Your SlideShare is downloading. ×
A Good Worker is Hard to Find: Skills Shortages in New Zealand Firms
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

A Good Worker is Hard to Find: Skills Shortages in New Zealand Firms

806

Published on

Our work on skills is cited by the New Zealand Ministry of Economic Development in the document titled A Good Worker is Hard to Find: Skills Shortages in New Zealand Firms.

Our work on skills is cited by the New Zealand Ministry of Economic Development in the document titled A Good Worker is Hard to Find: Skills Shortages in New Zealand Firms.

Published in: Business, Health & Medicine
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
806
On Slideshare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. A Good Worker is Hard to Find: Skills Shortages in New Zealand Firms Penny Mok, Geoff Mason, Philip Stevens and Jason Timmins Ministry of Economic Development Occasional Paper 12/05 April 2012 ISBN: 978-0-478-38237-2 (PDF) ISBN: 978-0-478-38236-5 (online)
  • 2. Ministry of Economic Development Occasional Paper 12/05A Good Worker is Hard to Find: Skills Shortages in New Zealand FirmsDate: April 2012Author: Penny Mok Treasury Geoff Mason National Institute of Economic and Social Research, London Philip Stevens Ministry of Economic Development Jason Timmins Department of LabourAcknowledgementsThe authors would like to thank Belinda Buxton, Rosie Byford, Elizabeth Chisholm,Hilary Devine, Sid Durbin, Richard Fabling, Bette Flagler, Julia Gretton, Hamish Hill,Bill Kaye-Blake, Peter Nunns, Meighan Ragg, Lynda Sanderson, Steven Stillman,Richard White and participants in the Industry Training Federation, the Labour,Employment and Work, the New Zealand Association of Economists, and the NewZealand Centre for SME Research conferences (and others whom we haveneglected to mention). We acknowledge two lots of funding from the Cross-Departmental Research Funding for the production of IBULDD (and thence the LBD)and for the Impact of Skills on New Zealand Firms project.Contact: Occasionalpapers@med.govt.nzDisclaimerThe views, opinions, findings, and conclusions or recommendations expressed in thisOccasional Paper are strictly those of the author(s). They do not necessarily reflectthe views of the Ministry of Economic Development or the Department of Labour,Treasury or the National Institute of Economics and Social Research. The Ministrytakes no responsibility for any errors or omissions in, or for the correctness of, the
  • 3. information contained in these occasional papers. The paper is presented not aspolicy, but with a view to inform and stimulate wider debate.Access to the data used in this paper was provided by Statistics NZ in accordancewith security and confidentiality provisions of the Statistics Act 1975. Only peopleauthorised by the Statistics Act 1975 are allowed to see data about a particular,business or organisation. The results in this paper have been confidentialised toprotect individual businesses from identification.The results are based in part on tax data supplied by Inland Revenue to Statistics NZunder the Tax Administration Act 1994. This tax data must be used only for statisticalpurposes, and no individual information is published or disclosed in any other form,or provided back to Inland Revenue for administrative or regulatory purposes. Anyperson who had access to the unit-record data has certified that they have beenshown, have read and have understood section 81 of the Tax Administration Act1994, which relates to privacy and confidentiality. Any discussion of data limitationsor weaknesses is not related to the datas ability to support Inland Revenues coreoperational requirements.
  • 4. AbstractThis paper examines the determinants of firms’ skill shortages, using a specially-designed survey, the Business Strategy and Skills (BSS) module of the BusinessOperations Survey. We combine the BSS module with additional data on firms in theStatistics New Zealand’s prototype Longitudinal Business Database (LBD). Wefocus on vacancies that were hard-to-fill because the applicants lacked thenecessary skills, qualification or experience - which we define as skill-related reasons(skill shortage vacancies). We also contrast these with vacancies that were hard-to-fill for non-skill-related reasons.JEL Classification: J24, J31, L60Keywords: skill shortages, hard-to-fill vacancies, Business Operations Survey, Heckman Selection model i
  • 5. Executive SummarySkills are an important determinant of the economic performance of people, firms,industries and economies. Shortages of skilled labour directly constrain productionand prevent firms from meeting demand and using available inputs efficiently. Theyinhibit innovation and the use of new technologies. This may have longer-termimpacts on the way firms do business, in terms of their location, size, structure,production methods and product strategy.In this paper we have examined factors relating to firms’ skill shortages. Our focushas been on vacancies that were hard-to-fill because the applicants lacked thenecessary skills, qualification or experience – what we have defined as ‘skill shortagevacancies’ (SSVs). We have also contrasted these with vacancies that were hard-to-fill for reasons other than the skills of applicants.Worker turnover is a basic fact of economic life. In any given year one in eightworkers leaves their employer. Searching for new staff is, therefore, a position mostbusinesses find themselves in, regardless of their situation. Almost half ofbusinesses find some of these vacancies hard-to-fill and more than a thirdexperience difficulties in finding workers with the required skills, qualifications orexperience they need.We have examined patterns of skill shortage vacancies using an econometrictechnique that allows us to account, in part, for the interrelationship between thelikelihood of a firm posting a vacancy and for those vacancies being hard-to-fill forskill-related reasons.Large firms are more likely to have vacancies – they have more workers and so thechance of at least one of them leaving is consequently higher. However, larger firmsdo not appear to find it more difficult to source workers with the required skills thansmaller firms, once we account for other characteristics. As we might expect, whenfirms’ output grows, the need to seek additional staff also grows. However, this isonly in the years the firm is expanding. Expanding sales is not a good predictor offuture additional staff requirements.Businesses can experience skill shortages internally or externally. A shortage in theskills it requires can manifest itself: (a) in terms of the ability of its existing staff to do ii
  • 6. their job; or (b) in terms of its ability to find appropriately skilled workers throughrecruitment. We have found evidence that these two phenomena co-exist.Businesses experiencing internal skill gaps are also more-likely to have skill shortagevacancies. We also find evidence of persistence in this relationship; businessesexperiencing internal skill gaps are also more likely to have SSVs in the followingyear.Businesses have a ‘make or buy’ option when it comes to skills. If they cannotsource the skills internally or externally, they can up-skill existing or new workersthrough training. Firms’ training choices are the subject of a companion paper to this(Mason et al., 2012). In this paper, we find that firms who undertake training aremore likely to have vacancies, suggesting that their response to skill shortages islikely to be a mixture of recruitment and up-skilling strategies.Our results confirm earlier work that highlights the importance of two distinctions:First, it is important to distinguish between firms’ general recruitment activity and ithaving skill shortage vacancies. Second, we must distinguish between firmsexperiencing recruitment difficulties related to skills and those for other reasons (suchas the firm not offering good enough pay and conditions). Failing to account for thismay cause us to misdiagnose the problem and will therefore undermine theappropriateness of any policy prescription.In any market there will be a number of potential customers that cannot purchase allthey would wish because they are unwilling or unable to pay the market price. Thisdoes not mean that the market is malfunctioning. There is an important question forresearchers and policymakers to ask before interpreting reports of unfilled vacanciesof any kind, and of SSVs in particular, as a ‘skills problem’. This question is: Why dobusinesses that cannot fill a vacancy not simply raise the wages they are offering?There a number of reason why firms may not be able to do so. First, it may bebecause the firm cannot afford to pay any more. If this is the case, the problem is notthe labour market or the education system, but the firm’s own productivity. Firms thatthat undertake activities associated with high productivity are likely to both have ahigher demand for skills and be more able to afford to pay skill premia.Second, the supply of labour takes a long time to adjust. Because of the length oftime required to educate and train skilled workers, it takes a long time for changes in iii
  • 7. wages to influence changes in skill acquisition (particularly if we are relying on therather indirect mechanism of changes in relative wages influencing the choices ofstudents at school) and immigration is restrained by, among other things, legalbarriers and the costs of relocation.We have found that the businesses that feel the applicants they are attracting do nothave the required skills, qualifications or experience are those that are paying higherwages. The fact that businesses that pay higher wages than others in the sameindustry and region are actually more likely to experience SSVs suggests that raisingwages is not sufficient to fill these vacancies. In contrast, when we focus onvacancies that are hard-to-fill for reasons other than skill, paying higher wages doesappear to be a successful policy to fill vacancies. Firms that pay higher relativewages are less likely to have non-skill-related hard-to-fill vacancies. The implicationof this is that SSVs and NSRs are different phenomena.One important issue researchers have to confront when undertaking analysis of thistype is that of causality. It is one thing to establish a statistical regularity – acorrelation between firm characteristics, activities or environment and an outcomesuch as external skill shortages. It is another thing to interpret this as a causalrelationship. In this paper, we have uncovered some clear statistical regularities.These are consistent we with certain predictions we have discussed in this paper andprovide us with useful information to aid our understanding of how the landscape ofskills in New Zealand and, in particular, we done this from the perspective whatbusinesses are looking for, rather than what the education and training system hasprovided. iv
  • 8. Table of ContentsAbstract ...................................................................................................................... iExecutive Summary ................................................................................................. iiTable of Contents ..................................................................................................... vList of Figures ......................................................................................................... viiList of Tables .......................................................................................................... vii1. Introduction ....................................................................................................... 12. Background – Skills and Economic Performance .......................................... 2 2.1. Skills and Firm Performance ........................................................................ 2 2.2. Labour Turnover and Vacancies .................................................................. 5 2.3. Skill Shortages ............................................................................................. 6 2.4. Skills and the Labour Market ........................................................................ 7 2.5. Firm’s Responses to Skill Shortages ............................................................ 8 2.6. Summary .................................................................................................... 103. Data and Preliminary Analysis ....................................................................... 12 3.1. Business Operations Survey (BOS) ........................................................... 12 3.2. Vacancies ................................................................................................... 14 3.3. Hard-to-fill vacancies .................................................................................. 16 3.4. Skill and non-skill-related shortage vacancies ............................................ 19 3.5. Internal skill gaps........................................................................................ 204. Method.............................................................................................................. 21 4.1. Basic econometric model ........................................................................... 22 4.2. Selection equation – The likelihood of vacancies ....................................... 23 4.3. Probability of SSVs..................................................................................... 265. Results ............................................................................................................. 30 5.1. Using contemporaneous variables ............................................................. 30 5.2. Using lagged variables ............................................................................... 35 5.3. Skill and non-skill related hard-to-fill vacancies .......................................... 396. Summary and Conclusions ............................................................................ 41References .............................................................................................................. 45Appendix A1. Data Appendix .............................................................................. 53 A1.1 BOS Variables ............................................................................................ 53 6.1.1. Module A (2005 – 2008) ....................................................................... 53 6.1.2. Module B – Innovation (2007) ............................................................... 54 6.1.3. Module C – Employment Practices (2006) ............................................ 55 6.1.4. Module C – Business Strategy and Skills ............................................. 55 A1.2 LEED/PAYE Data ....................................................................................... 58 A1.3 Other LBD Data (AES, BAI, and IR10) ....................................................... 60 v
  • 9. Appendix A2. Comparison of hard-to-fill vacancies in Module C withrecruitment difficulties in Module A ...................................................................... 62 vi
  • 10. List of FiguresFigure 1 Skill shortages and Hard-to-fill vacancies in Green et al. (1998) .................. 7Figure 2 Vacancies, hard-to-fill and skill shortage vacancies ................................... 14List of TablesTable 1 Vacancies, hard-to-fill, skill and non-skill-related shortage vacancies, % .... 15Table 2 Businesses reporting vacancies, % ............................................................. 16Table 3 Businesses reporting hard-to-fill vacancies, % ............................................ 17Table 4 Year-to-year rank correlation between reported recruitment difficulties ....... 19Table 5 Skill-related reasons for hard-to-fill vacancies, by size ................................ 20Table 6 Existing staff with skills required to do their job ........................................... 21Table 7 Principal component factor analysis of business strategy variables ............ 25Table 8 Factor loadings (pattern matrix) and unique variances ................................ 25Table 9 Principal component factor analysis of 2007 business strategy ................... 26Table 10 Factor loadings (pattern matrix) and unique variances .............................. 26Table 11 Principal component factor analysis of census variables ........................... 29Table 12 Rotated factor analysis .............................................................................. 29Table 13 Rotated factor loadings (pattern matrix) and unique variances .................. 29Table 14 Results - SSVs - Contemporaneous variables - Selection equation .......... 32Table 15 Results – SSVs – Contemporaneous variables – Outcome equation ........ 34Table 16 Results - SSVs - Lagged variables - Selection equation............................ 36Table 17 Results - SSVs - Lagged variables - Outcome equation ............................ 38Table 18 Results – Different definitions of dependent variable ................................. 40Table 19 Staff occupation/role variables ................................................................... 58Table 20 Weighted counts of firms reporting levels of difficulty recruiting new staff in Module A by Module C hard-to-fill vacancy response .................................. 63 vii
  • 11. A Good Worker is Hard to Find: SkillsShortages in New Zealand Firms1. IntroductionSkills are an important determinant of the economic performance of people, firms,industries and economies1. Shortages of skilled labour directly constrain productionand prevent firms from meeting demand and using available inputs efficiently (Haskeland Martin, 1993a; Stevens, 2007). Indirectly, skill shortages inhibit innovation andthe use of new technologies. This may have longer-term impacts on the way firmsdo business, in terms of their location, size, structure, production methods andproduct strategy (Mason and Wilson, 2003; Durbin, 2004; Mason, 2005).The success of policies to enable the emergence and performance of successfulbusinesses depends upon having a workforce with the appropriate skills. There are,however, concerns that New Zealand is experiencing a shortage of workers withparticular skills2. If New Zealand is to raise productivity and improve its internationalcompetitiveness, it is important to understand whether skill shortages do indeed exist,how they manifest themselves and develop policies to address them.In this paper we investigate skill shortages deriving from recruitment difficulties atfirm level. That is, vacancies that are hard-to-fill because applicants do not have the1 See Card (1999) or Dickson and Harmon (2011) for an overview of the results on the individualreturns to education and skills, Abowd, Kramarz, Margolis (1999) or Haskel, Hawkes and Periera(2005) for evidence on the relationship between firm performance and skills, Kneller and Stevens(2005) present international evidence for skills at the industry level in OECD economies and Stevensand Weale (2004) or Madsen (2010) discuss evidence on the relationship between education levelsand growth.2 ‘Skills shortage hits agricultural science ‘, Dominion Post, 13/01/12, ‘Half of Kiwi companies facingskill shortages’, NZ Herald, 30/11/11, ‘Academic warns of major skills shortage’, Waikato Times,4/3/11 1
  • 12. skill, qualifications or experience the business requires. In order to carry out thisanalysis, we make use of a specially-designed survey, the Business Strategy andSkills (BSS) module of the Business Operations Survey (BOS) 2008. We combinethis with information from other sections of the current and previous years’ BOS anddata from the prototype Longitudinal Business Database (LBD). The LBD includesinformation from tax and survey-based financial data, merchandise and servicestrade data and a variety of sample surveys on business practices and outcomes.This allows us to link the responses of the BSS module to a wealth of information onfirms to inform our analysis.The rest of this paper is as follows. In section 2 we discuss some of the issues thatinfluence the interrelationship between the skills of the workforce and theperformance of businesses. In section 3 we describe our data and present somedescriptive information on New Zealand firms’ experiences of skills. We set out ourempirical model in section 4 and our results in section 5. Section 6 concludes.2. Background – Skills and Economic PerformanceHuman capital (and education in particular) has been found to be an importantdeterminant of economic development and explanator of international differences inaggregate economic growth or productivity (Barro and Sala-i-Martín, 1995; Stevensand Weale, 2004; Kneller and Stevens, 2005; Madsen, 2010). There is a largeliterature examining the links between the availability of skills and the performance offirms. These range from detailed case studies (see Keep, Mayhew and Corney,2002, for a summary of those undertaken by the UK National Institute for Economicand Social Research) to econometric analysis using firm-level data and industry levelmeasures of skill shortages (e.g. Haskell and Martin, 1993b, 2001; Haskel, Martinand Periera, 2005; Stevens, 2007). In this section, we consider the importance ofskills for firms, how labour markets match the skills of workers to their uses in firms,and how firms respond to skills shortages. We also consider what we mean by a‘skill shortage’. Can such a thing exist in a well-functioning economy?2.1. Skills and Firm PerformanceThere are essentially two elements to the link between skills and firm performance.First, skills represent a basic input into the firm’s production technology. Individuals 2
  • 13. with higher skills have more human capital and so produce greater output. There is alarge literature that consistently finds positive private and social returns to skill andeducation in particular (Card, 1999; Psacharopoulos and Patrinos, 2004; McMahon,2004).Higher skill levels do not only increase firm productivity through the direct impact onthe worker’s own productivity. They also create synergies with other productiveinputs, such as other workers, physical and knowledge capital, R&D or newtechnologies.An obvious example of how skills enhance the performance of other workers is theskills of managers, who organise and shape production. The quality of managementin businesses is an important predictor of business performance across a number ofdimensions in many countries (Bloom and Van Reenen, 2007, 2010; 2011), includingNew Zealand (UTS, 2009).Another way skilled labour increases the productivity of other workers is through whatare known as ‘knowledge spillovers’ (Arrow, 1962; Battu, Belfield and Sloane, 2004;Audretsch and Feldman, 2004). Spillovers occur when other workers pick up theseskills through observation, interaction or tuition. These spillovers can occur within thefirm and across the boundaries of the firms through direct relationships (i.e. withsuppliers and customers) or through geographic proximity (Glaeser, Kallal,Scheinkman and Shleifer, 1992; Audretsch and Feldman, 2004).For a long time it has been suggested that skilled labour is more complementary withcapital than unskilled labour (Griliches, 1969; Duffy, Papageorgiou, and Perez-Sebastian, 2004). Many capital goods (e.g. computers) require skills to operate(Autor, Katz and Krueger, 1998; Morrison-Paul and Siegel, 2001; Falk, 2004). Autor,Levy and Murname (2003) found that over three decades, computer capital hasreplaced workers undertaking routine or repetitive manual and cognitive tasks andcomplemented those doing non-routine problem-solving and communications tasks.As well as physical capital, higher skill levels enable a workforce to deal with thecurrent technology and, moreover, they enable the firm to better adapt to newtechnology (Machin and Van Reenen, 1998). For economists, technology is takengenerally to include the means whereby inputs are brought together to produceoutputs, including organisational and product strategy. Mason (2005) found that the 3
  • 14. ability of firms in a number of industries to execute ‘high value’ business strategieswas contingent on the skills of their workforce. Work such as Abowd, Haltiwanger,Lane McKinney and Sandusky (2007) finds a strong positive empirical relationshipbetween advanced technology and skill. Skills are required not only for the creationof new technology, but also both for its dispersion and implementation (Rosenberg,1972; Lane and Lubatkin, 1998; Hall and Khan, 2003; Kneller and Stevens, 2006).Investing in skillsBoth workers and firms have an incentive to increase productivity through investing inskills, but their incentives are different. Individuals have an incentive to invest in skillformation because this enables them to function better in the economy and society.Crucially, they provide the basis for them to earn an income. Firms also have anincentive to increase the skills of their workforce because this enables them toperform more tasks or increase the productiveness with which they perform existingtasks. Because the skills embedded in workers can be increased by investing inthem, economists use the term ‘human capital’.Firms do not have the incentive to invest in all types of human capital equally.General human capital is of value to all employers, whereas specific human capital isvaluable only to specific firms or groups of firms (Becker, 1962, 1994; Stevens,1994) 3 . Firms will tend to under-provide training of all skills that have somegenerality to them (i.e. they are of use to other firms) because of the risks of staffleaving or being poached; there is a risk that they will pay the costs and other firmswill get the benefits. Because they may not reap all of the rewards – staff may moveelsewhere or be poached by other firms – firms may invest less in such training thatwould be socially optimal4. The likelihood of staff using recently acquired humancapital by moving elsewhere will depend of the outside options they have, inparticular the wages on offer.3 The specificity of skills may in also be determined by the structure of the market. Acemoglu andPischke (1999) argue that certain skills might be potentially useful to other firms (i.e. general) maybecome specific because of market imperfections. For example, if there were only one producer in asector a particular skill might only be of use to them, whereas if a number of firms were operating inthe same area they could be competing for the same labour.4 Both firms and individuals may under invest in skills and education if there are externalities. Thismight include agglomeration benefits (such as knowledge dispersion), matching externalities, benefitsin terms of lower crime, higher levels of political engagement and health. 4
  • 15. This is where labour differs from other factors of production. It is not tied to the firm,beyond the terms of its contractual relationship. It can decide to take its skill and usethem elsewhere.2.2. Labour Turnover and VacanciesLabour turnover is a basic fact of economic life (Davis, Halitwanger and Schuh, 1996;Davis, Faberman and Haltiwanger, 2006, 2010). For example, from the Septemberto the December quarter of 2010, total employment in New Zealand expanded from1,777,110 to 1,810,580, an increase of 33,4705. This change was dwarfed by thealmost half a million workers either leaving or starting jobs from which it resulted.Because of the dynamic nature of the labour market, we must be careful not toequate vacancies with job creation. Firms can have positive values for both hiresand separations (Davis, Faberman and Haltiwanger, 2006). In the December 2010quarter, 12.7% of workers in New Zealand became separated from their jobs6. Thisis not merely due to firms shrinking (what is called ‘job destruction’). The total jobdestruction in the December 2010 quarter was 90,890 – barely two-fifths of the totalworker separations.Firms will post vacancies for two reasons: First, to replace staff who have quit, retiredor been fired; Second, to fill new roles that have been created by the expansion ofthe firm. These two sources of vacancies are likely to have different causes.Voluntary separations may be higher in good times (as the likelihood of alternativeemployment increases), but involuntary separations may be lower (as firms seek toreduce employment) (Pissarides, 2000). New roles are more likely to be createdwhen a firm is expanding7.Given this discussion, we might reasonably expect the number of vacanciesgenerally increase with the number of employees. However, some work has foundvacancy rates to decline with firms size (e.g. Holtz, 1994).Note that our discussion relates to vacancy rates. Our data is a binary variablestating whether or not a firm posts a vacancy. We would generally expect5 Source: Statistics New Zealand Linked-Employer Employee Database. Data downloaded from:http://www.stats.govt.nz/tools_and_services/tools/TableBuilder/leed-quarterly-tables.aspx6 Source: Authors’ calculations based on LEED data. This separation rate is calculated as ‘Workerseparations’ in the December 2010 quarter divided by the average of ‘Total filled jobs’ in theDecember and September quarters.7 For a textbook description of the standard dynamic model of labour demand, see Nickell (1996). 5
  • 16. phenomena that increase the numbers or rate of vacancies to also increase theprobability of the firm reporting at least vacancy.2.3. Skill ShortagesThe primary focus of this paper is on difficulties firms face in filling roles for reasonsrelated to the qualifications or experience of applicants – what we call skill shortagevacancies. However, this is not the only way in which skill shortages can manifest.Much early work in this area did not distinguish between the types of skill shortage. Itfocussed on ‘skill shortages’ as a whole and their impacts. This was driven in part bythe data that were available to researchers. For example, in the UK, theConfederation of British Industry has collected information on whether firms outputwas limited by shortages of skilled labour since the early 1970s8.However, it has become clear that it is important to distinguish between shortages ofskills in existing workers and shortages of staff in the labour market with appropriateskills (Green, Machin and Wilkinson, 1998; Mason and Wilson, 2003). These havebeen called ‘internal skill gaps’ and ‘external skill gaps’, respectively (e.g. Forth andMason, 2004)9. The primary focus of this paper is on external skill gaps.When examining external skill gaps, it is important to be clear what we are measuring.As Green et al., (1998) point out, we must be careful not to simply equate them withvacancies that are hard-to-fill. As we shall see in section 3.3, there are manyreasons why vacancies may be hard-to-fill. These reasons range from the conditionsof work to the fact that firms are simply not paying the market wage; not all of themare skill-related. Because of this, authors such as Mason and Stevens (2003) havefocussed on the subset of vacancies that are hard-to-fill for skill-related reasons;specifically, by examining the reasons for vacancies being hard-to-fill and onlyconsidering those that relate to a lack of qualifications and/or experience inapplicants. This is the approach that informed the design of the relevant sections ofthe Business Strategy and Skills module of the BOS 2008, and one that underlies ouranalysis in this paper.8 This data has been used as an industry-level measure of shortages of skilled workers by Haskel andMartin (1993a), (1993b) and Stevens (2007).9 Although, others have called skills deficiencies relating to the external labour market ‘skills shortages’and those applicable to a firm’s existing workforce ‘skills gaps’ (Schwalje, 2011). 6
  • 17. We summarise the thinking on the various skill shortage concepts in Figure 1. Skillshortages can be internal to the firm, what we call internal skill gaps (ISG), orexternal, what we call skill shortage vacancies (SSV)10. As well as SSVs, vacanciescan be hard-to-fill for other reasons, what we call non-skill related vacancies (NSR).Figure 1 Skill shortages and Hard-to-fill vacancies in Green et al. (1998) Skill shortages Hard-to-fill vacancies ISG SSV NSR ISG = Internal Skill Gap; SSV = Skill Shortage Vacancy; NSR = Non-Skill-Related vacancyIn order to understand why firms might find it difficult to source skilled labour, weneed to consider the wider labour market. We turn our attention to this next.2.4. Skills and the Labour MarketModern theories of labour markets are based around the ability of the market tomatch individuals to jobs (Pissarides, 2000; Petrongolo, and Pissarides, 2001;Mortensen, 2005). Trading in the labour market is not without cost. It takes time andmoney for businesses to advertise for and assess potential employees. It takes timeand money for workers to search and apply for potential employment. We call these‘search costs’. Furthermore, costs are incurred when incoming workers start in anew role; workers need to learn the specific tasks involved in the role and the10 Of course some external skill gaps may not manifest themselves as SSVs because the firm may notbother to post vacancies they know they cannot fill. 7
  • 18. employer has to learn about the specific abilities of their new employee11. Anothercost is the time a worker or a firm stays in a state awaiting a ‘good’ match. Workersstay in less-preferable jobs or quit to unemployment while they wait for a better offerto come along. Firms either allocate a worker who is less aptly skilled to a role, orhold it open until the ‘right’ worker comes along. We call these ‘mismatch costs’.Average search costs and mismatch costs are likely to be lower in larger labourmarkets; the more jobs and workers there are, the more likely it is that a good matchwill exist. Thus, businesses and workers operating in big cities, for example, willbenefit from lower search costs, ceteris paribus. In part because of this, they are lesslikely to be stuck in ‘bad’ worker-job matches (as the cost of exiting a low productivitymatch is much lower). These ‘thick labour market agglomeration benefits’ can alsocome from co-location of economic activity with similar demand for skills; whilst thereis more competition between firms for skilled labour, there will also be a greatersupply (as workers with the appropriate skills are attracted to the area). Largermarkets will be driven less by market frictions and more by the fundamentaleconomic determinants of the productivity of the match (i.e. those that determine theproductivity of the firm, the productivity of the worker and their complementarity).Firms’ ability to recruit and retain skilled staff will depend on local labour marketconditions. These are typically summarised using the local unemployment rates orproportions of the population with particular qualifications (e.g. Mason and Stevens,2003). The firm’s ability to recruit new staff will also be affected by the ‘outside wage’.The opportunity cost of taking up a job at the firm for the worker is the value of thenext best potential offer foregone. This will depend on the distribution of job offersand associated wages. These can be measured by average wage in similar jobs –for example other jobs in that industry and/or region.2.5. Firm’s Responses to Skill ShortagesHow firms respond to labour market conditions will affect their performance. This ispart of how firms compete. The obvious question to ask if a business cannot fill avacancy is: ‘why don’t they just pay more?’ Here we discuss two reasons why thismay not be possible. First, it may be because the firm cannot afford to pay any more.11 For more on the training of incoming staff by New Zealand firms, see the companion paper to this(Timmins, et al., 2012). 8
  • 19. In this case, the problem is not the labour market, but the firm’s own productivity.Firms that that undertake activities associated with high productivity may well bothhave a higher demand for skills but also be more able to afford to pay skill premia.Firms with some degree of market power may not be more productive, but will beable to compete more aggressively for scarce inputs such as skilled staff.Second, the supply of labour takes a long time to adjust; it takes a long time forchanges in wages to influence changes in skill acquisition (particularly if we arerelying on the rather indirect mechanism of changes in relative wages influencing thechoices of students at school) and immigration is restrained by, among other things,legal barriers and the costs of relocation.In an early study of apparent shortages of engineers and scientists, Arrow andCapron (1959) defined a skill shortage as ‘a situation in which there are unfilledvacancies in positions where salaries are the same as those currently being paid inothers of the same type and quality’ (1959: 301). Although they recognised thatexcess demand for skills in competitive labour markets should put upward pressureon salaries, Arrow and Capron suggested that, even in a competitive labour market,a steady increase in demand over time for skilled workers could produce a ‘dynamicshortage’ if there were factors impeding rapid salary increases by employers such asdelays in accepting the needs for such increases, the further time needed toimplement them and a reluctance to incur increased salary costs for existing high-skilled employees as well as new ones. At the same time supply responses to anysalary improvements could be slowed down by the length of time required to educateand train skilled workers, as shown by Freeman (1971, 1976) in the case ofengineers and scientists.Later studies have recognised that firms have a range of potential non-salaryresponses and ‘coping mechanisms’ available to them when confronted by shortfallsin skills – such as asking existing employees to work longer hours, making increaseduse of subcontractors or retraining existing staff to develop the skills in shortage. In astudy based on data from the 1984 Workplace Industrial Relations Survey in the UK,Haskel and Martin (1993b) found no evidence of firms setting higher wages inresponse to difficulties in recruiting skilled workers. Indeed, they cited other UKsurvey evidence to suggest that salary responses were much less important thanother means of addressing skilled recruitment difficulties. 9
  • 20. Increased training provision is a potentially important non-salary response to externalskill shortages which, like salary increases, should help alleviate the shortages inquestion rather than just help firms to cope with them12. However, just as some firmsmay elect to ‘live with’ external skill shortages for periods of time rather than incur thecosts of raising salaries for new recruits with knock-on effects on existing salarydifferentials, some may also be reluctant to respond immediately to skill shortages byincreasing training provision for existing workers. Such reluctance could reflectimperfect information about the costs and benefits of training versus other potentialresponses to skill shortages.But economic theory also points to other possible explanations for variation betweenfirms in the extent and speed of training responses to different kinds of skill shortage.For example, Stevens (2007) suggests that the speed of firms’ adjustment behaviouris likely to vary inversely with the degree of competition in their particular labourmarkets for skills specific to their industries. More generally, resource- andknowledge-based theories of the firm suggest that heterogeneity of firms’ trainingresponses to external skill shortages is to be expected. This is because the ability ofany firm to provide training will be strongly conditioned by the specific resources andcapabilities (such as management skills and training capacity) which it hasaccumulated over time (Teece, Pisano and Shuen, 1997; Eisenhardt and Martin,2001; Teece, 2007).2.6. SummaryThis brief overview of the literature suggests a number of avenues for us to pursue inour analysis. First, we need to differentiate between firms’ general recruitmentactivity and it having skill shortage vacancies. Labour turnover is a basic fact ofeconomic life. Fast-growing and large firms will be more likely to be seekingadditional staff. Second, we need to distinguish between firms experiencingrecruitment difficulties because of the availability of labour with the appropriate skillsand those for other reasons, such as the firm merely not offering good enough payand conditions. Failing to account for this may cause us to misdiagnose the problemand will therefore undermine the appropriateness of any policy prescription. In any12 We examine the relationship between training and skill shortages in more detail in Mason, Mok,Stevens and Timmins, (2012). 10
  • 21. market there will be a number of potential customers that cannot purchase all theywould wish because they are unwilling or unable to pay the market price.Our quick survey of the literature suggests a number of patterns we might expect tosee, or hypotheses we might test:Firms are more likely to suffer SSVs the greater the proportion of higher skilled staffthey are looking for. Firms operating in tighter local labour markets will find it moredifficult to fill roles. Firms with high-productivity strategies will have a higher demandfor skills, but will be more able to pay skill premia. Firms with a degree of marketpower are likely to be more able to pay skill premia relative to their competitors.From a policy perspective, we wish to know whether these skill shortages are ‘real’ ornot. In any market, there will always be potential buyers who would wish to buy more,but are unwilling to do so at the current price. We know that labour markets doadjust supply to meet demand instantaneously; in the short run, the labour supplycurve is likely to be almost vertical13. Thus, a positive shock to demand (caused, forexample, by a firm getting a large export order) is not likely to create aninstantaneous increase in the supply of labour. In the short run, the effect may onlybe to increase wage inflation, as the firm outbids its competitors. In some cases thefirm may source skilled workers internationally, but this takes time and resources.This suggests two alternative hypotheses, one for a labour market with some slackand one where the supply of labour is tight. If the market is working satisfactorily,firms that offer higher wages will be less likely to experience SSVs. If there areshortages of skilled labour and the firms with the highest demand for skills are thosethat are the most productive, it may be the firms that pay the highest wages that haveSSVs.As we shall see below, because we only have a cross-section of data relating toSSVs, it is likely to be difficult to distinguish some of these phenomena.Nevertheless, we shall examine both the contemporaneous relationship and thatbetween current SSVs and previous years’ firm characteristics.13 there may of course be some economically inactive workers that are on the margins of entering thelabour market, such as single parents for whom the costs of childcare may raise the reservation wagehigher than their contemporaries 11
  • 22. 3. Data and Preliminary AnalysisThe data come from Statistics New Zealand’s prototype Longitudinal BusinessDatabase (LBD). The LBD is built around the Longitudinal Business Frame (LBF), towhich are attached, among other things, Goods and Services Tax (GST) returns,financial accounts (IR10) and aggregated Pay-As-You-Earn (PAYE) returns, allprovided by the Inland Revenue Department (IRD). The full LBD is described inmore detail in Fabling, Grimes, Sanderson and Stevens (2008) and Fabling (2009).The survey data considered in this chapter relate to the Business Operations Survey(BOS).The administrative data we use have four sources: the Linked Employer EmployeeDatabase (LEED), the Business Activity Indicator (BAI) dataset of GST returns, theAnnual Enterprise Survey (AES) and IR10 forms. These are described in more detailin the Data Appendix.3.1. Business Operations Survey (BOS)The Business Operation Survey (BOS) is an annual three part modular survey, whichbegan in 2005. The first module is focussed on firm characteristics and performance.The second module alternates between biennial innovation and business use of ICTcollections. The third module is a contestable module that enables specific policy-relevant data to be collected on an ad hoc basis14. The BOS is conducted using two-way stratified sampling, with stratification on rolling-mean-employment (RME) andtwo-digit industry according to the ANZSIC system15. The survey excludes firms withfewer than six RME and firms in the following industries: M81 GovernmentAdministration, M82 Defence, P92 Libraries, Museums and the Arts, Q95 PersonalServices, Q96 Other Services, and Q97 Private Households Employing Staff. The2008 survey achieved an 81.1% response rate (after adjusting for ceases), a total of5,543 usable responses, representing a population of 36,075 firms.The BOS is something approaching best practice in such surveys internationally. Ithas removed replication of surveys16 – and thus reduces respondent load and makes14 In 2005 and 2009 this was a ‘Business Practices Module’ and in 2006 an ‘Employment PracticesSurvey’. The 2007 module was on ‘International Engagement’.15 Note that there was some minor additional stratification conducted at the three-digit level.16 Prior to the BOS, surveys tended to occur on a fairly ad hoc basis – one assumes when policy-makers were considering a particular issue. Thus there was a Business Practices Survey in 2001, an 12
  • 23. sampling simpler. It is explicitly designed with a panel element; enabling moresophisticated analysis to be undertaken allowing us to better understand issues ofcausality and – as the panel element increases – dynamic issues17.We do not use SNZ-imputed values in cases of item non-response where it isimpossible to obtain them by simple edit rules (e.g. more than one expenditurecategories are missing).When we discuss questions from the BOS, we shall denote the question by theirmodule followed by their question number. So, for example, question 15 of Module Cis denoted C15. When there are multiple categories of response in the samequestion (e.g. the different occupational groups in question C18, we use the codeincluded on the form (e.g. C1801). Question numbers will relate to the 2008 BOS,unless otherwise specified.The ‘Business Strategy and Skills’ (BSS) ModuleThe Business Strategy and Skills (BSS) module of the 2008 Business OperationsSurvey was produced as part of the ‘Impact of Skills on New Zealand Firms’ project.This project involved the Ministry of Economic Development, the Department ofLabour, New Zealand Treasury and the Ministry of Research, Science andTechnology and was partly funded by the Cross Departmental Research Pool. Themodule was designed by the project team in conjunction with Statistics New Zealandand Geoff Mason, from the National Institute of Economic and Social Research inLondon.In this section we consider all firms that report vacancies of any kind and focus in onhard-to-fill vacancies and what we call skill shortage vacancies (SSV). The skillshortage vacancies are defined as vacancies that were hard-to-fill because theapplicants lacked the necessary skills, qualification or experience, which is a subsetof hard-to-fill vacancies. As there are other reasons for having hard-to-fill vacancies,this allows us to distinguish the non-skill-related vacancies from the skill shortagevacancies.Innovation Survey in 2003 and a Business Finance Survey in (2004). Elements of each of these areconsidered either every year as part of the Business Performance Module (Module A) or every two ormore years (i.e. the Innovation Module is run every other year and the Business Practices Module wasrun in 2005 and is scheduled to repeat in 2009).17 The panel element is in fact larger than it first seems as there is considerable overlap with previoussurveys, such as the 2001 Business Practices Survey (Fabling, 2007). 13
  • 24. The overall percentages of firms reporting each type of vacancy are depicted asconcentric circles in Figure 2. More detail on the construction and patterns of ourmeasures of vacancies, skill and non-skill-related shortage vacancies are set out inthe following sections.Figure 2 Vacancies, hard-to-fill and skill shortage vacancies Skill shortage vacancies (36%) Hard-to-fill vacancies (49%) Vacancies (77%) Figure shows the percentage of firms that report each type of vacancy Figures based in sample strata and weights Note that figures for the percentage of businesses with vacancies and hard-to-fill vacancies will not match the tables in the Statistics New Zealand Hot of the Press release because: (a) we use a slightly different sample and (b) we do not use imputed values3.2. VacanciesRespondents were asked: ‘During the last financial year, has this business had anyvacancies?’ (C14). The first row of Table 1 summarises these data. We shall returnto this table when we discuss hard-to-fill and skill shortage vacancies below.Overall, 76.6% of firms reported that they had posted a vacancy. As one mightexpect (given the greater number of employees), the likelihood of posting a vacancyincreases with firm size. 14
  • 25. Table 1 Vacancies, hard-to-fill, skill and non-skill-related shortage vacancies, % Business size Overall E <20* 20≤E<50 50≤E<100 E≥100Vacancies 71.8 89.5 93.6 95.2 76.6Hard-to-fill Vacancies 44.2 58.8 65.2 73.4 48.7Skill Shortage Vacancies 32.1 43.1 48.6 58.6 35.7 Table shows percentage of firms reporting each type of vacancy Figures based in sample strata and weights Business size (E) is measured by rolling mean employment, or RME. Note that the figure for business size being fewer than 20 RME is not all firms in the total business population with fewer than 20 RME, but rather firms in the BOS sample. For more on these see the Data Appendix. Note that figures for number of businesses with vacancies and hard-to-fill vacancies will not match the tables in the Statistics New Zealand Hot of the Press release because: (a) we use a slightly different sample; (b) we do not use imputed values; and (c) we use rolling mean employment (RME) from the 2008 financial year, rather than 2007.We can break the reporting of vacancies down by occupation. Respondents thatreported they had posted vacancies in the last year were asked a follow-up question:‘During the last financial year, how many vacancies has this business had for thefollowing roles?’ (C15). The responses to this question are set out in Table 2.It is for ‘clerical, sales and services workers’ that the greatest proportion of firms hadvacancies, followed by ‘labourers, production, transport or other workers’. Thisreflects the greater number of staff in these occupations 18. This is not quite trueacross all firm sizes. ‘Managers’ is the second most popular category for firms withmore than one hundred employees (and also, marginally, for those with between 50and 99 employees).18 See Figure 6 of Stevens (2012) 15
  • 26. Table 2 Businesses reporting vacancies, % Business size Overall E<20* 20≤E<50 50≤E<100 E≥100Managers 10.2 23.2 40.1 62.4 15.8Professionals 11.5 19.1 29.1 42.6 14.8Technicians and associate profs 8.7 19.3 25.7 38.5 12.4Tradespersons and rel. workers 20.4 25.5 27.5 36.3 22.1Clerical sales and service workers 28.6 45.2 58.7 72.8 34.5Labourers, production, transport or 26.2 41.1 47.3 53.3 30.7other workersAll occupations 68.5 86.1 89.4 88.0 73.1 Table presents data from questions C14: ‘During the last financial year, has this business had any vacancies?’ and C15: ‘During the last financial year, how many vacancies has this business had for the following roles?’ Table shows percentage of firms reporting each type of vacancy Figure for ‘all occupations’ does not match that for Table 1 because some firms do not report the occupations of their vacancies Figures based in sample strata and weights Business size (E) is measured by rolling mean employment, or RME. * Note that the figure for business size being fewer than 20 RME is not all firms in the total business population with fewer than 20 RME, but rather firms in the BOS sample. For more on these see the Data Appendix. Note that figures for number of businesses with vacancies and hard-to-fill vacancies will not match the tables in the Statistics New Zealand Hot of the Press release because: (a) we use a slightly different sample; (b) we do not use imputed values; and (c) we use rolling mean employment (RME) from the 2008 financial year, rather than 2007.3.3. Hard-to-fill vacanciesRespondents were asked: ‘During the last financial year, was this business easilyable to fill all vacancies with suitable applicants?’ (C16). Those who answered ‘no’ tothis question were classified as having a hard-to-fill vacancy. The second row ofTable 1 (repeated at the bottom of Table 3) summarises these data. Well over half ofthe firms that have vacancies find them hard-to-fill (47.9% compared to 76.6%).Again, the probability of having a hard-to-fill vacancy increases with firm size, withalmost three-quarters of firms with rolling mean employment of one hundred or morehaving hard-to-fill vacancies. 16
  • 27. Table 3 Businesses reporting hard-to-fill vacancies, % Business size Overall E<20* 20≤E<50 50≤E<100 E≥100Managers 4.9 10.1 16.7 29.4 7.2Professionals 7.1 11.6 16.1 24.8 9.0Technicians and associate profs 4.6 8.4 11.3 19.4 6.1Tradespersons and related workers 13.7 16.3 15.1 19.8 14.4Clerical sales and service workers 10.2 14.2 18.1 24.6 11.8Labourers, production, transport orother workers 12.3 17.6 19.9 22.9 13.9All occupations 44.2 58.8 65.4 73.6 48.7 Table presents data from questions C16 ‘During this last financial year, was this business easily able to fill all vacancies with suitable applicants?’ and C18: ‘Mark all that apply/ for this business, which roles were hard-to- fill?’ Table shows percentage of firms reporting each type of vacancy Figure for ‘all occupations’ does not match that for Table 1 because some firms do not report the occupations of their hard-to-fill vacancies Figures based in sample strata and weights Business size (E) is measured by rolling mean employment, or RME. Note that the figure for business size being fewer than 20 RME is not all firms in the total business population with fewer than 20 RME, but rather firms in the BOS sample. For more on these see the Data Appendix. Note that figures for number of businesses with vacancies and hard-to-fill vacancies will not match the tables in the Statistics New Zealand Hot of the Press release because: (a) we use a slightly different sample; (b) we do not use imputed values; and (c) we use rolling mean employment (RME) from the 2008 financial year, rather than 2007.Respondents that reported that they found some vacancies hard-to-fill were asked:‘For this business, which roles were hard-to-fill?’ (C18). ‘Tradespersons and relatedworkers’ were the occupations with which most businesses had recruitmentdifficulties (Table 3). This reflects a more even spread of hard-to-fill vacancies byfirm size and hence the influence of the greater number of small (6-19 employees)firms.‘Managers’ were the role for which most large (100+) firms found difficult to fillvacancies. Given that managers represent a relatively small proportion of total staff,and one that has an important impact on firm performance (Bloom and Van Reenen,2007, 2010; UTS, 2010), this is a worrying result.Comparison with Recruitment Difficulties in Module AIn order to place these figures in context, it is useful to compare these results with theresponses to a similar question that is asked in Module A. In a section onemployment, firms are asked ‘Over the last financial year, to what extent did thisbusiness experience difficulty in recruiting new staff for any of the following 17
  • 28. occupational groups?’ (A33). The different language used in module C (vacancy plushard-to-fill vacancy) and Module A (‘difficulty in recruitment’), differences in theresponse categories and the influence of previous questions mean that these are notidentical19. The results of such a comparison are presented in Appendix A2. Insummary, there is a high degree of correlation between the two measures, but thatthis is not total. The majority of respondents that said they had a severe difficulty inrecruiting each type of staff in Module A said that vacancies were hard-to-fill inModule C.The Persistence of Recruitment DifficultiesNotwithstanding the differences between the Module A recruitment difficultiesvariables and the Module C hard-to-fill vacancies variable, we can use the Module Aquestion to consider the persistence of recruitment difficulties. In order to prevent usbecoming overwhelmed with tables, we focus on the year-to-year rank correlationsbetween the responses to the recruitment difficulty question (A33) for the 2005, 2006,2007 and 2008 surveys.As we can see from Table 4, the correlations are both high and significant.Interestingly, whilst the correlation does decline over time, this decline is not verysteep. There may be a number of reasons for this. One might be that firms withrecruitment difficulties are more likely to have had them in the past, but notnecessarily every year. Another may be a function of the attrition in the survey, asfirms enter and leave.Whatever the reasons for the particular pattern, there is clearly considerablepersistence in the existence of recruitment difficulties. We shall consider thepredictive power of previous recruitment difficulties for the probability of the firmposting a vacancy in our econometric analysis in the following section.19 For more on the subject of how respondents interpret and answer questions in business surveys,with particular reference to the BOS, see Fabling, Grimes and Stevens (2008, 2012). 18
  • 29. Table 4 Year-to-year rank correlation between reported recruitment difficulties 2005 2006 2007 2008 2005 2006 2007 2008Managers and professionals Tradesmen and related occupations2006 0.5836*** 1 0.4827*** 1 (0.0000) (0.0000)2007 0.4522*** 0.5402*** 1 0.4506*** 0.4592*** 1 (0.0000) (0.0000) (0.0000) (0.0000)2008 0.5151*** 0.5225*** 0.5450*** 1 0.3668*** 0.4462*** 0.6254*** 1 (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)Technicians and assistant professionals Other occupations ***2006 0.4534 1 0.3357*** 1 (0.0000) (0.0000)2007 0.4225*** 0.5687*** 1 0.3105*** 0.3874*** 1 (0.0000) (0.0000) (0.0000) (0.0000)2008 0.4161*** 0.4183*** 0.6455*** 1 0.3011*** 0.3362*** 0.4067*** 1 (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) p value in parenthesis * significant at 10%; ** significant at 5%; *** significant at 1%3.4. Skill and non-skill-related shortage vacanciesRespondents that had hard-to-fill vacancies were asked ‘For which of the followingreasons did this business find it hard-to-fill vacancies?’ (question C17). They weregiven twelve categories, from which they could choose as many as they wished.Those that replied ‘applicants lack the work experience the business demands’ or‘applicants lack the qualifications or skills the business demands’ were defined ashaving skill shortage vacancies (SSVs) 20 . Around the same numbers of firmsreported each of these reasons across all firm sizes (Table 5). However, only aroundhalf of businesses that reported that applicants lacked appropriate experience alsoreported that they lacked the appropriate qualifications or skills (and vice versa).Thus, we should be careful as considering these two explanations as synonymous.Table 5 shows that there are attributes workers obtain through their on-the-jobexperience that businesses value over and above qualifications and, in particular,separate from the things that they call ‘skills’. It may be that experience connotes adepth of skill rather than merely its presence.20 For the full list of responses, see SNZ (2008). 19
  • 30. Table 5 Skill-related reasons for hard-to-fill vacancies, by size Business Size RME 20 ≤ RME 50 ≤ RME 100+ Total <20 <50 <100 RMEApplicants lack work experience the 25.4 34.7 37.9 48.1 28.4business demandsApplicants lack qualifications or skills 25.1 35.2 38.3 47.5 28.2the business demandsLack either qualifications or work 13.7 16.2 20.9 21.8 14.7experienceLack both qualifications and work 18.4 26.8 27.7 36.8 20.9experience Table presents data from questions C17 ‘For which of the following reasons did this business find it hard-to-fill vacancies?’ Table shows percentage of firms reporting each response Figures based in sample strata and weights Business size (E) is measured by rolling mean employment, or RME. Note that the figure for business size being fewer than 20 RME is not all firms in the total business population with fewer than 20 RME, but rather firms in the BOS sample. For more on these see the Data Appendix. Note that figures for number of businesses with vacancies and hard-to-fill vacancies will not match the tables in the Statistics New Zealand Hot of the Press release because: (a) we use a slightly different sample; (b) we do not use imputed values; and (c) we use rolling mean employment (RME) from the 2008 financial year, rather than 2007.Firms that replied having other reasons besides ‘applicants lack the work experiencethe business demands’ or ‘applicants lack the qualifications or skills the businessdemands’ were defined as having non-skill-related shortage vacancies (NSR). Weshall use this information allows us to better understand what is unique about skill-shortage vacancies than other recruitment difficulties that would have differentimplications for policy.3.5. Internal skill gapsIn this section, we explore the internal skill gaps from the BOS 2008 and skillimprovement needs obtained from the computer-aided telephone interview (CATI).From the BOS, respondents were asked ‘How many of this business’s existing staffhave the skills required to do their job?’ for 6 given occupation groups of managers;professionals; technicians and associate professionals; tradespersons and relatedworkers; clerical, sales and service workers; and labourer, production, transport orother workers. Respondents were further asked the reasons for their staff not havingthe skills required. Table 6 shows that more than half of the firms have all theirmanagers and clerical, sales and service workers with skills required. Internal skill 20
  • 31. gaps are more prominent for less skilled workers, such as clerical, sales and serviceworkers and labourers, production, transport and other workers.Table 6 Existing staff with skills required to do their job Proportion of staff with skills (%)Occupation Less than half Half or more All staffManagers 9.8 11.7 78.5Professionals 4.4 7.3 37.7Technicians and associate professionals 3.3 9.0 33.0Tradespersons and related workers 4.4 16.1 32.4Clerical, sales and service workers 4.4 20.3 60.6Labourers, production, transport and others 4.2 19.0 36.6 Table presents data from questions C20 ‘How many of this business’s existing staff have the skills required to do their job?’ Source: SNZ (2009) Note that the percentages in this table are taken from ‘SNZ (2009) and are not exactly comparable with the Table 1 and Table 2 and Figure 2. For more on this, see the footnotes to Table 1 and Table 2.4. MethodIn this section we outline our empirical strategy. First we set out our basiceconometric model. This is a two stage, ‘Heckman selection’ model, where we allowfor the interrelationship between the firm’s decision to look for staff externally andpost a vacancy and the probability that these vacancies prove to be hard-to-fill forskill-related reasons.21In the following two sections we consider the variables included in each stage of themodel. One of the key issues we face is that we only have one year of informationon our dependent variable, so our model is essentially a cross-sectional model. Thismakes interpreting coefficients as causal relationships, rather than merely uncoveringstatistically significant correlations, difficult. This is a problem faced by many studiesof this nature (e.g. Mason and Stevens, 2003). We seek to overcome (or at leastmitigate) this by using the fact that the Business Operations Survey is embedded inthe prototype Longitudinal Business Database (LBD). We use additional data from21 An earlier version of this paper (Mason et al., 2010), shows the importance of accounting for therelationship between in this way. In unpublished work we also considered using multinomial logit andbivariate probit models, but concluded that these were not appropriate also. 21
  • 32. previous years’ BOS and information on the firm we have from other sources,specifically tax and survey data.Another important issue to bear in mind is the fact that the unit of observation in ourempirical analysis is the firm (in LBD parlance, the enterprise). Thus, when weestimate the probability of a skill shortage vacancy, we are estimating the probabilityof a firm with a vacancy having at least one that is hard-to-fill for skill-related reasons.We are not estimating the probability of a particular vacancy being hard-to-fill for skillrelated reasons. For this we would need vacancy-level information. Our focus is onthe impact on firms rather than on individual vacancies.Finally, in order to aid interpretation, we also supplement our analysis of SSVs byconsidering different specifications of our dependent variable (whether the firm hasan SSV): we investigate specifications where the outcome is restricted to firms thathave SSV only (i.e. not for other reasons)4.1. Basic econometric modelThe primary model we estimate is the probability of reporting skill shortage vacanciesaccounting for sample selection. We model the first stage (the selection equation) asbeing the probability of reporting having posted a vacancy. The second stage is theprobability that a firm with vacancies finds at least one of them hard-to-fill for reasonsof skill. We also investigate different outcomes in the second stage equations (i.e.non-skill-related hard-to-fill vacancies and various variations on the SSV/NSR theme).The discussion of method that follows, however, focuses on the model for SSVs, tokeep the exposition clear.We suppose that the propensity of firm i to post a skill shortage vacancy can beexpressed as(1) SSVi   X i    1i where SSV* is the propensity to have a skill shortage vacancy, Xi is a (1×k) vector of kexplanatory variables,  is a (k×1) vector of parameters to be estimated and  i anerror term. We do not observe the SSV* terms, but the binary realisation of them.Therefore we assume: 22
  • 33. SSVi  0 if SSVi   0(2) SSVi  1 if SSVi   0This variable SSVi takes one of three values. If the firm reports a skill shortage, SSVtakes the value of 1. It takes the value zero if the firm does not have a skill shortagevacancy, but does have a vacancy. It takes a missing value when the firm does nothave any vacancies.The dependent variable in (2) for observation i is observed if:(3) Vaci  Z i   2i  0where 1 ~ N 0,1(4)  2 ~ N 0,1 corr 1 ,  2   When  ≠ 0, standard probit techniques applied to (2) yield biased results. Theselection model provides consistent, asymptotically efficient estimates for all theparameters in the model. For the model to be identified, we should have at least onevariable in the selection equation (3) that is not in the outcome equation.4.2. Selection equation – The likelihood of vacanciesIn our selection equation (the probability of reporting a vacancy), we include thefollowing variables in the X vector22. Scale will be an important issue; clearly themore employees a firm has, the more likely it is to have lost one and need toadvertise for a replacement. We measure firm size by the (natural) logarithm ofrolling mean employment plus a count of working proprietors (from the LinkedEmployer-Employee Database or LEED), ln(E). This highlights an important linkbetween the workers in a firm and the vacancies it posts. The more workers that quitor are fired from a firm, the more it must replace. We include a direct measure of thiswith the separation rates (the total number of employees that leave a firm in a yeardivided by the number of employees, again from the LEED), SepRate.It may be the case that occupations differ in their likelihood of quitting (because theyhave different outside options, for example). Therefore we include variables for the22 For more information on the variables see the Data Appendix to this paper. 23
  • 34. proportions of staff that are ‘managers’, ‘professionals’, ‘associate professionals ortechnicians’ or ‘tradespersons’, Propman to Proptrade.Firms that are already experiencing internal skill gaps may be more likely to attemptto fill these via external recruitment. Therefore, we include a variable that indicateswhether the firm feels that their staff lack the skills to perform their roles, ISG.We consider the firms productivity levels using both (the natural logarithm of) theirlabour productivity, ln(LP), and multifactor productivity, MFP. The latter is generatedas the residual from a series of industry-level production functions estimated by OLS.Relative wagesWorkers choose to accept a job offer contingent on their knowledge of all alternativeoffers. However, if their wages slip behind other alternatives (because wages in thefirm grow more slowly or wages in similar firms grow more rapidly), workers mayleave23. We therefore include the change in wages paid by the firm relative to otherfirms in the same industry, j, in the same region, r, (we discuss the calculation of thisin more detail in section 4.3 below), [ln(Wi) – ln(Wijr)].It may be the case that firms with market power in their product market have greaterscope to compete for scarce skilled labour (their supernormal profits are, in effect,shared between the owners of capital and labour). Because of this, we also includea dummy variable to indicate whether the firm is effectively a monopoly (Monopoly).Business StrategyTo investigate the relationship between he firms’ business strategy and the demandfor skills, we include a set of variables to indicate the business strategy of the firm.The variables we consider are: whether it was able to obtain a higher price than itscompetitors for its goods and services (Price_Leader); whether it developed orintroduced any new or significantly improved goods or services, operationalprocesses, organisational/ managerial processes or marketing methods (Innovate);whether it undertook research and development (R&D); whether its primary marketwas international, rather than local or national (Mark_Inter); whether it providedcustomised goods or services (Customise); whether it exported (Export); or whether itundertook outward foreign direct investment (ODI).23 For a discussion on the influences on individual’s decisions to quit, see Stevens (2005). 24
  • 35. There are a number of factors that make it difficult to include the whole suite ofpotential business strategy variables from the BOS in our analysis. First, we onlyhave a single cross-section of information on our dependent variables. Second, thebusiness strategy variables are highly correlated with each other. Third, businesspractices are highly correlated with firm performance and, thus, productivity and thefirm’s ability to pay higher wages than its competitors. Fourth, they are all binary orother forms of categorical variables. These combine to make analysis difficult andpotential make our estimated coefficients erratic and even counter-intuitive. Becauseof these issues, we also performed factor analysis and took the factor that had anEigenvalue of one 24 (which in our econometric analysis will be called BS1). Theresults of our factor analysis are set our it Table 7 and Table 8.Table 7 Principal component factor analysis of business strategy variablesFactor Eigenvalue Difference Proportion CumulativeFactor1 1.03421 0.79462 1.3842 1.3842Factor2 0.23959 0.23876 0.3207 1.7049Factor3 0.00083 0.06736 0.0011 1.7060Factor4 -0.06653 0.01844 -0.0890 1.6169Factor5 -0.08497 0.07256 -0.1137 1.5032Factor6 -0.15753 0.06092 -0.2108 1.2924Factor7 -0.21845 -0.2924 1.0000 LR test: independent vs. saturated: chi2(21) = 2374.58 Prob>chi2 = 0.0000Table 8 Factor loadings (pattern matrix) and unique variancesVariable Factor1 (BS1) Factor2 Factor3 UniquenessMark_Inter 0.4563 -0.2085 0.0097 0.7482Customise 0.1627 0.2250 0.0119 0.9228Price_Leader 0.2047 0.2074 0.0150 0.9148Innovate 0.2598 0.2691 -0.0100 0.8600R&D 0.4721 0.0954 -0.0127 0.7678Export 0.5731 -0.1229 0.0047 0.6565ODI 0.3725 -0.0765 -0.0093 0.8553The panel nature of the BOS allows us to consider the firms business strategy inyears prior to the current year’s vacancies (and SSVs). However, some of thevariables (i.e., those in Module C) are not available. Therefore, we consider adifferent set of variables and also undertake a separate factor analysis of these. The24 The so called ‘Kaiser-Guttman rule’ after Guttman (1954) and Kaiser (1970). 25
  • 36. variables we take from the 2007 BOS included the three variables from Module A inthe above analysis: whether they conducted R&D (R&D07), exported (Export07), orundertook ODI (ODI07). Because we have access to the innovation module B in 2007,we are able to divide their innovation activity into the four constituent parts, that is,new or significantly altered: products (prod_inno07), processes (proc_inno07),organisation (org_innov07) and marketing (mark_inno07). The results of our factoranalysis are set out in Table 9 and Table 10.Table 9 Principal component factor analysis of 2007 business strategyFactor Eigenvalue Difference Proportion CumulativeFactor1 1.49186 1.03350 1.1324 1.1324Factor2 0.45836 0.47050 0.3479 1.4803Factor3 -0.01214 0.08671 -0.0092 1.4711Factor4 -0.09885 0.03521 -0.0750 1.3961Factor5 -0.13406 0.01513 -0.1018 1.2943Factor6 -0.14918 0.08937 -0.1132 1.1811Factor7 -0.23855 . -0.1811 1.0000 LR test: independent vs. saturated: chi2(21) = 3637.14 Prob>chi2 = 0.0000Table 10 Factor loadings (pattern matrix) and unique variances Factor1Variable Factor2 Uniqueness (BS07)R&D07 0.4245 0.3145 0.7208Export07 0.3133 0.3712 0.7641ODI07 0.2351 0.2875 0.8621prod_innov07 0.5866 0.0259 0.6552proc_innov07 0.5428 -0.1862 0.6707org_innov07 0.5153 -0.2451 0.6744mark_innov07 0.5039 -0.2088 0.70254.3. Probability of SSVsIn addition to some of the control variables included in the selection equation, wealso investigate the relationship between SSVs and relative wages, the local labourmarket and types of staff for whom firms are seeking to hire. 26
  • 37. WagesOne might expect a major determinant of firms’ ability to fill any vacancy to be thewages it pays relative to other potential employers. We have discussed above whythis might not be as clear as it first seems and so the sign on the coefficient ofrelative wages will tell us something about the function of the New Zealand labourmarket.In considering the appropriate measure of outside wages, there are a number offactors to take into account. The unit of observation for the data on wages is theestablishment. We do not have information on wages by skill level. Our measure ofwages is the total wage bill divided by the number of employees on the 13 th of themonth. This can be aggregated from a monthly establishment measure to an annualenterprise measure using employment weights (or, equivalently, by summing wagecosts and rolling mean employment over the year and dividing one by the other)25.The two dimensions over which we can measure variation in the outside wage areindustry classification and geography. However, firms often operate in multipleindustries and/or multiple locations. Firm level analysis generally uses the concept of‘predominant industry’. This is typically measured by taking the industry in which thefirm employs the most workers. Thus, if a firm has two establishments, oneemploying 100 workers constructing boats and one employing 20 to sell marinesafety equipment, we say the firm is a boat-maker. If the firm has establishments inmore than one region, we could proceed by analogy to determine its ‘predominantregion’. This is the course followed by Grimes, Ren and Stevens (2012)26. Becausewe have information of the numbers of employees and wages at the establishmentlevel, along with the region in which that establishment is situated and its ANZSICcode, we can calculate a more sophisticated measure. We calculate the relativewage at the establishment level and aggregate it across industries and locations inwhich it operates using employment weights.In the analysis for this paper, we actually considered three outside wage measures:the average wage in the 4-digit industry; the average wage in the local region and;25 As with other measures using LEED employment, this measure does not take into account thehours worked by employees (we cannot calculate hourly wages), but it does take into account monthlyvariations within the year (e.g. seasonal working).26 This is because in the questions on internet connectivity in the 2006 Business Operations Surveyused by Grimes et al., respondents are answer with respect to their largest operation. 27
  • 38. the average wage of firms in the industry in the region. However, the measures arehighly correlated with each other at the enterprise level and the influence on theresults is marginal. Therefore, in this paper, we present results from only theindustry-regional measure as it reflects both the national market for skills of a similarnature and the geographic area.We chose the Territorial Authority as the appropriate region as we believe that thismost closely approximates a travel-to-work area27. In order to maintain cell sizes, thedefinition of industry on which we settled was the 2-digit level according to theANZSIC 1996 classification. In cases where the number of firms in a cell droppedbelow 50, we used data at the next level of industry aggregation.The industry-region outside wage for an enterprise (the unit of observation in the LBDthat broadly corresponds with what economists would call the firm) was calculated asthe average wage in the industry-region cell in which the establishments in theenterprise operated, aggregated using establishment employment weights. That is,the outside wage of enterprise i, consisting of L establishments, is given by L M Ei(5) Woi   W jrm l 1 m 1 Eikmwhere Wjrm is the average wage of the establishments in the 2-digit industry j in theTerritorial Authority r in month m (M will typically, but not exclusively, be 12), Ei is thetotal employment in the enterprise and Eikm is the employment in each establishmentin a given month.Local Labour Market ConditionsWe consider three measures of the local labour market conditions (beyond theindustry-region wage). These come from the 2006 Census of Population andDwellings. The first two are measures of the supply of labour at the two ends of theeducation distribution. First, we consider the percentage of the population with adegree or higher qualification. Second, we consider the percentage of the populationwith no qualifications. The third variable we consider is the unemployment rate in theTerritorial Authority.27 It is possible to calculate more sophisticated measures of outside wages and local labour marketconditions using more disaggregated (Area Unit or even Meshblock) data and weighting it by somefunction of the distance between the two areas. We leave this to later work. 28
  • 39. Because these variables are highly correlated, we perform a factor analysis. Thefactor analysis produces two primary factors with Eigenvalues of greater than one. Inorder to facilitate interpretation, we undertake factor rotation and obtain the twofactors (Census1 and Census2) as set out in Table 11 to Table 13. The first factor isprimarily made up of the percentage of the population with no qualifications andunemployment rate variables. The second is a combination of the percentage withdegree and unemployment rate.Table 11 Principal component factor analysis of census variablesFactor Eigenvalue Difference Proportion CumulativeFactor 1 1.50331 0.27826 0.5011 0.5011Factor 2 1.22505 0.95341 0.4083 0.9095Factor 3 0.27164 . 0.0905 1.0000 LR test: independent vs. saturated: chi2(3) = 3782.54 Prob>chi2 = 0.0000Table 12 Rotated factor analysisFactor Eigenvalue Difference Proportion CumulativeCensus1 1.41023 0.09211 0.4701 0.4701Census2 1.31812 . 0.4394 0.9095 Rotation method: orthogonal varimaxTable 13 Rotated factor loadings (pattern matrix) and unique variancesVariable Census1 Census2 Uniqueness% no qualifications 0.9318 -0.2150 0.0854% degree or higher -0.1096 0.9553 0.0754% unemployment 0.7279 0.5995 0.1108Occupational breakdown of vacanciesGiven the specific nature of many skills, an important element of the ease with whichfirms will be able to vacancies will be for whom they are seeking. We thereforeinclude four variables for the proportion of vacancies that are for the followingoccupational groups: managers (Vman), professionals (Vprof), technicians and associateprofessionals (Vtech) and tradespersons and related workers (Vtrade). The baseline is a 29
  • 40. combination of clerical sales and services workers and labourers, production,transport or other workers28.5. ResultsOur results are set out in three sections. In the first section (5.1) we examine SSVsusing contemporaneous variables. These will show us which types of firms arecurrently experiencing SSVs. Next we consider lagged values of the BOS and otherLBD variables (5.2). This allows us to look at the characteristics of firms prior to theyear in which they reported SSVs. Finally, we consider different specifications of thedependent variable to give some contextual information to help understand our mainresults (5.3).5.1. Using contemporaneous variablesThe results from our estimation using contemporaneous variables are set out inTable 14 and Table 15. The first two columns show the impact of our industrydummies, with column (1) excluding industry dummies and the subsequent columnsincluding them. In column (3) we introduce internal skill gaps and our businessstrategy variables. Column (4) replaces the strategy variables with the principalfactor and the final column presents a more parsimonious specification.Table 14 presents the results for the selection equation of our contemporaneousspecifications. Comparison of (1) and (2) suggests that the impact of our industrydummies on the remaining coefficients is relatively minor.In general, larger firms and those that are growing are more likely to be posting avacancy. The separation rate of the firm has no more predictive power beyond thetotal number of employees in explaining the variation in the probability of posting avacancy.Firms that are increasing their wages faster than their competitors tend to be morelikely to have a vacancy, although this is not statistically significant in allspecifications. Note, however, that in this contemporaneous specification this mayreflect firms raising wages in response to staff turnover or firm expansion. This issomething we return to when we consider our lagged specifications in section 5.228 We also included the first of these in preliminary analysis, but its coefficient was never statisticallysignificant. 30
  • 41. below. The productivity of the firm does not appear to help explain the probability of afirm posting a vacancy, ceteris paribus29.Firms that undertake training in the current year are more likely to have vacancies.Firms enjoying a monopoly in their product market and those with collectiveagreements are no more likely to have vacancies.None of the occupations appear to be any more likely to create a need for vacancies.When we account for the occupational make-up of the firms employment (in columns(1) to (3)), we find no relationship between this and the likelihood of posting avacancy. Because of this, we exclude the occupational variables from specifications(4) and (5).We introduce our measure of internal skill gaps (ISG) and business strategyvariables in column (3). Firms with internal skill gaps are no more likely to havevacancies, ceteris paribus. Firms that consider themselves to be price leaders andthose that have innovated are more likely to have vacancies, a result that isstatistically significant at the 1% level. Firms that undertake a high degree ofcustomisation and ODI are also more likely to have a vacancy, although these resultsare significant only at the 10% level.When we replace these binary variables with the factor (BS1), this is positivelycorrelated with the probability of having vacancies (columns (4) and (5)). This resultis significant at the 5% level.29 Note that we estimated a number of different specifications of the productivity variable and obtainsimilar results. We also estimated the models in Table 14 and Table 15 with both labour productivityand MFP, but the models with MFP either failed or experienced difficulty converging. We take this asa sign of a less well-specified model and so do not present the results. 31
  • 42. Table 14 Results - SSVs - Contemporaneous variables - Selection equation (1) (2) (3) (4) (5)ln(sales) *** *** *** *** *** 0.281 0.283 0.237 0.246 0.239 (0.071) (0.074) (0.075) (0.072) (0.068)[ln(Wi)–ln(Wijr)] * * ** 0.242 0.343 0.463 0.385 0.417 (0.273) (0.239) (0.268) (0.215) (0.203)SepRate 0.542 0.470 0.449 0.313 (0.447) (0.420) (0.401) (0.384) *** *** *** *** ***ln(E) 0.419 0.407 0.397 0.419 0.435 (0.049) (0.050) (0.052) (0.048) (0.044)ln(LP) 0.041 0.071 0.074 0.052 (0.049) (0.046) (0.052) (0.054) *** *** *** *** ***Train 0.821 0.831 0.821 0.860 0.852 (0.074) (0.097) (0.085) (0.080) (0.074)Union -0.002 -0.031 -0.038 (0.058) (0.049) (0.045)Monopoly -0.029 -0.057 0.009 (0.103) (0.096) (0.118)Propman 0.257 0.210 0.078 (0.439) (0.506) (0.440)Propprof -0.030 0.043 -0.075 (0.318) (0.323) (0.320)Proptech -0.215 -0.148 -0.256 (0.208) (0.242) (0.263)Proptrade 0.009 0.089 0.100 (0.126) (0.167) (0.130)ISG 0.002 0.017 (0.056) (0.047) ***Price_Leader 0.303 (0.060) ***Innovate 0.294 (0.074)R&D -0.068 (0.123)Mark_Inter 0.042 (0.161) *Customise 0.164 (0.088)Export 0.080 (0.121) *ODI 0.376 (0.196) ** **BS1 0.211 0.208 (0.085) (0.086) *** *** *** *** ***Constant -1.513 -1.816 -2.030 -1.652 -1.081 (0.489) (0.512) (0.595) (0.587) (0.108)Industry dummies      Robust standard errors in parentheses * ** *** significant at 10%; significant at 5%; significant at 1% 32
  • 43. Table 15 presents the coefficients in the outcome equation of our contemporaneousspecifications, i.e. the probability of an SSV. One clear result across thespecifications is that firms that pay higher wages than others in their industry andregion are more likely to be experiencing SSVs. This result is significant at the 5%level in the first two columns and at the 1% level in the remaining specifications.Once we have accounted for the probability of a firm reporting a vacancy, we find nostatistically significant relationship between the size of a firm in terms of employmentand the probability it reports an SSV.In contrast to our results in the selection equation for the probability of reporting avacancy, the occupational breakdown of vacancies is related to the probability thatthey are hard-to-fill for reasons of skill. Firms with a higher proportion of theirvacancies for professional, technicians and associate professional, and tradeoccupations are more likely to report SSVs. The proportion of vacancies that are formanagers is not significant. Because the coefficients on each variable are so similar(we can accept the restriction that they are identical) we replace them with a commonvariable for the proportion of vacancies which are for either professional, techniciansand associate professional or trades occupations (and drop the managementvariable) in columns (4) and (5).Of the two regional labour market variables, only Census2 is statistically significant (inall but column (2)). Firms operating in Territorial Authorities with a higher proportionof the population with degrees and higher unemployment are less likely to findvacancies hard-to-fill for reasons of skill. Regions in which there is both a highproportion of the population with no qualifications and unemployment are no easier tofill such vacancies.Firms with internal skill gaps also appear to also suffer external skill gaps in the formof SSVs. Of the business strategy variables, only the R&D variable is statisticallysignificant. When we replace the suite of business strategy variables with the BS1factor, it is statistically insignificant. 33
  • 44. Table 15 Results – SSVs – Contemporaneous variables – Outcome equation (1) (2) (3) (4) (5)ln(Wi) – ln(Wijr) 0.138 ** 0.146 ** 0.143 *** 0.194 *** 0.201*** (0.054) (0.063) (0.053) (0.061) (0.073)ln(E) 0.037 0.016 0.055 0.046 0.052 (0.049) (0.053) (0.047) (0.043) (0.040)Census106 -0.002 0.004 -0.001 -0.004 -0.001 (0.038) (0.035) (0.035) (0.038) (0.037) ** *** ***Census206 -0.073 -0.046 -0.080 -0.076 -0.070** (0.031) (0.030) (0.029) (0.027) (0.028)Union 0.115 0.121 0.108 (0.095) (0.093) (0.098)Monopoly -0.056 -0.019 -0.073 (0.075) (0.082) (0.079)Vman 0.091 0.120 0.145 (0.171) (0.187) (0.189) *** ***Vprof 0.518 0.459 0.534*** (0.164) (0.142) (0.166)Vtech 0.618*** 0.556*** 0.645*** 0.325*** 0.327*** (0.171) (0.158) (0.188) (0.068) (0.065) *** *** ***Vtrade 0.652 0.657 0.643 (0.115) (0.157) (0.110)ISG 0.117* 0.129** 0.139** (0.063) (0.059) (0.063) **R&D08 -0.261 (0.117)Innovate -0.006 (0.119)Mark_Inter -0.079 (0.109)Customise 0.070 (0.086)Price_Leader -0.081 (0.107)Export08 -0.039 (0.103)ODI08 0.061 (0.172)BS1 -0.082 (0.065)Constant -0.213 -0.100 -0.251 -0.175 -0.203 (0.239) (0.247) (0.311) (0.231) (0.180)athrho -0.610* -0.881* -0.535 -0.503 -0.424* (0.348) (0.469) (0.405) (0.315) (0.236)Prob  2 0.0795 0.0601 0.1862 0.1103 0.0723Observations 4485 4482 4335 4683 4821 See notes to Table 14 34
  • 45. Our estimate of  (or, rather, athro) is significant in specifications (1), (2) and (5) andborderline in (4). This suggests that in those specifications at least, it is important toaccount for the probability of posting a vacancy in the first place when one considerswhich types of firms are likely to report a SSV.5.2. Using lagged variablesThe results of our estimation using lagged variables are set out in Table 16 andTable 17. Once more, firms with more employees are more likely to have vacancies.Once we lag sales, we no longer see the positive relationship between sales growthand the probability of the firm posting a vacancy. Clearly, the relationship betweensales growth and vacancies are two sides of the same phenomena; as the firm growsit needs more labour input to produce the output and thus needs to post vacancies tofill the additional roles created. Because of the erratic nature of firm growth, asopposed to the level of sales, (Hull and Arnold, 2008) , current sales growth is not avery good predictor of future vacancies.Our training variable retains a similar sign if we use its current value or even if weuse a training variable taken from the 2006 Employment Practices Module C (column(10)).More productive firms are no more or less likely to post a vacancy in the followingyear. This result does not change whether we use labour productivity or MFP.In this analysis we consider the persistence of recruitment difficulties. Firms thatreported recruitment difficulties in 2007 are more likely to report a vacancy in 2008.This result is significant at the 1% level across all specifications. There is furtherevidence of persistence in recruitment difficulties when we include the second lag ofour RecDiff variable. Although the coefficient on the 2006 recruitment difficultyvariable is only significant at the 10% level, this affect is over and above that of the2007 variable. 35
  • 46. Table 16 Results - SSVs - Lagged variables - Selection equation (6) (7) (8) (9) (10)ln(sales)07 -0.027 -0.038 -0.038 -0.112 -0.102 (0.059) (0.062) (0.062) (0.070) (0.142)[ln(Wi)–ln(Wijr)]07 -0.105 -0.092 -0.150 -0.161 -0.405 (0.276) (0.266) (0.268) (0.280) (0.291)SepRate07 0.110 0.101 0.118 0.152 -0.326 (0.245) (0.269) (0.265) (0.309) (0.540) *** *** *** *** ***ln(E) 07 0.360 0.357 0.365 0.341 0.445 (0.049) (0.051) (0.053) (0.047) (0.059)ln(LP) 07 0.008 0.003 0.009 (0.059) (0.062) (0.062)MFP07 0.098 0.076 (0.100) (0.105) *** *** *** ***Train08 0.571 0.584 0.569 0.536 (0.104) (0.096) (0.106) (0.136) ***Train06 0.437 (0.126)Union07 0.252 0.212 0.292 0.291 (0.238) (0.251) (0.235) (0.283) *** *** *** *** ***RecDiff07 0.688 0.680 0.665 0.607 0.581 (0.073) (0.076) (0.077) (0.075) (0.096) *RecDiff06 0.220 (0.116)Monopoly07 -0.109 -0.104 -0.028 -0.037 (0.087) (0.092) (0.078) (0.079) ** *Export07 0.247 0.231 (0.125) (0.128)R&D 07 0.114 0.102 (0.149) (0.145)ODI07 0.255 0.255 (0.268) (0.260)Innovate07 0.131 (0.115)Prod_Innov07 0.063 (0.099)Proc_Innov07 0.033 (0.099)Org_Innov07 0.114 (0.095)Mark_Innov07 0.069 (0.090) ***BS107 0.175 0.088 (0.060) (0.092) * *** ***Constant -1.139 -1.101 -1.018 -0.806 -1.189 (0.638) (0.675) (0.685) (0.220) (0.245) * Robust standard errors in parentheses ** *** significant at 10%; significant at 5%; significant at 1% All models include industry dummiesTurning to the outcome equation (Table 17), the results with the lagged variables aresimilar to those with the contemporaneous variables, although once we include thepast recruitment difficulties and training variables from the 2006 survey (RecDif06 and 36
  • 47. Train06), the number of observations does fall somewhat and the significance of someof the contemporaneous variables drops (e.g. ISG and Vtech08). Once more, we findthat the positive impact of the relative wage variable remains when we use its lag.Firms that pay more than their peers are more likely to report SSVs. Dividinginnovation into the four different types does not uncover any differences in theirimplications for the likelihood of SSVs.Our measure of the correlation between the two equations, altrho, is statisticallysignificant in all but one of the specifications described in Table 16 and Table 17. 37
  • 48. Table 17 Results - SSVs - Lagged variables - Outcome equation (3) (4) (5) (6) (7) *** *** *** *** ***ln(Wi,07) – ln(Wijr,07) 0.205 0.212 0.199 0.295 0.361 (0.075) (0.075) (0.074) (0.059) (0.099)ln(E07) -0.018 -0.019 -0.028 -0.072 -0.083 (0.030) (0.030) (0.032) (0.050) (0.054)Census106 -0.011 -0.015 -0.014 -0.042 -0.057 (0.027) (0.029) (0.030) (0.051) (0.061) *** ***Census206 -0.029 -0.027 -0.032 -0.087 -0.153 (0.030) (0.031) (0.030) (0.028) (0.037) * * *ISG07 0.084 0.086 0.086 0.089 0.078 (0.049) (0.050) (0.050) (0.065) (0.097)Vman,07 0.087 0.081 0.083 -0.097 -0.134 (0.139) (0.134) (0.148) (0.297) (0.265) *** *** *** ** **Vprof,07 0.392 0.396 0.393 0.420 0.424 (0.115) (0.115) (0.118) (0.191) (0.197) *** *** *** *Vtech,07 0.403 0.392 0.408 0.365 0.220 (0.150) (0.148) (0.150) (0.194) (0.180) *** *** *** *** ***Vtrade,07 0.592 0.586 0.583 0.716 0.642 (0.104) (0.101) (0.108) (0.112) (0.179)Union07 0.210 0.240 (0.181) (0.182) ** **Monopoly07 0.159 0.164 (0.065) (0.065)Export07 -0.067 -0.071 (0.139) (0.141)R&D07 -0.164 -0.173 (0.144) (0.141) *** ***ODI07 -0.413 -0.406 (0.158) (0.155)Innovate07 0.002 (0.092)Prod_Innov07 0.028 (0.096)Proc_Innov07 0.080 (0.072)Org_Innov07 0.017 (0.100)Mark_Innov07 -0.102 (0.102)BS107 -0.061 (0.054) *Constant 0.205 0.200 0.194 0.283 0.390 (0.159) (0.161) (0.149) (0.247) (0.225) *** *** ***athrho -1.699 -1.543 -1.637 -1.774 -2.451*** (0.509) (0.397) (0.498) (1.249) (0.306)Prob  0.0008 0.0001 0.0010 0.1556 0.0000Observations 3,495 3,495 3,492 2,595 1,851 * Robust standard errors in parentheses ** *** significant at 10%; significant at 5%; significant at 1% All models include industry dummies 38
  • 49. 5.3. Skill and non-skill related hard-to-fill vacanciesIn Table 18 we present results for estimating a two-stage model with differentdefinitions of the dependent variable in the outcome equation. In columns (11) and(12) we model the probability of the firm posting a non-skill-related hard-to-fill-vacancy. In columns (13) and (14) we restrict the dependent variable to take apositive value when the firm has NSR only (i.e. no SSVs). The specifications aresimilar to (8) and (9) for SSVs in the previous section to allow comparability. Thedifferences between the them are that the first two columns measure productivityusing MFP and the final two columns use ln(LP).Perhaps the most obvious difference between the SSV specifications in section 5.1and (more importantly) 5.2 and the NSR specifications set out in Table 18 is theresult for the relative wage term (ln(Wi,07) – ln(Wijr,07)). In the results for our model ofSSVs, we find a statistically significant positive relationship between the wages thefirm pays relative to others in its industry and region, and the likelihood of it reportinga SSV. This is not the case when we consider NSRs.If we consider at all firms that have vacancies that are hard-to-fill for non-skill-relatedreasons (columns (11) and (12)), we find no evidence of a relationship betweenrelative wages and the probability of reporting an NSR. However, this may conflatethe influence of SSVs and NSRs, as many of the firms with NSRs will also haveSSVs. Therefore, if we focus on those firms that only have NSRs ((13) and (14)) adifferent picture emerges. Firms that pay higher relative wages are less likely tohave NSRs. When there are no constraints caused by a lack of skilled labour,markets appear to clear in the way predicted by standard economic theory. If a firmcannot find staff, it raises its wages relative to its competitors. Those that do notcompete in this way are more likely to suffer from vacancies that are hard-to fill. 39
  • 50. Table 18 Results – Different definitions of dependent variable (11) (12) (13) (14) NSR NSR NSR NSR Only OnlyOutcome equationln(Wi,07) – ln(Wijr,07) 0.019 -0.264*** -0.039 -0.306*** (0.063) (0.079) (0.059) (0.085)ln(E)07 -0.025 -0.073 0.011 -0.009 (0.040) (0.051) (0.040) (0.059)Census106 -0.035 0.033 0.001 0.038 (0.056) (0.032) (0.051) (0.050)Census206 -0.049 0.063 -0.026 0.040 (0.058) (0.045) (0.042) (0.064) *** *ISG 0.027 -0.186 0.077 -0.064 (0.058) (0.059) (0.044) (0.063) ***Vman -0.395 -0.410 -0.397 -0.508 (0.301) (0.407) (0.151) (0.351)Vprof 0.442** 0.102 0.353** 0.168 (0.195) (0.250) (0.164) (0.226)Vtech 0.307 -0.523*** 0.226 -0.574*** (0.222) (0.202) (0.167) (0.215)Vtrade 0.352*** -0.282** 0.238** -0.418*** (0.096) (0.127) (0.114) (0.124) *Constant 0.304 -0.292 0.243 -0.646** (0.159) (0.235) (0.150) (0.287)Selection equationln(sales)07 -0.077 0.025 -0.006 0.066 (0.076) (0.102) (0.063) (0.088)[ln(Wi)–ln(Wijr)]07 -0.136 0.208 -0.099 0.095 (0.285) (0.257) (0.258) (0.316)seprate07 0.056 0.205 0.223 0.361 (0.212) (0.287) (0.241) (0.295)ln(E) 07 0.358*** 0.414*** 0.370*** 0.419*** (0.040) (0.050) (0.048) (0.054)MFP07 0.069 0.075 (0.085) (0.085)ln(LP)07 0.027 0.062 (0.047) (0.047) *** *** ***Train08 0.563 0.583 0.623 0.636*** (0.125) (0.111) (0.097) (0.105) *** *** ***RecDiff07 0.621 0.495 0.652 0.514*** (0.065) (0.077) (0.068) (0.088)bss107 0.138 0.151* 0.171** 0.191*** (0.095) (0.081) (0.071) (0.068) *** *** **Constant -0.821 -0.911 -1.181 -1.691*** (0.145) (0.196) (0.482) (0.491) * *** ***athrho -1.619 -1.310 -1.453 -0.718** (0.963) (0.409) (0.352) (0.334)Observations 2,592 3,495 Robust standard errors in parentheses * ** *** significant at 10%; significant at 5%; significant at 1% Weighted and Stratified 40
  • 51. 6. Summary and ConclusionsSkills are an important determinant of the economic performance of people, firms,industries and economies. Shortages of skilled labour directly constrain productionand prevent firms from meeting demand and using available inputs efficiently. Theyinhibit innovation and the use of new technologies. This may have longer-termimpacts on the way firms do business, in terms of their location, size, structure,production methods and product strategy.The success of policies to enable the emergence and performance of successfulbusinesses depends upon having a workforce with the appropriate skills. If NewZealand is to raise productivity and improve its international competitiveness, it isimportant to understand whether skill shortages do indeed exist, how they manifestthemselves and develop policies to address them.In this paper we have examined factors relating to firms’ skill shortages. Our focushas been on vacancies that were hard-to-fill because the applicants lacked thenecessary skills, qualification or experience – what we have defined as ‘skill shortagevacancies’ (SSVs). We have also contrasted these with vacancies that were hard-to-fill for reasons other than the skills of applicants.Worker turnover is a basic fact of economic life. In any given year one in eightworkers leaves their employer. Searching for new staff is, therefore, a position mostbusinesses find themselves in, regardless of their situation. Almost half ofbusinesses find some of these vacancies hard-to-fill and more than a thirdexperience difficulties in finding workers with the required skills, qualifications orexperience they need.We have examined patterns of skill shortage vacancies using an econometrictechnique that allows us to account, in part, for the interrelationship between thelikelihood of a firm posting a vacancy and for those vacancies being hard-to-fill forskill-related reasons.Large firms are more likely to have vacancies – they have more workers and so thechance of at least one of them leaving is consequently higher. However, larger firmsdo not appear to find it more difficult to source workers with the required skills thansmaller firms, once we account for other characteristics. As we might expect, whenfirms’ output grows, the need to seek additional staff also grows. However, this is 41
  • 52. only in the years the firm is expanding. Expanding sales is not a good predictor offuture additional staff requirements.Businesses can experience skill shortages internally or externally. A shortage in theskills it requires can manifest itself: (a) in terms of the ability of its existing staff to dotheir job; or (b) in terms of its ability to find appropriately skilled workers throughrecruitment. We have found evidence that these two phenomena co-exist.Businesses experiencing internal skill gaps are also more-likely to have skill shortagevacancies. We also find evidence of persistence in this relationship; businessesexperiencing internal skill gaps are also more likely to have SSVs in the followingyear.Businesses have a ‘make or buy’ option when it comes to skills. If they cannotsource the skills internally or externally, they can up-skill existing or new workersthrough training. Firms’ training choices are the subject of a companion paper to this(Mason et al., 2012). In this paper, we find that firms who undertake training aremore likely to have vacancies, suggesting that their response to skill shortages islikely to be a mixture of recruitment and up-skilling strategies.Our results confirm earlier work that highlights the importance of two distinctions:First, it is important to distinguish between firms’ general recruitment activity and ithaving skill shortage vacancies. Second, we must distinguish between firmsexperiencing recruitment difficulties related to skills and those for other reasons (suchas the firm not offering good enough pay and conditions). Failing to account for thismay cause us to misdiagnose the problem and will therefore undermine theappropriateness of any policy prescription.In any market there will be a number of potential customers that cannot purchase allthey would wish because they are unwilling or unable to pay the market price. Thisdoes not mean that the market is malfunctioning. There is an important question forresearchers and policymakers to ask before interpreting reports of unfilled vacanciesof any kind, and of SSVs in particular, as a ‘skills problem’. This question is: Why dobusinesses that cannot fill a vacancy not simply raise the wages they are offering?There a number of reason why firms may not be able to do so. First, it may bebecause the firm cannot afford to pay any more. If this is the case, the problem is notthe labour market or the education system, but the firm’s own productivity. Firms that 42
  • 53. that undertake activities associated with high productivity are likely to both have ahigher demand for skills and be more able to afford to pay skill premia.Second, the supply of labour takes a long time to adjust. Because of the length oftime required to educate and train skilled workers, it takes a long time for changes inwages to influence changes in skill acquisition (particularly if we are relying on therather indirect mechanism of changes in relative wages influencing the choices ofstudents at school) and immigration is restrained by, among other things, legalbarriers and the costs of relocation.We have found that the businesses that feel the applicants they are attracting do nothave the required skills, qualifications or experience are those that are paying higherwages. The fact that businesses that pay higher wages than others in the sameindustry and region are actually more likely to experience SSVs suggests that raisingwages is not sufficient to fill these vacancies. In contrast, when we focus onvacancies that are hard-to-fill for reasons other than skill, paying higher wages doesappear to be a successful policy to fill vacancies. Firms that pay higher relativewages are less likely to have non-skill-related hard-to-fill vacancies. The implicationof this is that SSVs and NSRs are different phenomena.One important issue researchers have to confront when undertaking analysis of thistype is that of causality. It is one thing to establish a statistical regularity – acorrelation between firm characteristics, activities or environment and an outcomesuch as external skill shortages. It is another thing to interpret this as a causalrelationship. In this paper, we have uncovered some clear statistical regularities.These are consistent we with certain predictions we have discussed in this paper andprovide us with useful information to aid our understanding of how the landscape ofskills in New Zealand and, in particular, we done this from the perspective whatbusinesses are looking for, rather than what the education and training system hasprovided.The issues this research has raised and other questions unanswered suggest anumber of potentially useful avenues for future research. Perhaps the most obviousis the question of the impact of shortages of skilled labour. The LongitudinalBusiness Database has ample data to consider this question. We can use the panelelement of the Business Operations Survey to investigate the impact on future 43
  • 54. business practices. The following year’s BOS contains a ‘Business Practices’Module, which would be particularly useful for this. There is now enough informationto examine the impact of skills shortages on measures of firm performance, likeproductivity, profitability, employment or skill-intensive activities such exporting, R&Dand other innovation activity and outputs. It is possible, for example, to usetechniques previously used with the LBD to tackle issues of causality whenevaluating government business assistance programmes (Morris and Stevens, 200),the impact of broadband (Grimes, Ren and Stevens, 2012) or foreign acquisition(Fabling and Sanderson, 2012) on firm outcomes.This paper has focussed on external skill gaps – difficulties sourcing workers with therequired skills from the labour market. A fuller picture will only emerge when wehave also investigated the firms that suffer internal skill gaps – the fact that the firms;current workforce may not have all the skills it requires – and their impact of firmperformance.We have only lightly touched the issue of the persistence of skill shortages and foundprima facie evidence of persistence in recruitment difficulties. The increasing lengthof the BOS panel will enable researchers to use the Module A questions onrecruitment difficulties to investigate this. This is not, of as we have noted, the sameas a skill shortage, however. One means to overcome this and some of the issues ofcausality is to repeat the Business Skills and Strategy or something like it. Analternative option to a whole new survey module would be to append some of thequestions to another module, such as a repeat of the Business Practices Module. 44
  • 55. ReferencesAbowd, J.M., Haltiwanger, J., Lane, J., McKinney, K.L., and Sandusky, K., (2007), ‘Technology and the Demand for Skill: An Analysis of Within and Between Firm Differences’, Working Paper No 13043, National Bureau of Economic Research.Abowd, J. M., Kramarz, F. and Margolis, D. N. (1999), ‘High wage workers and high wage firms’, Econometrica, 67(2), pp. 251-333.Acemoglu, D. and Autor, D., (2011), ‘Skills, Tasks and Technologies: Implications for Employment and Earnings’, Chapter 12, pp. 1043-1171 in Ashenfelter, O. and Card, D. (eds.), Handbook of Labor Economics, vol. 4B, North Holland: ElsevierAcemoglu, D., and Pischke, S., (1999), ‘The Structure of Wages and Investment in General Training’, Journal of Political Economy, 107, pp. 539–572.Audretsch, D.B., and Feldman, M.P., (2004), ‘Knowledge Spillovers and the Geography of Innovation’, In J.V. Henderson and J-F Thisse (eds.), Handbook of Regional and Urban Economics, Volume 4, pp. 2713–2739.Arrow, K.J., (1962), ‘The Economic Implications of Learning by Doing’, Review of Economic Studies, 29.3, pp. 155–73Arrow, K.J., and Capron, W.M., (1959), ‘Dynamic Shortages and Price Rises: The Engineer-Scientist Case’, Quarterly Journal of Economics, 73(2), pp. 292-308.Ashton, D., Green, F., Sung, J. and James, D. (2002), ‘The evolution of education and training strategies in Singapore, Taiwan and S. Korea: a development model of skill formation,’ Journal of Education and Work, 15.1, pp. 5-30.Autor, D. H., Katz, L. F. and Krueger, A. B,. (1998), ‘Computing Inequality: Have Computers Changed the Labor Market’, Quarterly Journal of Economics, 113, 4, pp. 1169-1213.Barro, R.J., and Sala-i-Martín, X., (1995), Economic Growth, New York: McGraw Hill.Battu, H., Belfield, C.R., Sloane, P.J., (2004), ‘Human capital spillovers in the workplace: evidence for the service sector in Britain’, International Journal of Manpower, 25.1, pp.123-138 45
  • 56. Baumol, W., (1993), Entrepreneurship, management, and the structure of payoffs, Cambridge, MA: MIT Press.Becker, G.S., (1962), ‘Investment in Human Capital: A Theoretical Analysis’, Journal of Political Economy, 70.5, pp. 9-49.Becker, G.S., (1994), Human Capital: A theoretical and empirical analysis with special reference to Education. The University of Chicago Press.Bennett, J., and McGuinness, S., (2009), ‘Assessing the impact of skill shortages on the productivity performance of high-tech firms in Northern Ireland’, Applied Economics, vol. 41(6), pages 727-737.Bosworth, D., (1993), ‘Skill Shortages in Britain’, Scottish Journal of Political Economy, 40.3, pp. 241-271Bloom, N., and Van Reenen, J., (2007), ‘Measuring and Explaining Management Practices Across Firms and Countries’, Quarterly Journal of Economics, Vol. 122, No. 4, pp. 1351–1408.Bloom, N., and Van Reenen, J., (2010), ‘Why Do Management Practices Differ across Firms and Countries?’, Journal of Economic Perspectives, Volume 24, Number 1, pp. 203–224.Bloom, N., and Van Reenen, J., (2011), ‘Human Resource Management and Productivity’, in Orley Ashenfelter and David Card (eds.), Handbook of Labor Economics, Volume 4, Part B, pp. 1697–1767, North Holland: Elsevier.Card, D., (1999), ‘The causal effect of education on earnings’, in: O. Ashenfelter and D. Card (ed.), Handbook of Labor Economics, volume 3, chapter 30, pp. 1801-1863, North Holland: Elsevier.Card, D. and Lemieux, T. (2001), ‘Can Falling Supply Explain The Rising Return To College For Younger Men? A Cohort-Based Analysis’ The Quarterly Journal of Economics, 116(2), pp. 705-746Davis, S.J., Faberman, J., and Haltiwanger, J.C., (2006), ‘The Flow Approach to Labor Markets: New Data Sources and Micro–Macro Links’, Journal of Economic Perspectives, 20.3, pp. 3–26. 46
  • 57. Davis, S.J., Faberman, J., and Haltiwanger, J.C., (2010), ‘The Establishment-Level Behavior of Vacancies and Hiring’, NBER Working Papers 16265.Davis, S.J., Haltiwanger, J.C., and Schuh, S., (1996), Job Creation and Destruction, MIT press.Dickson, M., and Harmon, C., (2011), ‘Special Issue: Economic Returns to Education’, Economics of Education Review, 30.6, pp. 1115-1540.Dohmena, T.J., and Pfann, G.A., (2004), ‘Worker separations in a nonstationary corporate environment’, European Economic Review, 48, pp. 645-663Duffy, J., Papageorgiou, C., and Perez-Sebastian, F., (2004), ‘Capital-Skill Complementarity? Evidence from a Panel of Countries’, The Review of Economics and Statistics, 86.1, pp. 327-344Durbin, Sid, (2004), ‘Workplace Skills, Technology Adoption and Firm Productivity: A Review’, New Zealand Treasury Working Paper 04/16.Eisenhardt, K., and Martin, J., (2000), ‘Dynamic capabilities: what are they?’, Strategic Management Journal, 21, pp.1105-1121.Fabling (2007), ‘Just how innovative are New Zealand firms? Quantifying & relating organisational & marketing innovation to traditional science and technology indicators’, MED Occasional Paper 07/04.Fabling, R., (2009), ‘A rough guide to New Zealands Longitudinal Business Database’, Institute of Economic Research, Hitotsubashi University, Global COE Hi-Stat Discussion Papers, No. 103.Fabling, R., and Grimes, A., (2007), ‘HR Practices and Firm Performance: What matters and who does it?’, MED Occasional Paper 07/02.Fabling, R., Grimes, A., Sanderson, L., and Stevens, P., (2008), ‘Some rise by sin, and some by virtue fall: Firm dynamics, Market structure and Performance’, MED Occasional paper 08/01.Fabling, R., Grimes, A., and Stevens, P., (2008), ‘A Comparison of Qualitative and Quantitative Firm Performance Measures’, MED Occasional Paper 08/04. 47
  • 58. Fabling, R., Grimes A., and Stevens, P., (2012), ‘The Relatives are Fine: Use of Qualitative Firm Data in Economic Analysis’, Applied Economics Letters, 19:7, pp. 614-8.Fabling, R., and Sanderson, L., (2012), ‘Foreign acquisition and the performance of New Zealand firms’, Reserve Bank of New Zealand, DP2011/08.Forth, J., and Mason, G., (2004), ‘Information and Communication Technology (ICT) Skill Gaps and Company-Level Performance: Evidence from the ICT Professionals Survey 2000-01’, National Institute of Economic and Social Research, London. NIESR Discussion Paper No. 236. Downloadable at: http://www.niesr.ac.uk/pubs/dps/dp236.PDFFreeman, R.B., (1971), The Market for College Trained Manpower. Cambridge, MA: Harvard University Press.Freeman, R.B., (1976), ‘A Cobweb Model of the Supply and Starting Salary of New Engineers’, Industrial Labor Relations Review, 29(2), pp. 236-48.Glaeser, E., Kallal, H., Scheinkman, J., and Shleifer, A., (1992), ‘Growth of cities’, Journal of Political Economy, 100, pp. 1126-1152.Goldin, C., and Katz, L.F., ‘The Origins of Technology-Skill Complementarity’, Quarterly Journal of Economics, 113, pp. 693–732.Green, F.; Machin, S. and Wilkinson, D. (1998), ‘The meaning and determinants of skills shortages’, Oxford Bulletin of Economics and Statistics, 60(2).Griliches, Z., (1969), ‘Capital-Skill Complementarity’, The Review of Economics and Statistics, 51.4, pp. 465-468Grimes, A., Ren, C., and Stevens, P., (2012), ‘The Need for Speed: Impacts of Internet Connectivity on Firm Productivity’, Journal of Productivity Analysis, 37.2, pp. 187-201Guttman, L., (1954), ‘Some Necessary Conditions for Common Factor Analysis’, Psychometrika, 19.2, pp. 149-61.Hall, B.H. and Kahn, B., (2003), ‘Adoption of New Technology’, Working Paper No. 9730, National Bureau of Economic Research. 48
  • 59. Haskel, J., Hawkes, D., and Pereira, S. (2005), ‘Skills, human capital and the plant productivity gap: UK evidence from matched plant, worker and workforce data’, CEPR Discussion Paper 5334.Haskel, J. and Martin, C. (1993a), ‘Do Skill Shortages Reduce Productivity: Theory and Evidence from the United Kingdom’, Economic Journal, 103, pp. 386-394Haskel, J. and Martin, C. (1993b), ‘The causes of skill shortages in Britain’, Oxford Economic Papers, 45(4), pp. 573-588.Haskel, J. and Martin, C., (2001), ‘Technology, wages and skill shortages: evidence from UK micro data’, Oxford Economic Papers, 53, pp. 642-658.Heckman, J., (1979), ‘Sample selection bias as a specification error’, Econometrica, 47, pp. 153-161.Holzer, H.J., (1994), ‘Job Vacancy Rates in the Firm: An Empirical Analysis’, Economica, 61.241, pp. 17-36.Hull L., and Arnold, R, (2008), New Zealand firm growth as change in turnover, mimeo, Ministry of Economic Development.Kaiser, H.F., (1970), ‘A Second Generation Little Jiffy’ Psychometrika, 35, pp. 401-17.Keep, E., Mayhew, K., and Corney, M., (2002), ‘Review of the evidence on the rate of return to employers of investment in training and employer training measures’, Coventry, ESRC Centre on Skills, Knowledge and Organisational Performance, Research Paper No 34. Downloadable from: www.econ.ox.ac.uk/SKOPE/Working%20Papers/SKOPEWP34.pdfKneller, R., and Stevens, P., (2006), ‘Frontier Technology and Absorptive Capacity: Evidence from OECD Manufacturing Industries’, Oxford Bulletin of Economics and Statistics, vol. 68(1), pp. 1-21.Lane, P.J., and Lubatkin, M., (1998), ‘Relative absorptive capacity and inter- organizational learning’. Strategic Management Journal, 19(5), pp. 461-477.Machin S., and Van Reenen, J., (1998), ‘Technology and Changes in Skill Structure: Evidence from Seven OECD Countries’, Quarterly Journal of Economics, 113, 4, pp. 1215-44. 49
  • 60. Madsen, J.B., (2010), ‘The anatomy of growth in the OECD since 1870’, Journal of Monetary Economics, 57(6), 753-767.Maré, D.C., (2008), ‘Labour Productivity in Auckland Firms’, MED Occasional Paper Number 08/09. Downloadable at: http://www.med.govt.nz/templates/MultipageDocumentTOC____38579.aspxMaré, D.C., Fabling, R., and Stillman, S., (2010), ‘Immigration and Innovation’, draft Motu Working Paper, Motu Economic and Public Policy Research.Morris, M., and Stevens, P., (2010), ‘Evaluation of a New Zealand business support programme using firm performance micro-data’, Small Enterprise Research, 17.1, pp. 30-42.Mason, G., (2005), ‘In Search of High Value Added Production: How Important are Skills? Investigations in the Plastics Processing, Printing, Logistics and Insurance Industries in the UK’, Research Report RR663, Department for Education and Skills, London.Mason, G., Mok, P., Nuns, P., Stevens, P., and Timmins, J., (2012), ‘The Training’s A-Comin’ In: An analysis of training decisions using microdata’, MED Occasional Paper 12/06.Mason, G. and Wilson, R., (eds.), (2003), ‘Employers Skill Survey: New Analyses and Lessons Learned’, Report NALL1, Department for Employment and Skills, London. Downloadable at: http://www.dfes.gov.uk/research/data/uploadfiles/RR663.pdfMason, G., and Stevens, P., (2003), ‘The determinants of HTF and SSV in key occupational groups in Employers Skill Survey’, Chapter 7 of Mason and Wilson (eds.) (2003).McMahon, W.M., (2004), ‘The social and external benefits of education’, Chapter 6 in G. Johnes and J. Johnes, International handbook on the economics of education, Cheltenham: Edward Elgar.Morrison-Paul, C. J., and Siegel D., (2001), ‘The Impacts of Technology, Trade, and Outsourcing on Employment and Labor Composition’, Scandinavian Journal of Economics, 103, pp. 241-264. 50
  • 61. Mortensen, D.T., (2005), Wage Dispersion: Why Are Similar Workers Paid Differently? Cambridge, MA: MIT PressNickell, S., (1986), ‘Dynamic Models of Labour Demand’, in Ashenfelter, O., and R. Layard (eds.) Handbook of Labour Economics, Volume I. North Holland: Elsevier Science.van Ours, J., and Ridder, G., (1992), ‘Vacancies and the Recruitment of New Employees’, Journal of Labor Economics, 10(2), pp. 138-55.Petrongolo, B. and Pissarides, C.,(2001), ‘Looking into the black box: a survey of the matching function’, Journal of Economic Literature, 39(2), pp. 390-431.Pissarides, C., (2000), Equilibrium Unemployment Theory (2nd ed.), Cambridge, MA: MIT Press.Psacharopoulos, G., and Patrinos, H.A., (2004), ‘Human capital and rates of return’, Chapter 4 in G. Johnes and J. Johnes, International handbook on the economics of education, Cheltenham: Edward Elgar.Razzak, W.A. and Timmins, J.C., (2008), ‘A macroeconomic perspective on skill shortages and the skill premium in NZ’, Australian Economic Papers, 47(1), pp. 74-91.Romer, Paul M. (1990), ‘Endogenous Technological Change’, Journal of Political Economy, 98, pp. 71-102.Rosenberg, N., (1972), ‘Factors Affecting the Diffusion of Technology’, Explorations in Economic History, Vol. 10(1), pp. 3-33.SNZ (2009), ‘Business Operations Survey: 2008 Tables’, Statistics New Zealand, http://stats.govt.nz/methods_and_services/access-data/tables/business-op- survey-2008.aspxStevens, M., (1994), ‘A Theoretical Model of On-the-Job Training with Imperfect Competition’, Oxford Economic Papers, 46, pp. 537-562.Stevens, P.A., (2005), ‘The job satisfaction of English academics and their intentions to quit academe’, Discussion Paper 262, National Institute for Economic and Social Research, London. 51
  • 62. Stevens, P.A., (2007), ‘Skill shortages and firms employment behaviour’, Labour Economics, vol. 14(2), pages 231-249.Stevens, P., (2012), ‘Business Strategy and Skills in New Zealand’, MED Occasional Paper 12/06Stevens, P. and Weale, M., (2004), ‘Education and Economic Growth’, Chapter 4 in G. Johnes and J. Johnes, International handbook on the economics of education, Cheltenham: Edward Elgar.Schwalje, W., (2011) ‘What is the Difference Between a Skills Shortage and a Skills Gap?’, in NORRAG NEWS, Towards a New Global World of Skills Development? TVETs turn to Make its Mark, No.46, pp. 44-46.Tan, H., Savchenko, Y., Gimpelson, V., Kapelyushnikov, R., and Lukyanova, A., (2007), ‘Skills Shortages and Training in Russian Enterprises’, IZA Discussion Paper No. 275.Teece, D., Pisano, G., and A. Shuen (1997), ‘Dynamic capabilities and strategic management’, Strategic Management Journal, 18, pp.509-533.Teece, D.G., (2007), ‘Explicating dynamic capabilities: the nature and microfoundations of (sustainable) enterprise performance’, Strategic Management Journal, 28(13), pp. 1319-1350UTS (2009), ‘Management Matters in New Zealand - how does manufacturing measure up?’, Report for Ministry of Economic Development by University of Technology, Sydney. Downloadable at: http://www.med.govt.nz/templates/MultipageDocumentTOC____43278.aspx 52
  • 63. Appendix A1. Data AppendixA1.1 BOS Variables6.1.1. Module A (2005 – 2008)Data items for questions for Module A relate to the 2008 survey. Because ofchanges to questions, these may differ slightly in earlier years.Exporting (Export)This variable relates to question 11 of Module A: ‘Estimate from question 8 thepercentage of sales that came from exports’. Data Item A1101. The variable Exporttakes the value of 1 if the firm reports a non-zero percentage of sales as coming fromexports, zero otherwise.Research and Development (R&D)This variable relates to question 23 of Module A. Data item (A2300). The question is:‘For the last financial year, did this business undertake or fund any research anddevelopment (R&D) activities?’ The respondents are asked to include: ‘any activitycharacterised by originality: it should have investigation as its primary objective, andan outcome of gaining new knowledge, new or improved materials, products,services or process’; or ‘the buying abroad of technical knowledge or information’.They were asked to not include: ‘market research’; ‘efficiency studies’; or ‘stylechanges to existing products’. The variable R&D takes the value of 1 if the responseis yes, zero otherwise.Ownership of overseas businesses (ODI)This variable relates to question 25 of Module A. Data item (A2500). The questionasked is ‘As at the end of the last financial year, did this business hold any ownershipinterest or shareholding in an overseas located business (including its own branch,subsidiary or sales office)? The variable ODI takes the value of 1 if the response isyes, zero otherwise.Recruitment difficulties (RecDif)This variable relates to question 33 of Module A: ‘Over the last financial year, to whatextent did this business experience difficulty in recruiting new staff for any of the 53
  • 64. following occupational groups?: managers and professionals; technicians andassociate professionals; tradespersons and related workers (includingapprenticeships); all other occupations’ (Data Items A3301 to A3304). The variableRecDif takes the value of 1 if the respondent reports a severe difficulty in any of thethree occupations, zero otherwise. For more on this variable, including anexamination of its persistence and a comparison with the hard-to-fill vacancy variablefrom Module C of BOS 2008, see the discussion in section 3.3 and Appendix A2.Collective agreements (union)This variable relates to question 36 of Module A: ‘As at the end of the last financialyear, what percentage of this business’s employees were covered by a collectiveemployment agreement?’ (Data item A3600.) The variable union takes the value of 1if the respondent reports any value above zero and zero otherwise.Innovation (innovate)This variable relates to question 42 of Module A. Data item (A4200). The question is:‘In the last financial year, did this business develop or introduce any new orsignificantly improved: goods or services; operational processes;organisational/managerial processes; marketing methods?’ The variable innovatetakes the value of 1 if the response is yes, zero otherwise.Competition (monopoly)Competition is measured through binary variable monopoly. These variables relate toquestion 47 of Module A. Data item (A4700). The respondents were asked ‘Howwould you describe this business’s competition?’ The variable monopoly takes thevalue of 1 if the response is yes, zero otherwise.6.1.2. Module B – Innovation (2007)Product innovation (Prod_Innov07)This variable relates to question 3 of Module B: ‘During the last 2 financial years, didthis business introduce onto the market any new or significantly improved goods orservices?’ Data Item B0300. Respondents are asked to exclude the selling of newgoods or services wholly produced and developed by other businesses. The variableProd_Innov07 takes the value of 1 if the firm has innovated, zero otherwise. 54
  • 65. Operational process innovation (Proc_Innov07)This variable relates to question 7 of Module B: ‘During the last 2 financial years, didthis business implement any new or significantly improved operational processes (i.e.methods of producing or distributing goods or services)?’ Data Item B0700. Thevariable Proc_Innov07 takes the value of 1 if the firm has innovated, zero otherwise.Organisational/managerial process innovation (Org_Innov07)This variable relates to question 10 of Module B: ‘During the last 2 financial years, didthis business implement any new or significantly improved organisational/managerialprocesses (i.e. significant changes in this business’s strategies, structures orroutines)?’ Data Item B1000. The variable Org_Innov07 takes the value of 1 if the firmhas innovated, zero otherwise.Marketing innovation (Mark_Innov07)This variable relates to question 12 of Module B: ‘During the last 2 financial years, didthis business implement any new or significantly improved sales or marketingmethods which were intended: to increase the appeal of goods or services forspecific market segments; to gain entry to new markets’ Data Item B1200. Thevariable Mark_Innov07 takes the value of 1 if the firm has innovated, zero otherwise.6.1.3. Module C – Employment Practices (2006)Training (train06)This variable relates to question 13 of Module C. Data item (C1300). Respondentswere asked: ‘Over the last financial year, have employees in this business receivedtraining of any type?’ The variable train06 takes the value of 1 if the response is yes,zero otherwise.6.1.4. Module C – Business Strategy and SkillsMarket focus (mark_int)This variable relates to question 2 of Module C: ‘In the last 2 financial years, whatmarket accounted for the largest proportion of this business’s total sales of goods orservices?’ Data item (C0200). Respondents could answer one of either ‘local’,‘national’ or ‘international’. The variable mark_int takes the value of 1 if therespondent indicates that their market focus is ‘international’ and zero otherwise. 55
  • 66. Customisation (Customise)This variable relates to question 6 of Module C: ‘Which of the following bestdescribes this business’s goods or services?’ Data item C0600. Respondentsanswer from one of the following four options: ‘standard range of goods or services’,‘minor differences in goods or services according to customer requirements’,‘substantial differences in goods or services according to customer requirements’, or‘don’t know’. The variable Customise takes the value 1 if the respondent answers‘substantial differences in goods or services according to customer requirements’,zero otherwise.Price leadership (Price_Leader)This variable relates to question 8 of Module C: ‘How often is this business able toobtain a higher price than competitors for its main goods or services?’ Data itemC0800. The variable Price_Leader takes the value 1 if the respondent answers‘always’, zero otherwise.Occupational vacancy rates (Vman, Vprof, Vtech, Vtrade, Vproftechtra)These variables relate to the question 15 from Module C: ‘During the last financialyear, how many vacancies has this business had for the following roles? managers;professionals; technicians and associate professionals; tradespersons and relatedworkers; clerical, sales and services workers; labourers, production transport or otherworkers’. Data items C1501 to C1506. Each variable relates to the proportion ofvacancies that are made up of the respective occupations as a proportion of allvacancies. For example, that for Vman equals the following data items:C1501/( C1501+ C1502+ C1503+ C1504+ C1505+ C1506).Internal skill gaps (ISG)This variable relates to question 20 from Module C: ‘How many of this business’sexisting staff have the skills required to do their job?’ Data items C2000 to C2006.There are four response categories: ‘less than half’, ‘half or more’, ‘all staff’ or ‘nostaff of this type’. The variable ISG is a binary variable, taking the value of one if therespondent answers that ‘less than half’ for any of the occupations, zero otherwise. 56
  • 67. Training (train)This variable relates to question 22 of Module C. Data item (C2200). Respondentswere asked: ‘During the last financial year, have the staff of this business receivedtraining of any type?’ The variable train takes the value of 1 if the response is yes,zero otherwise.Occupational breakdown of staff (Propman, Propprof, Proptech, Proptrade)This variable relates to questions 10-13 of Module C. Note that every year in ModuleA, businesses are asked to provide a breakdown of their staff by four occupations(A3201-A3204). Respondents are asked to copy the totals from Module A into boxesin Module C. They are then asked to provide a further breakdown of two of these(‘Managers and professionals’ and ‘all other occupations’). We calculate theproportion of the workforce in each of the occupations, with ‘labourers, production,transport or other workers’ as the baseline category. 57
  • 68. Table 19 Staff occupation/role variables DataOccupation/role Variable itemManagers 30 C1002 Propman(i.e. those who supervise staff or determine policy and future direction)Professionals 31 C1003 Propprof(i.e. those who have specific expertise, but no managerial responsibility)Technicians and associate professionalsTechnicians and associate professionals perform complex technical or C1101 Proptechadministrative tasks, often in support of professionals or managers (e.g.technical officer, building inspector, legal executive)Tradespersons and related workersTradespersons and related workers perform tasks requiring trade specific C1201 Proptradetechnical knowledge. Include all apprentices and trade supervisors (e.g.electrician, mechanic, hairdresser, baker).Clerical, sales and service workers C1302 Baseline(i.e. those who perform administrative, sales or customer service tasks)Labourers, production, transport or other workers C1203 Baseline(i.e. those who operate vehicles or equipment or perform manual tasks)A1.2 LEED/PAYE DataOur data on employment come from the Linked Employer-Employee Database. Ithas two components, counts of employees and working proprietors.EmployeesEmployment is measured using an average of twelve monthly PAYE employeecounts in the year. These monthly employee counts are taken as at 15th of themonth. This figure excludes working proprietors and is known as Rolling MeanEmployment (RME).30 In Module A, where Mangers and professionals are grouped together, there are separatedescriptions of managers and professionals. In addition to the description of managers given in thequestion in Module C, respondents are also offered two examples: ‘General Manager’ and ‘FinanceManager’31 In Module A, respondents are also offered a different description: ‘Professionals perform analytical,conceptual or creative tasks with skills equivalent to a bachelor degree or higher (e.g. accountant,engineer, journalist, computer programmer)’. 58
  • 69. Working proprietorsThe working proprietor count is the number of self-employed persons who were paidtaxable income during the tax year (at any time). In LEED, a working proprietor isassumed to be a person who (i) operates his or her own economic enterprise orengages independently in a profession or trade, and (ii) receives income from self-employment from which tax is deducted.From tax data, there are five ways that people can earn self-employment incomefrom a firm:  As a sole trader working for themselves (using the IR3 individual income tax form [this is used for individuals who earn income that is not taxed at source]);  Paid withholding payments either by a firm they own, or as an independent contractor (identified through the IR348 employer monthly schedule);  Paid a PAYE tax-deducted salary by a firm they own (IR348);  Paid a partnership income by a partnership they own (IR20 annual partnership tax form [this reports the distribution of income earned by partnerships to their partners] or the IR7 partnership income tax return);  Paid a shareholder salary by a company they own (IR4S annual company tax return [this reports the distribution of income from companies to shareholders for work performed (known as shareholder-salaries)]).Note that it is impossible to determine whether the self-employment income involveslabour input. For example, shareholder salaries can be paid to owner-shareholderswho were not actively involved in running the business. Thus there is no way oftelling what labour input was supplied, although the income figures do provide somerelevant information (a very small payment is unlikely to reflect a full-year, full-timelabour input).SeparationsLabour separation rate (SepRate) is measured as the annualised number ofseparations to the firm divided by RME. Note that we also considered measures ofnet employment growth (accessions less separations) and labour turbulence 59
  • 70. (accessions plus separations), but neither of these were statistically significant in ourspecificaitons.WagesWages are calculated as ‘total employee gross earnings’ from the LEED database,divided by RME (i.e. excluding working proprietors). This data comes from theEmployers Monthly Schedule (EMS).Relative wagesOur measure of relative wages measures the wages of the firm relative to other firmsin the same industry in the same Territorial Authority. Because many of the firms aremulti-plant firms we account for this by calculating employment weighted values.The entity in the LBD that corresponds to the concept of a ‘firm’ or ‘businesses’ ineconomic parlance is called the enterprise. Just as a firm can be made up of manyplants or establishments, we say an enterprise can have many PBNs32. In the LBDwe have information on the location, industry, [wages] and employment at theestablishment (PBN) level.First we calculate the average wage a described above for each plant in each month.These can be averaged across the establishments within an enterprise and acrossmonths in the year using employment weights. This is mathematically equivalent tothe average wage calculated at the level of the enterprise. We can then calculate theaverage wage of an establishment relative to other firms in the same 2-digit industryand the same Territorial Authority. Note that if there were fewer than 30 firms in thesame 2-digit industry and the same Territorial Authority, we calculated this at theone-digit level.A1.3 Other LBD Data (AES, BAI, and IR10)Sales (ln(sales))Our measure of sales from the Business Activity Indicator (BAI) database. TheBusiness Activity Indicator uses GST data from the Inland Revenue matched to theStatistics NZ Business Frame. The BAI data come from the Goods and Services Tax32 In fact, the geographic unit that makes up an enterprise is called the geographic using (or GEO).However, after work conducted to improve the longitudinal linking of GEOs created a new identifier,the Permanent Business Number, hence the name. 60
  • 71. return form, GST 101. For more information on this and other LBD datasets, seeFabling, Grimes, Sanderson and Stevens (2008) and Fabling (2009).Labour Productivity (ln(LP))Labour productivity is calculated from the AES, BAI, IR10 and LEED data. Theprimary data source for gross output, intermediate consumption, changes instocksand capital services is the AES. These are supplemented with data from the IR10financial accounts form when missing. Finally, for the purposes of calculating labourproductivity, if the data is still missing, we calculate value added as sales minuspurchases from the BAI. Labour input is total labour input (RME plus the count ofworking proprietors).Multifactor productivity (MFP)This is generated from a simple 2-digit industry-level OLS regression of the log ofgross output on the logs of intermediate consumption, capital services and labourinput. In earlier work, we also considered a measure based on Levinsohn and Petrin(2003), but this had little effect on results. 61
  • 72. Appendix A2. Comparison of hard-to-fill vacancies in Module C with recruitment difficulties in Module AIn this appendix, we compare responses to the hard-to-fill vacancies question inModule C with those to a similar question that is asked in Module A. Module A is acore module, asked every year, containing questions about businesses’ generalactivities, environment and performance. In a section on employment, firms areasked ‘Over the last financial year, to what extent did this business experiencedifficulty in recruiting new staff for any of the following occupational groups?’ (A33).Respondents are asked to rank these difficulties on a three-point scale (‘no difficulty’,‘moderate difficulty’ or ‘severe difficulty’) along with a ‘don’t know’ or ‘not applicable’category. Whilst these two questions (C18 and A33) appear to be measuring thesame concept, we must be wary of expecting responses to be identical. In Module Athis is a generic stand-alone question (although the previous question does ask themto break down staff numbers by occupation). In Module C, respondents have justbeen asked a number of questions about vacancies during the last financial year andbeen asked to provide reasons why their business found it hard-to-fill vacancies.This, combined with the different language used in module C (vacancy plus hard-to-fill vacancy) and Module A (‘difficulty in recruitment’), are likely to create a certainamount of dissonance. For more on the subject of how respondents interpret andanswer questions in business surveys, with particular reference to the BOS, seeFabling, Grimes and Stevens (2008, 2012).With these caveats I mind, Table 20 sets out a comparison of responses to questionsC18 and A33 for 2008. Because Module C has a more disaggregated set ofoccupational classifications we present cross-tabulations with both the morenarrowly-defined occupational groups and for both combined.The first panel of Table 20 compares responses to the first occupational group inA33, ‘managers and professionals’ (A3301), with those for ‘managers’ (C1801),‘professionals’ (C1802) and both C1801 and C1802 combined. The final columncounts there being a hard-to-fill vacancy if the respondent ticks the box for eitherC1801 or C1802.Perhaps the first thing that is clear is that there is a high degree of correlationbetween the two measures, but that this is not total. Just under three-quarters of 62
  • 73. respondents that said they had a severe difficulty in recruiting managers andprofessionals in Module A said that vacancies for either managers or professionalswere hard-to-fill in Module C. Nevertheless, over a quarter did not.Table 20 Weighted counts of firms reporting levels of difficulty recruiting new staff in Module A by Module C hard-to-fill vacancy response Module C Managers and Professionals Managers & Managers Professionals professionalsModule A No H2F H2F No H2F H2F No H2F H2FNo Difficulty 4,767 279 4,782 270 4,578 471Moderate difficulty 3,900 1,140 3,846 1,191 3,030 2,010Severe difficulty 1,908 1,242 1,488 1,665 873 2,280Dont know 1,206 57 1,158 105 1,119 144Not applicable 19,215 351 19,077 489 18,792 774Total 30,999 3,072 30,348 3,720 28,389 5,679Pearson test2 294.48 502.97 702.50F 31.77 54.28 75.15p 0.0000 0.0000 0.0000 The three tables are cross-tabulations of Q18 from Module C (columns) and Q33 from Module A (rows) In this particular table, the first pair of columns relates whether firms ticked box C0801 (manger roles were hard-to-fill (H2F)), the second pair to C0802 (professionals). The final pair of columns relate to firms who ticked neither (No H2F) or one of the two boxes (H2F). The rows represent responses to A3301. Counts based on population weights and random-rounded to base 3 The test of independence that is displayed by default is based on the usual Pearson  statistic for two-way 2 tables. To account for the survey design, the statistic is turned into an F statistic with non-integer degrees of freedom by using a second-order Rao and Scott (1981, 1984) correction. Note that this test is performed on the first three rows only (i.e. those that report whether they have a difficulty) 63
  • 74. Table 20 (continued) Module C Technicians and Tradespersons associate and related professionals workersModule A No H2F H2F No H2F H2FNo Difficulty 3,645 84 4,278 219Moderate difficulty 3,267 963 4,146 2,262Severe difficulty 1,509 1,164 1,329 2,970Dont know 1,116 24 1,086 30Not applicable 20,991 378 16,533 366Total 30,528 2,619 27,372 5,844Pearson test2 340.62 631.65F 58.18 92.67p 0.0000 0.0000 See notes for table 1a The pairs of columns relate to data items C1803 and C1804 The rows relate to the responses to data items A3302 and A3303 Counts based on population weights and random-rounded to base 3Table 20 (continued) Module C All other occupations Clerical, sales and Labourers, production, Both service workers transport or other workersModule A No H2F H2F No H2F H2F No H2F H2FNo Difficulty 4,575 528 4,512 588 4,071 1,032Moderate difficulty 7,575 2,400 7,449 2,526 5,481 4,491Severe difficulty 1,392 753 1,080 1,071 477 1,671Dont know 1,212 72 1,080 204 1,011 273Not applicable 9,501 426 9,381 546 8,991 936Total 24,255 4,179 23,502 4,935 20,037 8,400Pearson test2 116.07 216.88 384.94F 12.53 27.96 51.68p 0.0000 0.0000 0.0000 See notes for table 1a The pairs of columns relate to data items C1805 and C1806 and the combination of the two, as set out in the footnote to table 1a The rows relate to the responses to data items A3304 Counts based on population weights and random-rounded to base 3 64
  • 75. Of the two occupational groups with direct mappings from A33 to C18, the correlationis less strong for ‘Technicians and associate professionals’ (A3302 and C1803), butsimilar for ‘Tradespersons and related workers’ (A3303 and C1804)33.The final panel of Table 20 shows the results for another category that is split intotwo groups in Module C. The ‘all other occupations’ (A3304) category in Module Amaps to two categories in Module C: ‘clerical, sales and service workers’ (C1805)and ‘labourers, production, transport or other workers’ (C1806). The results for thefinal combined group are similar to the ‘managers and professional’ combined group.Whilst the correlation is not very strong between the combined A3304 and thecomponent parts C1805 and C1806 individually, that with the combined group ismuch higher. More than three quarters of the respondents that stated they hadsevere difficulty recruiting ‘all other occupations’ in Module A answered that theyfound vacancies for either ‘clerical, sales and service workers’ or ‘labourers,production, transport or other workers’.33 Note that for the question in Module A (A33), the occupational title includes the parenthesis‘(including apprenticeships)’. 65

×