, the next best performing sector.

• Cash flow, on a superficial level, presents more of a problem for construction. The per capita requirements for working capital appears to be no higher than for manufacturing and less than for agriculture. However, it should be pointed out that the ratio of working capital to fixed capital is much higher for construction than for certain other sectors. Thus 60% of all capital requirements for construction are for working capital as compared to 25% for manufacturing and 3% for energy and water supply.

• Construction companies face similar problems to those from other sectors in obtaining credit. It can be argued that the preponderance of small firms in construction could exacerbate the problem to a limited extent. This is reflected in the statistics for receiving orders in construction.

If an explanation is to be found for the high incidence of insolvency in construction, it is necessary to identify the distinctive features of construction. These concern issues such as the impact of fluctuating demand, the method of tendering and the management of the industry.

Fluctuating demand

It has often been cited that construction suffers from fluctuations in demand. This may arise due to some sort of accelerator mechanism or because of the effect of government policy. Many commentators, for example Lean & Goodall (1966) and NEDO (1975), have argued that construction was used by successive UK governments as an economic regulator. This is liable to affect public sector projects and commercial and speculative developments.

If the volume of work, in a given sector of construction, falls, it is likely that the turnover of firms operating in that sector will fall. This could create problems in both profitability and cash flow if the firm had high fixed costs since there would be less revenue available to cover this. Adjustment by the firms to the lower level of demand could prove costly. Alternatively, a firm may be persuaded by high profit margins during the boom period of the cycle to expand beyond the limits imposed by their managerial competence. This would lead to inefficiency, which is likely to be masked during the period of high demand. As the period of slump arrives, the inefficiency would be exposed and insolvency could follow unless the management of the firm is improved quickly. Finally, any firm engaged in speculative development failing to read economic trends could be left with a large volume of unsold units. This could lead to losses if the units were sold off cheaply or to problems of cash flow and debt charges if they were retained in the hope of a market recovery.

This explanation is superficially very attractive. However, is should be treated with some caution since there is no evidence that firms involved in aspects of work prone to fluctuating demand have a worse insolvency record than other companies. Those most affected include the larger general builders and civil engineering contractors and the speculative specialists. Indeed, it can be argued that this category is less prone to insolvency. Insolvency, with a few notable exceptions, appears to affect the smaller contractors and subcontractors who are more likely to be engaged on repair and maintenance work and on smaller new developments. Hillebrandt (1984) suggests that receiving orders issued to the self employed in construction is inversely related to the volume of repair and maintenance work.

Clearly, while it would be unwise to rule out fluctuations as a cause of construction insolvency, it does not provide all the answers.

Tendering and pricing policies

The widespread use of competitive tendering in construction may have an impact on the insolvency statistics. This leads to considerable abortive work on unsuccessful tenders and increases contractors’ overheads. It also magnifies the problem of fluctuating demand for the individual firm.

Apart from those operating in the speculative side of the industry, the tendering system limits the degree that companies can control their own turnover and destiny. Thus a construction firm is dependent upon the client or the design team to include them on the short-list for a given project and then they face competition to win the contract. In the 1960s, it was assumed that the probability of winning a project under competitive tender was between one in four and one in six (Lansley et al, 1974). It is likely that the odds are much longer, now possibly up to one in twenty for certain contracts.

The approach to letting projects makes it more difficult for builders to employ market pricing on contractual work. The approach lends itself to cost-plus-profits pricing rather than charging what the market will stand. The sealed-bid tendering system restricts information and precludes the formation of a market price. By contrast, the speculative sector employs market pricing throughout. In the case of speculative housebuilding, there is a highly advanced market in operation. There is extensive knowledge of house prices for given locations. The speculative house builder can control output to balance demand in the same way as a manufacturer. Many speculative house builders use discounts and special offers such as subsidized mortgages and free carpets and curtains to attract buyers when necessary. Advertisements can also be used to influence the market and increase demand.

One development from the early 1980s has been the spread of tendering as a means of letting work to new areas such as local government services and office cleaning. Prior to this initiative, such services were usually allocated to direct labour. A marked increase in insolvency within this sector of services through the 1980s and 1990s might support this particular thesis.

Management problems

This encompasses all factors within the control of and specific to the individual company. By contrast the issues raised above relate to the whole construction sector. Construction has had a reputation for poor quality management as compared to other industries. This can apply at a number of different levels of organization:

• Management on site covers such areas as lack of control operations, poor control of labour and materials, and under utilization of plant. It also encompasses inadequate bookkeeping, lack of site records and photographs, and non-identification of potential claims. While any contractor can suffer from these problems, the successful firms have evolved accounting systems incorporating early warning signals to alert senior management. A company without internal rigorous accounting procedures may fail to notice that a cumulative problem exists until it is too late to do anything about it. If this is repeated on a number of projects, this could be very serious.

• Management at the head or regional or local office is concerned with issues such as estimating, planning and purchasing. Clearly, failings in the estimating department will be disastrous and could lead to either declining turnover and falling profits or expanding workload with very tight margins and cash flow difficulties. In addition, lack of control by the office-based management over operations on site or inadequate backup to the site-based management could contribute to the problems raised above.

• Strategic management includes decisions emanating from the board of directors and the senior management. Decisions on which sectors of the market to concentrate and the geographical limits of operations would fit into this category. An example of a poor strategic decision was the move by many contractors to effectively abandon secure but low margin work such as council house maintenance during the upswing in construction activity in the early 1970s. This lead to contractors building up their direct labour departments as a response. When the boom subsequently collapsed in the mid 1970s, many contractors found themselves frozen out of such work for some years to come. Equally, any contractor investing large sums on what subsequently turned out to be a passing fad, for instance, industrialized building, would constitute a bad strategic decision. Finally, falling into the pitfalls associated with fluctuating demand as outlined above could also come into this category.

This is a difficult area to obtain hard information because at best the data will be of commercial sensitivity while at worst it will non existent. Companies failing due to bad management practices are probably unaware as to why they failed since if they had realized the problem existed it could have been corrected.

Experimental design for empirical analysis

Approach to analysis

The empirical part of this paper is based on a statistical time series analysis of insolvency data, as the dependent variable, against independent variables representative of the causal factors outlined above. This will attempt to produce a multiple regression model of construction insolvency in the UK and also to identify the most important causal factors

The dependent variable

A twenty sector industrial analysis of company insolvency and bankruptcy of the self employed is published for England and Wales by the Department of Trade and Industry. Unfortunately, no comparable figures, broken down by industry, are available for Scotland and Northern Ireland. This will present problems for the analysis in that the economic variables are presented for the UK. Thus, the data for England and Wales is used as a proxy for UK insolvency data.

In the case of company insolvencies, this will not be critical since the total number is small for Scotland (less than 3% of the total for Great Britain) and will certainly be even smaller for Northern Ireland. There is, in any event, no reason to believe that the pattern of insolvency will vary throughout the UK. In any event, many of the companies operating in Northern Ireland and Scotland have head offices based in England. This may be part of the explanation for the low corporate insolvency in Scotland.

In the case of individual bankruptcies, there is more concern in that with the different legal system direct comparison is not possible. The number of sequestrations in Scotland were quite high as compared to receiving orders in England and Wales. In 1992, for example, there were 10,845 sequestrations as against 36,703 bankruptcy orders and arrangements. This is out of line with the expected number given the size of the Scottish economy. A change in the legal situation, in April 1993, had the effect of dramatically reducing the number of sequestrations in Scotland. The number of sequestrations remains significant after the legal change.

The analysis consequentially relies on a time series of company liquidations in England and Wales over the period 1969 to 1995 as the dependent variable for the following reasons:

• Company insolvency is more significant than individual bankruptcy. While the numbers concerned are about equal, as illustrated by Figure No 3, companies are of greater economic importance and contribute more turnover.

• Data for company insolvency is published in a uniform basis for a longer period than individual insolvency. Company data is available in annual and quarterly form from 1969 to date, whereas individual insolvency data is only available in a compatible format from 1983 onwards.

• The problems associated with the different legal definitions in Scotland and lack of suitable data, subdivided by industry, makes it desirable to eliminate individual insolvency from the analysis.

• There appears to be a fair correlation between the two data series, with a value of coefficient of determination of over sixty percent. There appears to be some divergence in the mid-1980s with a decline in individual insolvency accompanied by an increase in company insolvency. Otherwise the two series move in the same direction. Thus, the results of the analysis using company liquidations should be representative of all insolvency.

The analysis uses the number of insolvencies rather than failure rate. The argument for failure rate is that if the number of firms is rising we should expect more insolvencies. However, the actual number of insolvencies is a more meaningful variable to model.

The dependent variable is defined as follows:

Construction insolvency = C (Variable 01)

The independent variables

Seven causal factors were identified above. Of these, it is not possible to test for two of the factors using industry-wide statistics.

First, the impact of tendering and pricing policies cannot easily be identified since it is not possible to obtain separate insolvency statistics for contractual firms and speculative firms. No insolvency data is available for contract cleaning companies etc. to test the impact of the spread of tendering to non construction activities.

Second, it is not possible to come up with an adequate test for the impact of management problems on insolvency. While it might be possible to produce a test based on, say, the number of graduates employed in the industry. This would almost certainly give negative results since the greatest influx of graduates in management in UK construction came in recent years as insolvency statistics have grown.

The five remaining factors are dealt with as follows:

Profitability can be represented at industry level by the percentage return on capital invested. This is computed by taking the aggregate profit for the construction industry and dividing by an estimate of the cost expended on fixed and variable capital assets. This variable is defined as follows:

Cash flow can, theoretically, be modelled by use of data on working capital requirements. This identifies the capital tied up in work-in-progress, unsold speculative units, and holdings of materials and components. The data is published annually in the UK National Accounts. This will have to be presented in constant price terms. The use of indices to translate current prices to constant prices can lead to inaccuracy. A better approach might be to use a ratio of working capital divided by construction output with both in current prices. This variable is defined thus:

Unfortunately, in the UK, data on working capital for construction is partially estimated using assumptions of six weeks work. Thus, the ratio of working capital divided by construction output is liable to be fairly constant. As an alternative to working capital, borrowing by construction companies can be used as a proxy for cash flow. Figures for advances from UK and foreign banks to the domestic construction industry are published by the CSO in Financial Statistics. The ratio of bank borrowing divided by construction output will be used as follows:

The domino theory can be modelled using lagged data for company insolvency for all sectors of the economy. This will cover insolvency of both clients and contractors. The source of this data is the Department of Trade and Industry as before:

The variables used and the sources of data are summarized in Table No 3.

Credit availability presents more of a problem to model. The most obvious approach is to take a broad measure money supply such as M3 or Sterling M4. The problem with this for a time series analysis, is the need to adjust for inflation. A better approach is to use a ratio of money supply divided into the gross domestic product for the economy as a whole. This obviates the need to adjust for inflation and removes a source of error. It also sets the money supply in context with the demands placed on it. This ratio equates to the velocity of circulation of money in the economy. Figures are published in Financial Statistics. This is not likely to have a significant impact on insolvency after the demise of credit squeeze.

Money supply velocity V = GNP / M4 (Variable 06)

Where: GNP = Gross national product

M4 = Broad measure of money supply

Cost of credit interest rates could be used as a proxy for the availability of credit. It might be expected that shortages of credit will push up interest rates. Unfortunately, in the UK, prior to the removal of controls on export of capital in the 1980 budget, credit was often controlled administratively without the use of interest rates. Interest rates can, thus, be expected to give unreliable results prior to 1980, but should give improved results after that date. The rate used is the minimum lending from the Bank of England until this was discontinued and the rate used by the major clearing banks thereafter:

Interest rate = R (Variable 07)

Fluctuating demand can be measured in terms of the modulus of the percentage changes in gross output for the construction industry. Long time series of construction output are published in constant price terms in Housing and Construction Statistics. Gross output is preferable to value added for this analysis because it better represents the financial outlay for the contractor. This is defined as follows:

Change in output DQ = ( |Q_t -Q_t-1 | /Q_t ) x 100% (Variable 08)

Where Q_t = Output for construction in quarter t

Hypotheses

Economic theory suggests that insolvency can stem from lack of profitability, cash flow and credit problems or fluctuating demand. Five hypothesis emerge from the theory. Thus construction insolvency could be caused by:

For each of the above variables, a null hypothesis H_Ø that the particular variable has no significant impact on construction insolvency is tested against an alternative hypothesis H₁. If the null hypothesis can be rejected at the appropriate level of significance then the alternative hypothesis, that the variable does have an impact, can be accepted.

Testing of the hypotheses

The impact of each of the above variables on construction insolvency is tested using a multiple regression model. The efficacy of the multiple regression model is measured by the coefficient of determination ( r² ) — the proportion of variation in the dependent variable explained by variations in the independent variables — and also by test of significance using analysis of variance. The regression model will produce an equation of the form:

The tests of the individual hypotheses are accomplished by reference to the individual coefficients in the regression formula. The first check is to see if the sign of the coefficient b _I is the same as the theory would suggest.

The second check is to use use a student’s t-test of to see if the coefficient b _I is significantly different from zero.

The multiple regression model is based on quarterly data from 1969 through to 1994. Only two variables of the eight variables are published on an annual basis and these are transferred to a quarterly format. The statistical analysis was carried out using SPSS for Windows 6.0. The data used is included in the Appendix .

Empirical analysis

Initial regression model

The initial regression model was run with all seven independent variables included. The results were fairly encouraging with coefficient of determination of just over 93%. The analysis of variance F-test produced a high value for F of 184.962. This corresponds to a level of confidence of over 99.9% in the regression model. The individual a and b coefficients and t-values are included in Table No 4. The following conclusions can be drawn from these results as presented in Table No 5 :

Hypothesis I: The results show that return on capital invested has a negative impact on insolvency. In other words higher profitability leads to lower levels of insolvency. This accords with the theory. The level of significance for the coefficient to be not equal to zero is 0.0001. Thus, we can reject the null hypothesis that profitability has no effect on construction insolvency with a level of confidence of 99.99%.

Hypothesis II: There are mixed results in this case although both variables have positive regression coefficients. This suggest that higher requirements for both working capital and borrowing have the effect of increasing insolvency. Again, this might be expected from the theory. Variable No 04, dealing with borrowing, gives highly significant results whereas Variable No 03 gives more marginal results. We can reject the null hypothesis that cash flow has no impact on construction insolvency with over 99.99% confidence using Variable No 04 and with marginally under 95% confidence using Variable No 02.

Hypothesis III: The impact of lagged general insolvency on construction insolvency also has the expected positive coefficient. This has a highly significant t-statistic and the null hypothesis can be rejected with more than 99.99% confidence.

Hypothesis IV: In the case of cost and availability of credit the model produces negative results. Apparently both higher interest rates and tighter monetary policy have the effect of reducing construction insolvency. While the Variable No. 06, velocity of circulation of money, is highly significant, Variable No. 07, interest rate, is rather suspect with a level of confidence of only 46%. We cannot reject a null hypothesis that credit availability and cost has no impact on insolvency from the above test.

Hypothesis V: This test also gave a negative outcome. The regression coefficient is negative implying that the greater the fluctuation in construction output the lower construction insolvency will become! While the level of confidence is only around 80% for the negative sign, the null hypothesis cannot be rejected in this case.

Analysis of initial results

Surprising, for insolvency, a variable considered as a lagging indicator, the power of the model was not enhanced in any case by lagging the other variables with the exception of Variable 05.

The strong positive results for Variables 02, 04, and 05 fit in with the theory and the literature and no further analysis is necessary. The outcomes worthy of investigation are low confidence in Variable 03, 07 and 08, and also the negative outcome, indicated by the ‘wrong’ sign on the regression coefficients for Variable 06, 07, and 08.

Variable 03: The relatively disappointing results in this case can probably be explained in terms of the poor quality of available statistics. The figures include estimates for materials-on-site and for uncompleted and completed but unsold dwellings as well as land banks. This gives data rather biased towards the housing sector. Work-in-progress for non-housing work has not been included since 1985. In addition, the figures when included prior to 1985, appear to be have been based on estimates of the value of, say, six weeks work rather than on surveys of contractors. Thus, by the time these figures are divided by gross construction output to give the working capital ratio, they will have been trivialized almost out of existence. Despite this the results were positive and sufficiently significant to warrant the variable being retained.

Variable 06: The very strong negative results were surprising. It is a reasonable assumption that a tight monetary regime will result in gross national product growing at a faster rate than that of the money supply and hence increase the velocity of circulation of money. A possible explanation for the negative impact of this on insolvency is that credit is the mirror image of borrowing. Hence if higher borrowing increases the debt to companies and hence insolvency, then restriction in credit will reduce borrowing and insolvency. An economy without credit would probably have low insolvency, but it would also have low economic growth. The disuse of credit squeeze as an instrument of governmental economic policy from 1980 onwards makes this explanation more plausible. It should be pointed out that an analysis over the years 1969 to 1980 produced similar results to the period 1969 to 1994. Because of the significance of this variable, it was retained in the model but redefined as an indicator of cash flow rather than credit.

Variable 07: Little can be drawn from the negative sign of the regression coefficient, given the level of significance of this variable. To set the coefficient value of -180.5 in context, a 95% confidence interval for this goes from +396.5 to -757.5! All we can conclude is that the cost of credit has little or no effect on construction insolvency. There is no justification for this variable to be retained in the regression model. Removal of Variable 07 has no detrimental effect on the coefficient of determination of the regression model.

Variable 08: There is rather more confidence, circa 80%, in the negative coefficient for this variable than for the previous variable. A likely explanation is that it is not fluctuations in output that creates problems for contractors but decline in output. The sample frame follows from a then post-war peak in construction output. It covers major recessions in the mid-1970s, the early-1980s and the early-1990s. Despite this there are still more than enough quarters with increasing output to counteract the quarters when output declined. However a test using the negative value, where appropriate, and zero in other cases failed to give positive results. It would appear that fluctuations in output do not have a detrimental effect on insolvency and hence there is no case for variable to be retained within the revised regression model. If the variable is redefined to be an increase in output, with declining output included with negative values, this produces the expected negative coefficient. Thus, increasing output reduces insolvency. The confidence in the sign of the coefficient remained very low and the revised variable was excluded for this reason despite fitting in with the theory and the literature.

Intercept: The intercept in the model is positive and there is near 99% confidence in the its sign. The size and sign of the intercept will affect the predicted level of insolvency. There is nothing in the theory or the literature to test.

Modified regression model

The modified model was run with five explanatory variables, Variable 02 to 06. The overall power of the model was not affected by the removal of Variable 07 and 08. The coefficient of determination remained at around 93%. There were minor effects on the variable coefficients. While this had no effect on the results of the hypothesis tests it lead to slight changes in the levels of confidence in two cases.

The main change is a reduction in the confidence level for the coefficient for Variable 03 to around 87%. Despite this sharp fall in confidence, Variable 03, on balance, warrants retention within the revised model.

The coefficients, t-values, and other results are presented in Table No 6.

Results

Regression model

The revised regression model is presented in the following equation:

C = 220.97 - 296.34 (P) + 109.73 (W) + 640.96 (F) + 0.091 (L_t-4) - 210.73 (V) (Equation 02)

Thus, starting from a base level of insolvency of around 220, it can be expected to increase as borrowing, general insolvency and working capital increase. Similarly, increases in profitability and velocity of circulation will lead to predicted reductions in construction insolvency. For example, an increase of return on capital invested from 20% to 30%, will be expected to reduce construction insolvency by around 30 (296.34 by 10% » 30). Similarly an increase in the velocity of circulation from 2.00 to 3.00 might be expected to reduce insolvency by just over 200.

Factors contributing to insolvency

From the above, it is possible to conclude to a very high level of confidence, over 99.99%, that both return on capital in construction and cash flow problems have an impact on construction insolvency. Cash flow problems are represented by the requirements for working capital, borrowing from banks, and the availability of credit. Lagged general insolvency figures to represent the domino effect also have an clear measurable impact.

It is also, possible to conclude that the cost of credit and fluctuations in construction output have no impact, at least not in the way that the theory would suggest. Despite the apparently negative results for these variables with the coefficients having the wrong sign, the results are not strong enough to disprove the theory.

Balance between the factors

To give an indication of the proportion of total construction insolvency attributable to each factor, the following formula based on Equation 01 can be used:

Ý_j = | b _j c _j | / [ |b ₁ c ₁ | + | b ₂ c ₂ | + | b ₃ c ₃ | + ...... + | b _n c _n | ] (Equation 03)

Where Ý_j = Estimate of share of dependent variable Y attributable to independent variable X_j

c _I = Mean of independent variable X_i

|b _i c _i | = Modulus of mean of independent variable X _i times its coefficient

The results of the above are included in Table No 7. The regression model using average values for each variable give an estimated level of construction insolvency of 377.42. This suggest that the credit and bank borrowing variables are by far the most important with shares of 38% and 32% respectively. By contrast, return on capital only contributed 7%, marginally less than working capital requirements with 8%. Lagged insolvency statistics came in the middle with 16%.

Variation in insolvency attributable to each factor

It may be more useful to assess the contribution of each independent variable to variations in construction insolvency. The range of each independent variable can be estimated by taking the minimum value from its maximum. If the range of each variable is multiplied by its coefficient, this can be used to assess the contribution from each to the total variation in insolvency. The following equation can be used:

Var.(Y_j ) = | b _j R_j | / [ |b ₁ R₁ | + | b ₂ R₂ | + | b ₃ R₃ | + ...... + | b _n R_n | ] (Equation 04)

Where Var(Y_j ) = Estimate of share of variation in dependent variable Y attributable to independent variable X _j

R_I = Range of independent variable X_i [Max(X_i) -Min(X_i)]

|b _i R_i | = Modulus of range of independent variable X_i times its coefficient

The results of this are presented in Table No 8. The results to not differ greatly in terms of magnitude and ranking from those in Table No 7. Again, cash flow related variables are the most significant. Bank borrowing (42%) is followed by lagged general insolvency (29%) and credit (18%). Profitability comes near the bottom of the ranking with 8% and working capital with 3%.

Conclusion

Key findings

The above empirical analysis lead to the generation of a fairly robust predictive regression model. It also confirmed much parts of the theory thrown up in the literature review via a series of hypothesis tests. It also suggests that cash flow related variables are responsible for the majority of construction insolvency and also for most of the variation in insolvency. While the return on capital invested is a significant factor in construction insolvency, less than 10% of insolvency appears to be attributable to this variable.

The apparently higher failure rates for construction firms as opposed to those in other sectors is not fully explained by the analysis. However, the limited rôle of profitability and the high profile of cash flow related factors may provide a partial explanation given the small size of many construction firms and the limited availability of working capital for small firms.

Future work

The data used for this analysis was flawed to some extent due to the use of different sampling frames. For example, the insolvency data used relates for England and Wales, data on construction output is for Great Britain, and economic data is for the UK. While it is likely that this will not effect the outcome, it might be profitable for a future analysis to examine Scotland and Northern Ireland to see if the same pattern is observed as for England and Wales. Another future analysis ought to concentrate on receiving orders and sequestrations for the self-employed to observe if this results in the same outcome as for company insolvency. In particular, the impact of the domino effect of corporate insolvency on individual bankruptcy warrants further analysis. Finally a comparative analysis against other sectors, such as manufacturing or agriculture, might throw up some additional insights into construction insolvency.

References

Department of Trade (1975) "Insolvency statistics in England and Wales", Economic Trends, 257, March 1975, Central Statistical Office, HMSO, London, pp. 119-123.

Hillebrandt, Patricia (1977): "Going bust: What are the facts?", Building, 11 February , 1977, pp 52-53.

Hillebrandt, Patricia (1984): Analysis of the British Construction Industry, Macmillan Press, London.

Lea, Eleanor; Lansley, Peter; & Spencer, Paul (1972) Efficiency and growth in the building industry: A study of twenty-three building firms, Ashridge College Management Centre.

Lowe, John G (1993): Construction productivity: an input-output approach, Unpublished Ph.D. Thesis, Heriot-Watt University, Edinburgh.

Lean, William; & Goodall, Brian; (1966): Aspects of land economics, Estates Gazette, London.

NEDO (1975): The Public Client and the Construction Industries. Building and Civil Engineering EDCs, National Economic Development Office, HMSO, London.