Tracing the connections: short-termism, training and recession

Abstract It is frequently claimed that some employers act to maximise short-term gains at the expense of long-term rewards, hence reducing the level of employee training. In addition, during a recession such employers are expected to be the keenest to make further cutbacks. This paper examines the empirical validity of these two claims by examining the links between three proxies for short-termism and the incidence and volume of training activity as well as recession-induced changes to training expenditure and the proportion of the workforce trained. The results are based on establishment-level data taken from 67,599 private sector employers in England in 2009 and enriched with data from other sources (with sample sizes falling accordingly). The results suggest that short-termism plays a role in explaining both the level of training activity supported by employers and its sensitivity to the economic cycle. However, the results are rather ambiguous with one of the proxies suggesting that, contrary to theoretical reasoning, training incidence and volume is higher, not lower, in establishments which belong to stock market listed rather than unlisted enterprises. To make further analytical headway, then, direct measures of short-termism are needed rather than indirect, albeit improved, measures of the type used here.


Introduction
In Anglo-Saxon economies, there is a widespread belief that employers are averse to making investments which may generate large returns in the future if this requires sacrifices in the present. This 'short-termist' mentality has featured in political discussions of the comparative performance of the UK economy prompting statements such as 'short-termism seems hard wired into our economy' (Ed Miliband, former Leader of the Opposition, Financial Times, 19 January Miliband, 2012). The concern is that short-termism places UK businesses at a comparative disadvantage with competitors, particularly those from Germany and Japan, where there is a stronger focus on developing long-term productive capacity (BIS, 2012;Gospel & Pendleton, 2005). Investments in activities such as research and development (R&D), replenishing the capital stock and other items which might improve future economic performance are, therefore, lower in the UK than elsewhere and more susceptible to cutbacks in recession.
While the existing literature on short-termism on business behaviour has focused on uncovering the link between short-termism and patterns of research expenditure, product development and investment in new equipment (e.g. Hughes, 2014;Demirag, 1998), its impact on training activity has received scant attention. Yet developing the skills of workers through training is a well established means of increasing the productive capacity of human resources and as such is a key aspect of human resource management (HRM) (Boxall, Purcell, & Wright, 2007). However, like other long-term investments, the fear is that it may come under threat in times of recession (e.g. UKCES, 2008). The contribution of this paper is a twofold response to this neglect. First, it presents new empirical evidence which tests the claim that short-termism reduces the level of training activity employers are willing to fund. Secondly, it sheds new light on the fear that in recession short-termist employers are the most likely to cut training activity (cf. Felstead, Green, & Jewson, 2012). However, given the single country focus of the data used, the argument that UK employers are more short-termist than their international counterparts cannot be ruled out.
The paper is structured accordingly. Section 2 outlines the literature on short-termism and identifies two sources of this behaviour. The first is generated 'outside' the firm by capital markets and the ownership structure of the business. This is exemplified by the impact that stock market listing is reckoned to have on organisational behaviour through the greater emphasis it places on share price growth and dividend payouts. The second source of short-termism is 'internal' to the firm and is, in principle, more widespread. It stems from managers' own preferences for short-term results that serve personal interests even if at the expense of long-term business benefits. Such opportunistic managerial behaviour is based managers' desire to quickly build a reputation for delivery before moving on to other roles.
In this paper, these two sources of short-termism are proxied by the survey measures and enhancement procedures outlined in Section 3. These are then used to analyse data taken from an establishment-level survey of human resource managers carried out in England in 2009 and referred to as the National Employer Skills Survey 2009 (NESS2009). Using data taken from the 67,599 (and sub-sets thereof) private sector establishments which took part, the paper examines the association that short-termism has with the level of employer-sponsored training and recession-induced movements in this activity -as measured by training expenditure per capita and the proportion of employees trained. However, the survey items which relate to short-termism have limitations and so only provide first approximations. These drawbacks are addressed by importing information from elsewhere into the primary data source. This represents an important methodological contribution to the debate. The two-part presentation of the results in Section 4 reflects this approach. The first is a descriptive account of how the original proxies for short-termism -based on survey items in NESS2009 -relate to the incidence of employer provided training, the volume of provision and its susceptibility to recession-induced change. The second set of results uses multivariate analyses to identify whether short-termism -where possible measured by improved proxies -has a significant impact on the decision to provide any training at all, the amount of training undertaken and the effect of recession on training activity, after holding other factors constant. Section 5 concludes by highlighting the need for more direct measures of short-termism rather than relying on indirect, albeit improved, measures of the type used here.

Existing literature
It is frequently pointed out that UK expenditure on long-term investment is lower than in competitor nations such as Germany, France and the US. According to one survey over 70% of business leaders put this down to short-termism (Cox, 2013, pp. 14-18); that is, decisions 'that generate an early pay-off relative to strategies what would have added much more value, but at a significantly later point in time' (Barker, 2012, p. 3).
The nature of corporate governance, and in particular the influence of the stock market on decision-making behaviour, is often cited as the source of this myopia (Laverty, 1996). The theory is that managers of listed companies operate under governance structures which give the interests of shareholders primacy -known as the 'outsider-orientated' model (Gospel & Pendleton, 2005). Shares in listed companies can be bought and sold on stock markets. This means that ownership can come to rest with 'external' shareholders whose connection to the business is purely financial and often fleeting. Ownership of listed companies, therefore, tends to be widely dispersed across a large number of shareholders who have limited commitment to the organisation's long-term prospects. Compared to those directly employed by the organisation, shareholders are unlikely to be exclusively or even mainly dependent on the organisation for their income. Thus, shareholder and worker interests may not always coincide, but to keep and attract investors, managers give the interests of shareholders priority. As a result, managers' planning horizons are shorter and greater importance is attached to the movement in quarterly financial indicators on which shareholders concentrate most (Konzelmann, Conway, Trenberth, & Wilkinson, 2006;Dickerson, Gibson, & Tsakalotos, 1995).
This theoretical position has received empirical support from two types of evidence (Aspara, Pajunen, Tikkanen, & Tainio, 2014;Hughes, 2014). One approach is to look at current share prices and assess how closely they reflect future dividends. Short-termism occurs where current share prices are below the level predicted by future earnings. In this situation, investors are acting myopically by attributing too low a present value to the stock's future earnings (i.e. discounting the future too heavily). Several studies have examined the share prices and dividend payments of firms listed on the London Stock Exchange using this approach (Davies, Haldane, Nielsen, & Pezzini, 2014;Miles, 1993). They conclude that there is significant evidence of stock market myopia. Moreover, research shows that in comparison to their Australian, German, Japanese and US counterparts, those who invest on the London Stock Exchange are 'markedly short-termist with medium-and long-horizon cash flows being consistently, and relatively highly, underweighted in current stock market valuations' (Black & Fraser, 2000, p. 39).
Another approach to detecting myopia is to survey the attitudes of corporate decision-makers to investigate whether they believe they are judged by stock market performance. A survey of finance directors, for example, found that investment decisions were taken with a closer eye on what impact these decisions had on short-term stock market performance than on the long-term benefits these investments might eventually yield (Grinyer, Russell, & Collison, 1998). Similarly, a survey of board directors suggests that the perception of short-term pressures from the stock market means that more emphasis is given to R&D projects with a quick, predictable and safe return (Demirag, 1998).
Unlisted companies, on the other hand, operate on 'insider-orientated' principles (Gospel & Pendleton, 2005). For these companies, there is no organised market where shares are bought and sold, hence the route to exit for existing shareholders is limited. In addition, without a market for shares, management performance cannot be assessed by movements in the share price, thereby removing this source of control. As a result, the theory suggests that time horizons for investments in unlisted companies are longer.
Corporate governance theory, then, suggests that 'outsider-orientated' companies are more constrained than those with an 'insider-orientation' . This can have an impact on HRM practices leading to a reduction in management-worker dialogue over workplace change, reduced autonomy for workers in terms of how work is carried out and greater use of individualised systems of pay (Conway et al., 2008). It can also have an important bearing on how organisations respond to changing economic conditions and how these responses affect workplace outcomes such as training activity (Gospel & Pendleton, 2005;Black, Gospel, & Pendleton, 2007). In line with this paper's overarching hypothesis, then, the expectation is that establishment-level training activity is likely to be lower and more susceptible to cutbacks where enterprises are listed on the stock market compared to otherwise identical private sector establishments. However, existing evidence is weak. For example, analysis of the Workplace Employment Relations Survey 2004 (WERS04) does 'not support the argument that stock market listed workplaces are less likely to provide training than other private sector workplaces' (Pendleton & Deakin, 2007, pp. 348-349). In fact, listed workplaces were found to have a statistically higher probability of training larger proportions of workers. Nevertheless, the amount of training given in listed workplaces did not differ significantly from other parts of the private sector.
There is a second source of short-termism, unrelated to the actual or perceived preferences of financiers as expressed in stock market pressures. This is referred to as 'managerial opportunism' , and is generated within businesses (Laverty, 1996;Jackson & Petraki, 2011). It stems from the fact that managers have their own objectives and ambitions which are shaped by how their achievements are recognised by employers (Narayanan, 1985). As a result, managers take decisions which produce short-term results that serve their own interests at the expense of decisions which might be more favourable for the organisation in the long-term. This is a condition known as moral hazard. It is made worse by compensation packages based on performance measures such as bottom line profit. A review of payment systems in 45 large corporate entities in the UK, US and Germany, for example, shows that managers' time horizons are shortened the more closely their pay is linked to profits and/or market share, and vice versa (Coates et al., 1995). Labour turnover among managers -through internal or external job movesalso means that managers tend to favour short-term projects over long-term ones since they are better placed to claim credit for projects they initiate and deliver. Managers who champion long-term projects, on the other hand, risk not being able to take the credit (and rewards) when the project eventually delivers favourable outcomes (Laverty, 2004).
On the basis of managerial opportunism, one would expect short-term managers to quickly react to falling output by laying off staff, thereby protecting shortterm profits on which their own reputations and rewards are based. However, quickly shedding labour carries long-term costs if the productive capacity of the business and its ability to grow and maintain future market share is weakened. On the other hand, long-term employers are more likely to hoard labour in the face of a falling order book. Despite sacrificing a fall in short-term profits, additional training might even be given to existing employees to make them more productive and better able to increase future market share.
Managerial opportunism, and hence short-termism, can also be reflected in the organisation's capacity for innovation in the field. Becoming a market leader, and maintaining that position, requires significant investment. The discovery and development of break-through products and/or the pioneering of new production methods are costly undertakings. They are also risky and uncertain and, even if successful in generating long-term payoffs, managers who secured the necessary resources may not get the credit. Market leadership, therefore, is indicative of organisations which are willing to invest in the long-term. The expectation, then, is that relatively high levels of training activity will be part of a market leadership strategy and more insulated from cuts in times of recession. Those who simply implement and mimic the changes not of their own making, on the other hand, are likely to give investment in training a low priority with the training they do provide being more susceptible to cuts in an economic downturn.

Data sources
This paper is based on an analysis of the National Employer Skills Survey carried out in March to July 2009 (NESS2009). A total of 67,599 private sector establishments operating in England took part in telephone interviews lasting 10-20 min. Respondents were the most senior person responsible for human resource and personnel issues at the site. Interviewees provided information on the level of training activity undertaken in the last 12 months and the reported impact of the 2008-2009 recession on this activity. The survey also collected some data on short-termism. However, the research reported here makes improvements to the latter by matching into this data-set information collected from two administrative sources. First, from the Financial Analysis Made Easy (FAME) database, information is extracted relating to stock market listing. Secondly, data are imported on employment and turnover change in the two-year period up to the time establishments were surveyed for NESS2009. These data are taken from the relevant Business Structure Databases (BSD). These two different administrative data sources, then, provide enhanced measures of two of the three proxies for short-termism examined in this paper, namely stock market listing and changes to the labour-output ratio.
However, it must be remembered that not all NESS2009 respondents agreed to their survey data being linked to administrative sources. Around a fifth refused to give their consent to data linkage. Furthermore, incomplete data records meant that common identifier codes were successfully added to around half of these establishment addresses (Evans & Welpton, 2009). Administrative data of the type outlined above were added to the resulting sample of 33,319 private sector establishments. The multivariate analyses which follow are based on this reduced sample with all of this analysis carried out in the ESRC's Secure Data Service (SDS). 1 The descriptive statistics, on the other hand, are based on the full sample of 67,599 establishments where appropriate (non-training employers, for example, were not asked follow-up questions about non-existent training activity).
The loss of around half of the original sample raises the possibility that the reported results based on the reduced sample are biased. To examine this bias a series of logistic regressions were carried out. These show that the reduced sample is, for example, skewed towards smaller workplaces and those operating in the hotel and restaurant sector, while under-representing establishments in finance and transport. Also over-represented are establishments offering training, hence biasing the sample and reducing the number of establishments who chose not to invest in training at all and possibly excluding establishments where the links between the decision to undertake training or not and short-termism are strongest. However, the reduced sample does not differ significantly from the full sample in terms of the other dependent variables, namely training volume and recessioninduced changes to the amount spent on training and the proportion of workers trained. Nonetheless, these sample biases are recognised and acknowledged, and need to be borne in mind when interpreting the results.

Dependent variables
The paper analyses four dependent variables focusing on training incidence, volume and two recession-induced changes to training activity. The derivation of each is briefly outlined in what follows. Respondents to NESS2009 were asked whether over the past 12 months they had 'funded or arranged any off-the-job training or development for employees at this site' or whether they had 'funded or arranged any on-the-job or informal training and development' . Establishments are defined as trainers or non-trainers accordingly. The training incidence rate, therefore, refers to the proportion of establishments undertaking training regardless of the extent to which they were doing so. However, as a follow-up, those providing training were asked how many staff they had trained over that period and for how many days on average. By multiplying these two figures and then dividing the result by the number of workers on the payroll at the establishment, an establishment-level estimate of training volume is derived with the intensity indicator for non-trainers set at zero. However, those undertaking training but failing to provide an estimate of how many they were training and for how long, had to be dropped from this part of the analysis (hence the sample sizes vary accordingly, cf. Table 1).
The final two dependent variables focus on the impact of the recession on training activity. Respondents were asked: ' As a result of the [2008][2009] recession have the following increased, stayed about the same or decreased at this establishment' . A list of eight statements were read out, two of which focused on training activity as measured by 'expenditure on training per employee' and 'the proportion of employees provided with training' . Those not providing training were routed past these two questions and so the sample sizes for these two dependent variables fall accordingly. For analytical purposes, those reporting decreases are defined as 'cutters' , those reporting no change are denoted as 'stickers' and those reporting increases are defined as 'boosters' . Where necessary, these labels are prefixed with 'expenditure' and 'coverage' in order to indicate which aspect of training activity is under discussion. However, the number of 'cutters' may be underestimated since it excludes those who cut training as a result of the recession more than 12 months before the interview. It should also be noted that the coverage and expenditure statements apply across all the departments, occupations and contract types that constitute the reporting unit, respondents were asked to give an 'averaged' account of the impact of the recession on training in the establishment. The Table 1. Descriptive statistics on private sector employer training: incidence, volume and recession-effects.
Notes. column and row percentages do not always add up to 100% because of a small number of 'don't know' responses. sample sizes vary according to the number of respondents who were asked the question (i.e. those not offering training were not asked about the impact of the recession on their training activity since none was reported).
Source. own analysis of the standard ness2009 data-set. training volume indicator is also based on an average estimate of the number of days of training each trainee received. Furthermore, all four training variables are subject to measurement error since respondents were required to recall activities, provide estimates and identify the effects of recession. However, by carrying out interviews with the most senior person responsible for HRM at the establishment, the research was designed to minimise these errors.

Measures of short-termism
Management scholars and financial economists have devoted considerable effort to examining the existence or otherwise of stock market myopia and its effects on business performance. HRM scholars have also begun to examine the connection between corporate governance and the nature of employment relations (e.g. Konzelmann et al., 2006;Edwards & Walsh, 2009). However, the data-sets HRM scholars use rarely collect information on whether a public limited company (PLC) is listed on the stock market or not. NESS2009 is no different. Like many surveys, it asks respondents whether the organisation to which the establishment belongs sells shares to the public, but not whether these shares are traded on the stock market, and therefore bought, sold and held widely. Analysts are therefore left with imprecise measure of exposure to outside pressures. This is a major drawback since stock market listing is the key mechanism through which ownership is widely dispersed and short-termism is expected to flourish. These short-term pressures may, in turn, translate into lower levels of training activity and a greater willingness to cut back on training in times of recession. The result is a poorly defined indicator which may conceal the full effects of short-termism.
To address this limitation, NESS2009 is enriched with data imported from FAME. This is a database comprising financial and descriptive information on the UK operations of companies extending over 10 years. For present purposes, its importance is that it contains information on the company's legal form and, in particular, whether its shares are quoted on a stock market. To maximise the linkage, rate historic FAME data relating to June 2009 were used in order to coincide with the period of NESS2009 data collection. Using this improved proxy, 1017 NESS2009 establishments, which were part of a PLC whose shares were listed, were identified. According to theory, these are likely to be the establishments most exposed to short-termism.
The second indicator of short-termism focuses on employment levels falling faster than output as employers protect the present in preference to the future. The suggestion here is that short-termist employers are among the quickest to downsize. This will also be reflected in a lowering of employers' willingness to invest in training and increase in the likelihood that they will cut back on training in times of recession. While NESS2009 has a question on how staffing levels have changed as a result of the 2008-2009 recession, the responses given only indicate in which direction, if any, establishment staffing levels moved -up, down or remaining the same. So, from the raw NESS2009 data, there is no way of telling how significant these changes were for the establishment and whether or not these changes were above or below the reactions of other establishments operating in the same sector.
The real issue for this paper is the identification of employers who over-or under-reacted to changing economic circumstances by cutting employment levels faster (or slower) than output. Unfortunately, this information was not collected by NESS2009, but employment and financial turnover data of this sort are available from administrative sources such as BSD. However, BSD collects enterprise level data whereas NESS2009 is collected at establishment-level. This presents a problem for analysis since for multi-establishment enterprises there is no way of knowing how to apportion the reported enterprise-level employment and turnover data to establishments. This means that the improved proxy for over/under-reaction to falling output can only be used for single establishment enterprises. A small number of cases were also lost due to the analytical requirement of adding in data relating to recent changes employment and turnover which covered the period leading up to the time when interviews for NES2009 were carried out. For a sub-set of establishments -those operating as single establishment enterprises Table 2. multivariate analysis of private sector employer training: incidence, volume and recession-effects, all establishments operating in england.
Notes. controls were included for sector (13 dummies), region (8 dummies) and workplace size (4 size bands, but continuous and squared in linear regressions). *** =1% significance level; ** =5% significance level; * =10% significance level. Source. own analysis of sDs held ness2009 data-set with importation of data from fame, June 2009. -this improved proxy for changes in the labour-output ratio is entered into the regressions, while for the all establishment analysis, the more widely available yet less precise survey-based measure is used (cf . Tables 2 and 3). However, NES2009 does contain a relatively good survey item on market leadership. Respondents were asked to compare their establishment with others in the industry in terms of the degree to which it exercises market leadership.
Respondents were asked to position the establishment on a 1-5 scale ranging from 'often lead[s] the way' 'in terms of developing new products, services or techniques' at the top to it 'very rarely lead[s] the way' . In what follows, a threepoint scale of market leadership is created by keeping the top and bottom of the scale and collapsing the intermediate categories. This is used in the descriptive and multivariate analyses reported below. The survey also collects some limited background data on establishments. These data are used as control variables in the regressions which follow. These controls are: economic sector (14 are identified); region (9 areas are specified); and size of the establishment (as measured by the number of staff on the payroll regardless of their employment status). All the results are weighted to take into account reported variations in response rates among establishments and hence produce representative results (Shury et al., 2010, Appendix A6).

Descriptives
According to NESS2009, around two-thirds of establishments (65.3%) reported that training was undertaken in the 12 months before the survey was carried out. In volume terms, this translates into 4.3 days training provided on average for every employee in the establishments surveyed. Of those who provided training, 19.5% said that they had reduced training expenditure per capita, 70.2% made no change and 8.0% said that they had boosted their per capita training spend. A similar pattern is in evidence as regards training coverage with 13.0% categorised as cutters, 76.9% as stickers and 8.7% as boosters (see Table 1).
What role does short-termism play in whether or not employers provide employee training, the volume of training provided and, if they are trainers, how do deteriorating economic circumstances impact on variations in training expenditure and the proportion of the workforce in receipt of training? The first step in answering this question is to cross-tabulate the four dependent variables against the three survey-based proxies for short-termism discussed above.
First, we examine the raw corporate governance indicator. In short, the cross-tabular results for this indicator fail to support the hypothesis that shorttermist employers are less likely to train, provide lower volumes of training and are more prone to cut back on training in recession. On the contrary, training incidence and volume were higher, not lower, in establishments which operated as part of a private sector PLC. They also accounted for a higher, again not lower, proportion of establishments which boosted rather than cut training expenditure and coverage. For example, around a fifth (20.4%) of establishments reporting a cut to training coverage were part of a PLC compared to over a quarter (27.3%) of establishments reporting a boost. Initially, then, descriptive support for the short-term hypothesis is lacking. However, no data are available on whether establishments were part of an enterprise which was listed on the stock market and at greatest exposure to outside pressures. By using a better corporate governance measure, deploying multivariate techniques and carrying out significance tests, this hypothesis will be subject to further scrutiny later in the paper.
However, other descriptive findings give stronger support for the short-termism hypothesis. For example, those who increased staffing as a result of the recession were also among those most likely to offer training to their staff as well as supporting higher volumes of training per employee (see Table 1). This pattern is repeated for recession-induced changes to training activity. Among those reporting cuts to their training expenditure per capita around three-fifths (60.3%) also said that they had reduced staffing levels compared to less than a fifth (18.4%) of those who boosted expenditure (see Table 1). The reverse applies with increases in staff numbers being closely correlated with boosts to training expenditure. Over a quarter (27.3%) of training boosters had increased staffing as a result of the recession, whereas only a fraction (4.1%) of cutters reported that they had increased training expenditure per head at the same time. The pattern for training coverage is similar and, if anything, even more pronounced. However, given that new recruits are those who require most training it is perhaps not that surprising to find such a relationship. In these circumstances, ceteris paribus, a fall in recruitment will necessitate falling per capita training expenditure as well as falling training coverage.
A more general drawback of this analysis is that it cannot be deduced from these data whether those surveyed were over-or under-reacting to the recession since the survey does not provide data benchmarking the actions of establishments against those of their peers. In other words, to what extent do responding establishments report wilder movements in their labour-output ratios as compared to the sector average and what associated effects do these exaggerated movements have. This drawback is addressed using improved measures taken from administrative data in the multivariate analyses which follow.
The extent to which the establishment offers market leadership to the sector is the third proxy measure of short-termism. The suggestion from the crosstabulations is that market leaders were more likely to offer employees at least some training, provide staff with more intensive training and were more likely to boost rather than cut training expenditure and coverage as a result of the recession. On the other hand, training incidence and volumes were lower among those who rarely led the sector with 'new products, services or techniques' . Less innovative employers were also more likely to cut rather than boost training. The column percentages rise or fall accordingly and provide some initial support for the shortterm hypothesis (see Table 1).

Multivariate analyses
The next step in the analysis is to use, where possible, the improved measures of short-termism which capture: (a) stock market listing; and (b) changes to the labour-output ratio over two years covering the period before, during and towards the end of the 2008-2009 recession. In what follows, these improved proxy measures of short-termism are used along with the market leadership measure taken directly from NESS2009. Where this is not possible, proxies based on survey responses are used instead. All of these proxies are entered as independent variables, along with limited set of controls, into a series of logistic, linear and ordered probits as appropriate (with odds ratios and coefficients reported accordingly). The aim of the procedure is to highlight whether the short-termism proxies are significantly correlated with training incidence, the volume of that investment and the decision to cut, maintain or boost training activity in the recession.
It is frequently claimed that stock market listing puts pressure on employers to adopt a cost-cutting approach to HRM, with training suffering as a consequence. Despite this claim, the existing empirical evidence for the connection is weak. The FAME enriched NESS09 data allow further scrutiny of this finding and also uses, like previous research, training incidence and training volumes as outcome indicators. However, the present analysis is based on a representative private sector sample much larger than the WERS04 analysis which was based on a sample of 1563 British workplaces (Conway et al., 2008). In addition, the analysis presented in this paper examines whether listed workplaces were more likely than unlisted ones to be among those who cut training activity as a result of the recession rather than maintaining or boosting it. The results -like those of WERS04 -are not consistent with the arguments that listed companies take a more 'hard-nosed' approach to developing their employees. Instead, listed workplaces were more likely to train their staff and offer longer periods of training than non-listed private sector establishments (i.e. the odds ratio is significantly greater than one and the linear coefficient is positively significant). Furthermore, listing appears to make no difference to employers' willingness to cut or boost training activity as a result of the recession, whether measured by training expenditure per head or the proportion of workers trained (see Table 2). The results, then, for the impact of stock market listing are inconsistent with the theoretical predictions reviewed above, thereby providing further doubt on the connection between stock market listing and downward pressure on training.
Theories of short-termism predict a positive and significant statistical relationship between establishment-level movements in employment and the provision of staff training, and vice versa. Using the survey-based measure of employment change, this hypothesis is supported by the data reported here (as indicated by an odds ratio exceeding 1, and appropriate negative and positive coefficients in Table  2). Increases in staffing are associated with an increased propensity to train (as compared with establishments which experienced no employment change), higher volumes of training and a greater willingness to increase the training activity which exists. While for establishments reporting falls in employment, the reverse story holds for the recession-induced changes to training activity indicators, but falls in staffing levels are not associated with falls in training incidence or volume. Furthermore, the control variables suggest that a few sectors, such as hotels and restaurants, which have high rates of labour turnover (Heap, 2005) have relatively high levels of training per head, after controlling for reported changes to establishment staffing levels. However, there is no evidence of widespread sectoral variation for either this or the other three outcome variables. The findings reported in this paper, then, hold regardless of sector.
However, as indicated earlier this survey-based proxy of short-termism does not indicate whether employment changes are in line or out of kilter with the movement in business turnover and whether these labour-output moves are common to the sector or not. Data matching rectifies this deficiency by using more precise measures of the rapidity with which employment change outpaces (or is slower than) changes to financial turnover as benchmarked against the sector average. We can, therefore, better determine where establishments with short-termist tendencies exist and where they do not. However, the matching process restricts analysis to single establishments; hence the results here are for a particular sub-set of respondents. Not surprisingly, only a few such establishments had their shares listed on the stock market and so the survey-based proxy for corporate governance is used in these regressions (see Table 3). 2 The results confirm the dampening effect that short-termism -as measured by sector-benchmarked recent changes to the labour-output ratio -has on an establishment's propensity to offer training and the volume of training per capita subsequently carried out. That said, this measure does not appear to explain movements up or down in training activity as a result of the recession (see Table  3). So, there is partial, but not complete, support for the detrimental impact that short-termism has on employers' training activity using this particular measure.
Cross-tabular results show that establishments rated as market leaders in the field were also those most likely to be among those offering staff training and providing higher volumes of training activity than other private sector employers. Market leaders were also more likely to report that their training activity had increased in the recession -both in terms of expenditure per head and employee coverage -and least likely to report making cuts. The reverse is true of market followers. Broadly speaking, this pattern holds up when subject to more stringent multivariate analysis which hold other factors constant -the coefficients for market leaders are positive and significant, while they are negative and significant for market followers (see Tables 2 and 3). Here, again, then there is support for the downward impact that short-termism has on employers' training behaviour.

Conclusion
This paper is focused on a large survey of employers which was conducted in 2009, around the time that one of the longest and most severe recessions in Britain was officially coming to an end. Employers were asked about whether they had offered training to staff in the last 12 months, if so what proportion of staff this covered and how long on average this training lasted. Those providing training were also asked what impact the recession had on this activity. The data-set, therefore, provides establishment-level information about the level of training undertaken and its susceptibility to change in times of recession as reported by managers responsible for its delivery. Using these data, the paper tests the connections between short-termism, training and recession.
According to the theoretical literature, short-termism has two main sources. The first comes from the corporate governance structure of the organisation and outsider scrutiny of managerial decisions. Organisations whose shares are traded on the stock market are closely scrutinised by shareholders whose connection to the business is purely financial and often short-lived. Their interest is likely to be narrowly focused on share price growth and the size of the dividend payout, and not on long-term investments in activities such as training. Unlisted companies are more 'internally-orientated' and are therefore less constrained and more focused on the long-term. The second source of short-termism stems from managers' opportunist interest in short-term results that serve personal interests even if at the expense of long-term business benefits. Such a mentality is evidenced by a quickness to reduce headcounts in the face of falling turnover and a reticence to innovate.
In the absence of direct evidence on short-termism, this paper uses three survey-based proxies as first approximations of the phenomenon. Additional information taken from administrative sources is added, where possible, and these improved proxies are used instead. By these means, the paper examines what impact stock market listing and changing labour-output ratios had on the decision to train and for how long as well as on actions to cut, maintain or boost training activity as a result of the 2008-2009 recession.
Despite theoretical expectations to the contrary, the paper offers mixed support for the association that short-termism has with training. Most strikingly, the weakest empirical support for these associations comes where theoretical argument for their existence is at its strongest, namely the impact of stock market pressure on managerial decision-making. Indeed, in multivariate analyses stock market listing is positively, not negatively, correlated with the incidence of establishments to offer training and to provide more of it. Furthermore, stock market listing does not appear to be associated with reported movements in training expenditure nor its coverage across the establishment's workforce. In line with previous research, based on similar but much smaller samples, the paper offers little support for the short-termist hypothesis when proxied by stock market listing.
However, in interpreting these results it must be remembered that no direct measure of short-termism is available. Instead, the paper uses a number of proxies measured in various ways and using a combination of different data-sets. The result may well be that proxies -such as stock market listing -used here and elsewhere are not sophisticated enough to capture short-termism or that some of the data matching techniques used produce a biased sample. It may be, for example, that short-termism is not a particular feature of listed companies as a whole, but rather a feature of a particular segment. Case study evidence, for example, suggests that listed companies with large continental European pension fund shareholders have more latitude to pursue long-termist investment strategies (Deakin, Hobbs, Konzelmann, & Wilkinson, 2002). On the other hand, private limited companies -such as those owned by private equity companies -may be more aggressive in pursuing short-term strategies than their stock market listed counterparts (Kosman, 2009). The result may be that stock market listing is a blunt proxy for short-termism and therefore it may not adequately identify businesses which take actions in the present for quick gains.
Nevertheless, stronger evidence is provided by other proxies linking shorttermism with the level of, and changes to, training activity. Movements in the labour-output ratio (benchmarked against the sector average) and market leadership indicators are both significantly correlated with the decision to train, the volume of training undertaken and the resilience of training to an economic downturn (at least in the case of market leadership).
There is, then, evidence in this paper to suggest that short-termism is detrimental to training activity, albeit not as a result of stock market listing as many theoretical predictions suggest, but as a result of more widespread managerial opportunism. However, to make further headway in the debate, better measures of short-termism need to be developed, perhaps using direct indicators of managers' attitudes such as those pioneered by some of the management research which has inspired this paper. The recent interest in short-termism and its impact on economic performance by policy-makers may prove the additional spur needed to collect the necessary data, and so place the study of short-termism and its effects on HRM issues such as training on a sounder and more robust empirical footing.

Notes
1. However, a few respondents failed to answer all of the questions necessary for the runs reported in Table 2, hence the difference with the total sample reported here. 2. It is not possible to run Tables 2 and 3 on the same samples since the two improved proxies can only be applied to different sub-sets of the sample. Furthermore, the outcome variables address different issues; two focus on the extent of training (with zero a possibility) and two on the susceptibility of pre-existing training to economic pressures. However, additional runs keeping the sample sizes of the two measures of the former and the two of the latter the same produce results very similar to Tables  2 and 3.