How To Know If The Internet Is A Labor Market Matchmaker

Transcription

1 THE INTERNET AS A LABOR MARKET MATCHMAKER HOW EFFECTIVE ARE THE NEW ELECTRONIC METHODS OF WORKER RECRUITMENT AND JOB SEARCH? Author: Peter Kuhn Keywords: job search, recruitment, Internet, matching, job boards Extended, Preliminary Version. Contains additional references to the literature ELEVATOR PITCH Since the Internet s earliest days, workers and firms have used a variety of online methods to form new matches. Until recently, however, there has been little evidence that any of these Internet-based tools have reduced labor market frictions enough to have measurable effects on unemployment durations. Recent studies as well as the developing use of social networking as a business tool however suggest that workers and firms are finally developing ways to use the Internet as an effective matchmaking tool. KEY FINDINGS Positive Key Findings (Pro) Negative Key Findings (Contra) + Looking for work and posting jobs online have very low costs compared to traditional job search and recruiting methods + Job applications submitted electronically can be searched and filtered at low cost + Workers who look for jobs online are not casual window shoppers : they search just as actively as workers who search off-line. + Recent research has finally detected an unemployment-duration-reducing effect of Internet job search + The advent of social networking as a job search tool promises to yield further improvements in labor market matching efficiency. - Most previous studies have found no frictionreducing effects of looking for work online. -The low cost of applying for jobs electronically can result in large numbers of inappropriate applications being submitted. -Internet job boards can be dominated by large numbers of stale resumes - Internet job search remains disproportionately anonymous and formal, i.e. it does not take advantage of pre-existing networks of friends, relatives or colleagues, which historically have been the most effective search and recruitment tools. + Electronic job boards are emerging as an important source of data on labor market trends

2 AUTHOR S MAIN MESSAGE 2 A number of online tools, including job boards and social networking, have the potential to significantly reduce frictions in labor markets. Perhaps surprisingly, much of this potential is only now being realized, as early evidence showed no friction-reducing effect. Job board data are also emerging as an important tool for the statistical description of labor markets, generating useful data for firms, workers and policymakers. MOTIVATION Over the past two decades, a large share of commerce in developed countries has moved from brick-and-mortar settings to the web. In the process, a number of markets (for example books and recorded music) have been fundamentally transformed. Entirely new market mechanisms, such as ebay, have been created. And in a number of cases, there is solid evidence that matching buyers and sellers on line has had beneficial effects on market functioning. Given all these developments, plus the fact that a large share of workers job search and firms recruitment activities has shifted online, it seems possible that on-line matching has improved the functioning of labor markets as well. In this article I review the evidence on whether this has actually occurred. TRENDS IN INTERNET JOB SEARCH Table 1 shows the percentage of jobseekers looking for work on line in the United States in two periods: and ; the numbers are taken from a recent study by Kuhn and Mansour (2012) that focused on persons aged in both periods. Not surprisingly, there was a massive increase in on-line job search over this ten-year period: the share of unemployed persons who said they used the Internet to look for jobs more than tripled, from 24 and 74 percent. Much of this increase was related to the diffusion of Internet access the share of unemployed persons who had home Internet access more than doubled--, but the increase was also driven by an increase in the use of the Internet for job search among persons who had home access. This suggests that the on-line tools available for job searching may have become more attractive to workers. Another contributing factor, of course, is the fact that the vast majority of job ads has migrated to the Internet: that s where the jobs are now. While unemployed persons, by definition, are looking for work, it is important to remember that the unemployed are not the only ones looking for jobs in any labor market. In fact, in most labor markets, a substantial share of jobseekers already has a job. While historical data is not available for this group, Table 1 shows that by 2008/09, 85 percent of employed jobseekers used the Internet as part of their job search. In short, over the past decade, jobseekers have dramatically expanded their use of the Internet as a means of finding new jobs, to the point where Internet job search is almost universal, at least among jobseekers who already have jobs.

3 3 Table 1: Trends in Internet Access and Job Search among young workers, United States Period and Sample: Home Internet Access Looking for work on line, given Access Looking for work on line 1998/2000: Unemployed /2009: Unemployed Employed Jobseekers Source: Kuhn and Mansour, WHO LOOKS FOR WORK ON LINE? Perhaps not surprisingly, both in 1998/2000 and 2008/09, Internet job searchers are, on average, younger and much better educated than jobseekers who don t look on line (Kuhn and Skuterud 2004; Kuhn and Mansour 2012). It is perhaps noteworthy, though, that by 2008/09, when Internet access was much more universal, this large education gap was evident even among jobseekers with home Internet access. Thus, even though they had the option to look for work on line from a home computer, many unemployed, less-educated jobseekers did not use the Internet as part of their job search activity. At neither point in time was there a gender gap in online search. Internet job search activity is also overrepresented in certain industries, such as IT, finance/insurancee and real estate (Marinescu and Wolthoff 2012). Somewhat more surprising are the ethnic and racial gaps in online search, at least in the United States where these data are available. For example, while black and hispanic workers were less likely to look for work on line than whites in 1998/2000, this gap was absent in the subsample of jobseekers who had home Internet access. By 2008/09, blacks and hispanics were just as likely to look for work on line as white jobseekers, even though they were less likely to have home Internet access. In fact, controlling for Internet access and a number of other worker characteristics, blacks were significantly more likely to look for work on line than whites in 2008/09. One possible explanation is that black workers have less access to informal referral networks for jobs than whites; this could force them to use more formal, impersonal methods like responding to job ads and posting resumes (Holzer, 1987), which until very recently was the main way the Internet was used to look for work. HOW DO WORKERS LOOK FOR JOBS ONLINE? While it is clear that Internet job search has increased dramatically over the past decade, what does it actually mean to look for work on line? Importantly, the rates of Internet job search displayed in Table 1 include any kind of online activity to look for work, including . What do we know about the ways in which workers use the new electronic tools that have emerged in the past decade in their job search strategies?

4 4 The first study of this question used data from 1998 (Kuhn and Skuterud 2000). While the survey the authors used did not ask respondents what types of Internet resources they used in their job search, it did ask jobseekers about the broader types of job search activities they engaged in. The results are shown in Table 2, separately for Internet and non-internet job searchers. Interestingly, while most activities are used in similar amounts by these two groups, Internet searchers were much more likely to look for work by sending out resumes and by placing or answering job ads. At the same time, they were less likely to contact employers directly about jobs, and less likely to contact friends or relatives about job possibilities. These differences confirm the notion that, at least in its early days, Internet job search was primarily an anonymous, formal process of looking at (and reponding to) ads and posting resumes. Table 2: Percent of Job Searchers using various search activities, by whether they engaged in any Internet job search, 1998/2000: Looked for Work on Line? Search Activities: No Yes Contacted employer directly Contacted public employment agency Contacted private employment agency Contacted friends or relatives Contacted school/university employment center Sent out resumes or filled out applications Checked unions or professional registers Placed or answered ads Other active methods Source: Kuhn and Skuterud (2000) In a followup survey using data from 2008/9, Kuhn and Mansour were able to directly measure which types of job search activities were done mostly on-line, and which were done mostly off-line. As in 1998/2000, the use of the Internet was quite similar for most job search activities, but two stark differences were apparent: sending out resumes and filling in applications was much more likely to be done online, the reverse was true for contacting friends and relatives about jobs. This reinforces the idea that as recently as 2008/09, the Internet was primarily a tool for formal, anonymous job search, not informal networking activities. Additional information on how workers use the Internet in their job search activity, including the average number of applications and their timing over a search spell is provided by Kudlyak et al. (2012). Finally, it is worth mentioning a small but growing way in which Internet job search is happening: sites like odesk, MTurk, Elance, Freelancer that intermediate the actual provision of labor online. See Agrawal et al. (2013) for a recent study of this type of Internet matching.

5 5 HOW EFFECTIVE IS INTERNET JOB SEARCH? As mentioned, the Internet has become a widely-used intermediary in many markets, and studies of its effects have, in many cases, shown significant improvements in market functioning. For example, one feature of a well-functioning market is that identical products should be priced identically. In a well-crafted early study, Brown and Goolsbee (2002) found that the advent of Internet exchanges reduced the dispersion of prices for a standardized product life insurance. Since the Internet made it easy to compare prices, firms charging above-market rates found it harder to attract customers. Another feature of well-functioning markets is that new matches are made quickly, so few resources lie idle. Pursuing this idea, Kroft and Pope (forthcoming) have shown that the entry of Craigslist -- which is widely used for home rentals-- into a city reduces the apartment vacancy rate. Given this, it is perhaps surprising that most early studies of the effects of the Internet on the speed and efficiency of labor market matching have not shown a beneficial effect. While Stevenson (2007, 2009) found some evidence suggesting that the Internet may have increased the rate at which employed workers change jobs, the first study to actually compare the job search outcomes of persons looking for work on line to similar workers not looking on line was Kuhn and Skuterud (2004). Using data from 1998 and 2000, we found much to our surprise that unemployed workers who looked for work on line were less likely to become re-employed than unemployed workers who did not look on line. This was true even when the authors controlled for a detailed set of worker and labor characteristics, as well as the worker s overall search intensity. One possible explanation for this puzzling result was the formal nature of Internet search: not only is formal search less effective, but persons who use it may be doing so precisely because their informal contacts and social networks are poor. Interestingly, Kuhn and Skuterud s results about the apparent ineffectiveness of Internet job search were confirmed by a very different study conducted by Kroft and Pope, covering the period Kroft and Pope (forthcoming) took advantage of the fact that the U.S. website, Craigslist, expanded dramatically during that period, but at different times in different cities. Since Craigslist quickly became one of the main job boards in those cities, it seems reasonable to ask whether its entry into a city reduced the city s unemployment rate, in the same way it reduced the rental apartment vacancy rate. Consistent with Kuhn- Skuterud (2004), they found no effect of Craigslist on the local unemployment rate. Most recently, Kuhn and Mansour revisited the issue of Internet search effectiveness using survey data conducted in 2008 and We did our utmost to use exactly the same sampling rules and statistical methodology used in Kuhn-Skuterud (2004). Much to our surprise, the counterproductive Internet search effect we found in 1998/2000 was reversed in In particular, we found that Internet searchers have unemployment durations that are about 25 percent shorter than non-internet searchers. While this does not prove that Internet use now has a positive, causal effect on workers search outcomes, it does indicate that something quite substantial has changed over the last decade in the relationship between online job search and job search success.

6 So, why does Internet job search appear to be effective at least for the individual workers who use it in 2008 but not ten years earlier? Turning first to Kuhn-Skuterud s hypotheses, Kuhn and Mansour show that, at least by 2008, Internet search was not a casual, low-intensity activity: Internet searchers engage in just as many distinct types of activities to find work, and do not disproportionately use search methods that are considered passive rather than active. As noted, we do find, however, that even in 2008 Internet search was still largely a formal job search channel; thus, any improvements in Internet job matching must be largely improvements in how these formal channels work. 6 What kinds of improvements might these have been? One candidate that leaps immediately to mind is design improvements on Internet job search sites. Certainly, the major sites have invested a great deal of money and effort into improving their product over this period, and the problems of early job sites such as large numbers of stale resumes: see Brencic are well known. That said, as Table 2 makes clear, a significant share of Internet search involves activities that don t use job boards at all. For this reason, we think that a more likely explanation, which is consistent with the huge rise in Internet job search over the past decade, is network externalities: Internet search is only effective for workers if a large selection of job ads is posted on line, and for firms if many resumes are posted. Through this simple mechanism, greater overall connectivity can raise a network s effectiveness for every individual user. INTERNET JOB BOARDS AS SOURCES OF LABOR MARKET DATA Internet Job Boards are a rich source of information on the number and characteristics of vacant jobs, the number and characteristics of jobseeking workers, and sometimes the results of the matching process between workers and jobs. Job boards have a number of unique strengths compared to traditional labor market data sources, and these strengths are being used to answer new questions that are difficult to study with traditional data. One key strength of job board data is simply that in many cases it s available where traditional, survey-based labor market data is weak or absent. This applies both to highly specialized, local, or rapidly-evolving labor markets in developed countries (e.g. the market for IT workers) where standard labor force surveys don t provide enough detail, and to markets in less-developed countries where standard labor force data is not available or of high quality. In such situations, data that is voluntarily provided by a job board, or even data that is scraped from the Internet by a user-designed web crawler can be a unique source of basic labor market information. Thus, for example, Kuhn and Shen have used scraped (2013a) and internal (2013b) job board data from China, and Delgado Helleseter (2013) has used scraped Mexican job board data to study a number of questions that can t be addressed with existing labor market surveys in those countries. While processing the data (especially converting free-form text fields in job descriptions or resumes into quantifiable data for statistical analysis) can be time consuming, the time and expense can compare very favorably to conducting a survey. A second advantage of job board data is its immediacy: because new job board information can be downloaded almost continuously, it can provide highly current information on labor market trends. For example, Monster.com publishes a monthly index of

7 7 online job demand, based on a number of corporate career sites and job boards in addition to Monster itself. Monster claims that its index provides useful advance information, since job advertising typically precedes actual hiring by one or two months. Monster also releases detailed reports on the on-line market for specific job categories (such as sales) in individual, U.S. cities. Reports like these may one day provide useful information on emerging skill shortages and surpluses, that can be used by workers, firms and local community colleges in refining their search and training strategies. Also, since 2005, the U.S. Conference Board has released its Help Wanted OnLine data series (HWOL), which measures the number of new, first-time online jobs and jobs reposted from the previous month for over 16,000 Internet job boards. It has now replaced the Board s Help Wanted Advertising Index (HWI) of print ads, which was discontinued in July 2008 after 55 years of operation. Another new and creative use of Internet data to measure labor market trends uses counts of keyword searches, such as Google Trends data. For example, D Amuri and Marcucci (2009) and Askitas and Zimmermann (2009) show that Google searches for terms like unemployment office, jobs and the names of job search engines predict unemployment rates in the U.S. and Germany respectively, providing a useful leading indicator. Related, Baker and Fradkin (2011) and Marinescu (2012) examine the link between the changes in the UI benefits and online search, with Baker and Fradkin using Google searches and Marinescu using using internal CareerBuilder data as their measure of search activity. Aside from availability and immediacy, another advantage of job board data is that it provides highly current information on the number and types of workers that firms currently looking for (in terms of qualifications, personality attributes, and even physical and family characteristics), and on the number and types of workers that are currently looking for jobs. Studies of employer demand for these worker characteristics include Delgado Helleseter and Kuhn (2013) on personality, Delgado Hellester(2013) on English language skills, and Kuhn and Shen (2013a) on gender. A final use of job board data is more abstract. Specifically, internal job board data contains detailed information on the mechanics of the labor market, of potentially great relevance to how economists should best model labor markets. It is, in fact, amazing how little labor economists know about the actual mechanics of how workers get assigned to jobs, such as the share of jobs that is filled via formal applications, the number of applications that are submitted to a typical ad, the number of applications workers make and where they apply. Filling some of these gaps, Brencic uses job board data to show that when employers are especially eager to fill a vacancy, employers specify fewer job requirements, provide more details about the job application process, fill the vacancies faster, and are more likely to fill the job with an underqualified worker (Brencic 2009, Brencic and Norris 2009, Brencic 2010). Brencic (2012) also shows that the fraction of job ads that advertise a wage varies widely across different labor markets, and declines with the job s skill level. Brown and Matsa (2012) show that workers online application behavior responds to new public information on corporate distress. These types of results shed useful light on economists theoretical models of labor market search and recruiting, including whether labor markets are better described by either of the two canonical models developed by economics Nobelist Dale Mortensen: wage bargaining models (Mortensen-Pissaridies 1994), or posted-wage models (Burdett-

8 Mortensen 1998). Also in this vein, Marinescu and Wolthoff s (2012) analysis of internal CareerBuilder data shows that labor markets are much more specialized than the users of above models have typically assumed. 8 EMERGING TRENDS AND FUTURE POTENTIAL Personal and professional networking sites, like Facebook and LinkedIn, are a key online tool whose implications for labor markets are not yet fully understood (though Manant et al. (2013) and Acquisti and Fong (2013) have studied whether employers consult applicants social network profiles when deciding whom to interview). Possibly of major importance, job search features have recently been introduced into existing networking sites, including LinkedIn s Apply within LinkedIn button and Facebook. Given Kuhn and Mansour s (2012) finding that informal on line job search is still rare but highly effective when it is used, these development might significantly improve matching efficiency in labor markets. Improvements in job boards suggestion systems which recommend matches to both firms and workers based on what has worked well in the past also offer significant potential (Horton 2012). At the same time, Internet-sourced labor market data, such as HWOL and Google Trends data on job searches, remain in their infancy. Much more remains to be done in harnessing the potential of the Internet as a source of highly detailed, real-time data on the functioning of labor markets. LIMITATIONS AND GAPS A key limitation of studies like Kuhn and Mansour (2012), which compare unemployment durations of Internet versus non-internet searchers, is that they cannot capture the general equilibrium effects of improved search technologies. It remains to be seen whether making each horse in the race for jobs faster yields substantial aggregate benefits in terms of improved match qualities and lower overall unemployment. Thus, additional research on the aggregate effects of the new search technologies would be very useful. And while Internet job boards have tremendous potential as data sources, it s important to remember that these data are not representative of the workers or jobs in the labor market as a whole. By their nature, ads represent vacant jobs, which means that growing and high-turnover industries, occupations and regions will be overrepresented. By the same token, unemployed, mobile and dissatisfied workers, as well as workers in highturnover occupations and industries will be overrepresented in job board resumes. Finally, as we have noted, young and highly educated workers are much more likely to jobsearch on line than other workers. Additional challenges affecting the representativeness of online job postings are multi-homing (the tendency of firms (workers) to post the same ad (resume) on multiple job boards) and duplicates (the tendency of employers on some sites to frequently re-post the same ad until a vacancy is filled. These technical issues need to be resolved to make job board data more useful as an indicator of national patterns and trends. (See Nakamura et al. (2009)).

9 SUMMARY AND POLICY ADVICE 9 Despite negative early findings, recent research suggests that on-line job search might indeed speed up unemployed workers re-employment. If so, the rapid diffusion of Internet job search suggests that the large rise in unemployment in the Great Recession may even understate its severity, compared to previous recessions. Macroeconomic policymakers may wish to take this into account. At the same time, policymakers may wish to use government resources to take greater advantage of the tremendous potential of Internet-sourced data on labor markets. Small infusions of public funding may yield great rewards by supplementing expensive, slow-moving surveys with Internet-sourced data on market conditions and emerging labor market shortages that are both relatively cheap to collect and available in real time. LITERATURE AND FURTHER READING Acquisti, A., and C. Fong (2013) An Experiment in Hiring Discrimination via Online Social Networks, work in progress, Carnegie Mellon University. Agrawal, A.K., N. Lacetera, and E. Lyons (2013) Does information help or hinder job applicants from less developed countries in online markets, NBER working paper Askitas, N., and K.F. Zimmermann (2009) Google econometrics and unemployment forecasting, IZA working paper Baker, S., and A. Fradkin (2011) What drives job search? Evidence from Google Search data, SIEPR working paper Brencic, V. (2009). Employers' hiring practices, employment protection, and costly search: A vacancy-level analysis. Labour Economics, 16(5), Brencic, V., & Norris, J. B. (2009). Employers' online search: An empirical analysis. Industrial Relations, 48(4), Brencic, V. (2010). Do employers respond to the costs of continued search? Oxford Bulletin of Economics and Statistics, 72(2), Brencic, V. Wage Posting: Evidence from Job Ads Canadian Journal of Economics 45(4) (Nov 2012): Brencic, V. (2013) Price-Setting by online platforms: Evidence from online job boards and resume banks, working paper, University of Alberta Brown, Jeffrey and Austan Goolsbee. Does the Internet Make Markets More Competitive? Journal of Political Economy, 110(3) (June 2002): Brown, J., and D.A. Matsa (2012) Boarding a sinking ship? An investigation of job applications to distressed firms, NBER working paper

10 Burdett, Kenneth and Dale T.Mortensen (1998) Wage Differentials, Employer Size, and Unemployment International Economic Review, vol. 39, no. 2, (May), pp D Amuri, F., and J. Marcucci (2009) Google It: Forecasting the US unemployment rate with a Google job search index, Bank of Italy working paper. Delgado Helleseter, Miguel. English Skills and Wages in a Non-English-Speaking Country: Findings from Online Advertisements in Mexico. unpublished paper, University of California, Santa Barbara, Delgado Helleseter, Miguel and Peter Kuhn The Price of Personality: Evidence from Computrabajo, in progress Holzer, Harry J. "Informal Job Search and Black Youth Unemployment." American Economic Review, June 1987, 77(3), pp Horton, J.J. (2012) Computer-mediated matchmaking: Facilitating employer search and screening, working paper, Harvard University. Kroft, Kory and Devin Pope. Does Online Search Crowd Out Traditional Search and Improve Matching Efficiency? Evidence from Craigslist, Journal of Labor Economics, forthcoming. Kudlyak, M., D. Lkhagvasuren, and R. Sysuyev (2012) Sorting by skill over the course of job search, Federal Reserve Bank of Richmond working paper Kuhn, Peter and Hani Mansour Is Internet Job Search Still Ineffective?, unpublished paper, University of California, Santa Barbara, August Kuhn, Peter and Kailing Shen. Gender Discrimination in Job Ads: Evidence from China Quarterly Journal of Economics, February 2013a Kuhn, Peter and Kailing Shen. "Do Chinese Employers Avoid Hiring Overqualified Workers? Evidence from an Internet Job Board" forthcoming, Research in Labor Economics 2013 Kuhn, Peter and Mikal Skuterud. Job Search Methods: Internet versus Traditional, Monthly Labor Review, October 2000: Kuhn, Peter and Mikal Skuterud. Internet Job Search and Unemployment Durations American Economic Review 94(1) (March 2004): Manant, Matthieu, Serge Pajak and Nicolas Soulié. (2013) Do recruiters Like it? Privacy and social network profile in hiring: A randomized experiment, working paper. ADIS, Department of Law, Economics and Management, Université Paris-Sud. Marinescu, Ioana. Online Job Search and Unemployment Insurance during the Great Recession. Unpublished paper, University of Chicago, Marinescu, Ioana and Ronald Wolthoff. Wages, Applications and Skills. Unpublished paper, University of Toronto, December 2012) Mortensen, Dale T; and Christopher Pissarides. Job Creation and Job Destruction in the Theory of Unemployment Review of Economic Studies, vol. 61, no. 3, July 1994, pp

11 11 Nakamura, A.O., K.L. Shaw, R.B. Freeman, E. Nakamura, and A. Pyman (2009) Jobs online, in Autor, D (Ed), Labor Market Intermediation (pp ). Chicago Press. Stevenson, Betsey. The Impact of the Internet on Worker Flows Unpublished paper, The Wharton School, University of Pennsylvania, Stevenson, Betsey. The Internet and Job Search, In David Autor, ed. Labor Market Intermediation Chicago: University of Chicago Press, 2009.