How Competitive Is “Competitive” Procurement in the Social Services?

The scholarship on contracting generally argues that markets for social services are weak and lacking in competition. Using data gathered from Florida’s largest social service agency, the Department of Children and Families, this article adds to the discussion by constructing a more rigorous measure of competition that accounts for the quality of bidding entities. The findings indicate that while the measures used in earlier studies align reasonably well with the raw number of initial responders to competitive solicitations, they tend to overestimate competition when the quality component is included in the analysis. That is, social service markets may be even weaker than previously reported. Furthermore, an examination of the relationship between competition and performance fails to find a significant association.

multiplicity of past studies examining this topic, what has been largely uninvestigated is the extent of competition as determined by the exact number of bidders drawn from contract documents (see Savas, 2002, for an exception) and, more importantly, the quality of those respondents.
The main purpose of this study is to better understand the nature of soft service markets by operationalizing competition in a more stringent manner, namely, the number of acceptable bidders, than has been done in the past. In doing so, I examine how scholars have previously identified and evaluated competition, compare some of these measures to my new construct, and assess whether and how closely they converge with one another. I also investigate the quality of losing bidders to see whether they are likely capable of enforcing market discipline. Finally, I perform a rudimentary exploration of the link between competition and performance to see whether the assertions of market theory regarding the merits of competition hold. Using data gathered from the Department of Children and Families (DCF), Florida's largest social service agency, the results indicate that the existing measures perform reasonably well in gauging competition when considering only quantitative aspects. However, they tend to overestimate the robustness of social service markets when compared with a more rigorous method assessing the quality component. In addition, there is a lack of a relationship between competitive sourcing and vendor performance. On the other hand, losing bidders appear to stack up pretty well when compared with winners on bid scores.
The article progresses as follows: It opens with a discussion of what competition is and why it matters as well as a review of the literature focused on defining competitive thresholds (i.e., How many bidders are needed to maximize the gains associated with competition?). This is followed by a discussion of how the concept has been measured and accounted for in past research as well as the contribution I make in this regard. Next come a review of the DCF's use of contracts and grants, a discussion of the data and how they are collected, followed by analyses of the data. Finally, the article closes with an assessment of the implications of this research and avenues for further inquiry.

Why Is Competition Important? How Much Is Enough?
When markets are not robust in terms of the numbers of buyers and sellers, inefficient allocations of resources tend to result as adequate price signals are not present and actors can influence prices (Vickrey, 1961). While it is possible that a public agency tasked with controlling the market "might be able to create the conditions whereby the marginal conditions for efficient resource allocation could be maintained," the costs of operating this type of a "command" system would likely be prohibitive (Vickrey, 1961, p. 8). The crux of the problem is that such an agency would be dependent on both the buyers and sellers to truthfully and accurately report their marginalvaluations and marginal-costs, respectively. A relatively low-cost solution to this dilemma is the introduction of auctions (Vickrey, 1961). While there are a variety of types of auctions (see Klemperer, 2004;Milgrom, 1989, for good reviews of such institutions), they share the common characteristic of "elicit [ing] information, in the form of bids, from potential buyers regarding their willingness to pay and the outcome-that is, who wins what and pays how much-is determined solely on the basis of the received information" (Krishna, 2002, pp. 6-7). It is important to point out that in this literature, there is considered to be no formal distinction between "normal" auctions, in which auctioneers are the sellers, and "procurement" auctions, in which auctioneers are the buyers (Klemperer, 2004;Krishna, 2002;McAfee & McMillan, 1987). Governments most commonly employ what are termed "first-price, sealed-bid" procurement auctions in attempts to competitively procure social services. In these types of auctions, prospective vendors submit confidential bids that are evaluated by government officials who, based on the information received in the bids, choose an organization with which to contract (Klemperer, 2004;Krishna, 2002).
However, the worry is that if there are not enough bidders for an auction, suboptimal results may be obtained. There are two primary concerns. First, small numbers might give bidders informational advantages in that they can deduce the capacities of their competitors and bid accordingly (i.e., lower than their valuation, but higher than they believe their competitors are capable of). At the extreme, a single bidder, if aware of its status, can offer an absurdly low bid that the government may be forced to take (see Klemperer, 2004, p. 107, for examples). Second, when there is a small number of competitors, collusion might be more common (Selten, 1973). In the case of auctions, this would mean that potential bidders coordinate their efforts to ensure a low price for the agreed upon winner (Krishna, 2002). This would especially be a concern when there are multiple lots for sale such that bidders can organize to ensure the colluders each gain their preferred "wins" (Klemperer, 2004).
While public administration scholars infrequently cite the "auction theory" literature reviewed above, their preferred lens, agency theory (Jensen & Meckling, 1976;Ross, 1973) addresses many of the same issues. In this model, vendors are seen as self-interested, and potentially opportunistic, participants that must be kept "in-line" by contracting governments (see Eisenhardt, 1989 andVan Slyke, 2007, for good discussions). Competition assists the principal (i.e., government) in two critical ways. First, as posited by auction theory, it forces bidders to reveal information they might not otherwise accurately share. In noncompetitive situations, vendors have an incentive to lead the principal to believe that production costs are higher than they actually are in an effort to extract rents. However, this incentive is diminished in competitive procurement environments, since high-priced bids lower the probability of winning contracts. Hence, in the presence of competition, potential vendors can be expected to bid closer to their true production costs (DeHoog, 1984). Second, in what could be considered an extension of auction theory (since auction theory does not directly address performance), competition can also serve to control postaward behavior. Vendors who underperform or are inefficient can be replaced by competitors that are "waiting in the wings." Even well-performing vendors might be at risk if innovative competitors discover improved production techniques that lower their costs below that of the incumbent (Savas, 2000). In sum, according to agency theory competition can be expected to lead to efficient, responsive, and high quality service delivery as producers strive to ensure their survival and maximize profits by providing the best service at the lowest possible price. This dynamic has been termed the "competition prescription" (Kettl, 1993). Cohen (2001, p. 434) highlights the importance of competition in stating that "the problem of organizational waste and inefficiency occurs in both [the public and private] sectors and stems from habits born of monopoly" (see also Osborne & Gaebler, 1992).
While it is generally accepted that "[i]ncreasing the number of bidders increases the revenue on average of the seller" (McAfee & McMillan, 1987, p. 711), an enduring question is exactly how many competitors must there be to gain these purported benefits? Over the years, a variety of economists and management scholars have addressed this question utilizing the auction theory framework. The most common approach has been to examine public or private auctions or markets for goods, commodities, or services to determine the number of participants required to minimize collusion and otherwise ensure attainment of the most competitive price (i.e., the lowest for buyers or highest for sellers). In one of the earliest studies, taking a game theoretic approach, Selten (1973) finds that collusion is less likely if a market has at least five firms. Fewer than this and the environment may become oligopolistic in nature, to the detriment of the buyer of the good or service. Bresnahan and Reiss (1991) examine five professional and retail markets (doctors, dentists, druggists, plumbers, and tire dealers) and come to the conclusion that markets should have three to five participants to ensure the benefits of competition are captured, but the key is the third entrant. While prices continue to fall with the addition of the fourth and fifth participants, the impacts are diminished. Looking at auction markets for municipal bonds and offshore oil leases and National Forest Service timber sales, Brannman, Klein, and Weiss (1987) find that competitive pricing is achieved when there are between about five and eight bidders. Keisler and Buehring (2005) model the impacts of competition when there are economies of scale to be had and government must buy a fixed quantity of a good from the market. They find that the largest reductions in vendor profit occur with the introduction of the third bidder and savings continue through at least the fifth entrant into the market. Their model further indicates that having more than five vendors might drive profits down to essentially zero and make the market unsustainable. In the end they recommend that "government buyers should attempt to have four vendors" (p. 313). MacDonald, Handy, and Plato (2002) study United States Department of Agriculture procurement of five commodities (all-purpose flour, bakery flour, pasta, vegetable oil, and peanut butter) and find that prices generally drop until there are four to five bidders. Finally, Gupta (2002) examines sealed -bid auctions for highway construction in Florida. Her results indicate "that a minimum of six to eight bidders are required to attain the competitive threshold. Once six to eight bidders enter the market, the low winning bid is unaffected by additional entry" (p. 22).
The above indicates it is likely that governments will see substantial cost savings if they can attract at least three bidders. However, attracting four to five appears best for achieving optimal gains. It should be noted that previous study findings are heavily drawn from commonly marketable business trades and the procurement of commodities, and as a result, the competitive thresholds identified in these studies may not be directly applicable to government outsourcing of public services. Nevertheless, the knowledge gleaned from this literature can be informative and serves as guidance to public management scholars as they pursue refinement in measuring competition among the broad scope of government services.

Measuring Competition
A review of the public management literature indicates that scholars generally adopt one of five methods to measure levels of competition. Each has its advantages and disadvantages and the appropriateness of any particular approach depends on the specific research question and design involved in a given study.

Case Studies and General Perceptions of Sufficiency
Estimates of the robustness of service markets are often inferred through the use of qualitative case study methods in which the researcher interviews persons associated with a service, program, or agency (e.g., DeHoog, 1984;Johnston, Romzek, & Wood, 2004;Smith & Smyth, 1996;Van Slyke, 2003). The focus of such studies is generally on acquiring in-depth knowledge of the workings of the related systems and relationships. Competition, per se, is not usually of key interest, although it is an important environmental factor that needs to be accounted for. As such, many of the findings regarding competition in such studies take on almost anecdotal qualities, where interviewees remark on a common lack of suitable vendors or discuss specific procurement competitions that failed to garner much interest from the private sector. As is always a concern in case studies, the generalizability of such findings is in question.
Another technique, most commonly employed in research utilizing large-N surveys, is to ask respondents about their perceptions regarding the sufficiency of competition in general terms. For example, the International City/County Management Association (ICMA) has been conducting their "Profile of local government service delivery choices" approximately every 5 years since 1982 (with the most recent iteration occurring in 2007). In these surveys (starting in 1988), ICMA asks a two-part question. First, respondents are queried as to whether their jurisdiction has encountered any obstacles to privatization. If yes, they are then asked to choose all applicable obstacles from a menu of possible options, one of which is a lack of competent vendors. Such information is, of course, of limited research value in that one cannot identify the true nature or extent of the problem. That is, one cannot discern whether the problem is widespread or confined to a single service, which makes evaluating the impact of competition problematic. This method is notably inferior to the case study method owing to this ambiguity.

Specific Recollections of Sufficiency
Another study using survey methodology adopts a similar approach, but uses a measure that is more fine-tuned and service specific. M. Lamothe (2012a, 2012b) get at competition through the use of a 5-point scale question in which the respondent chooses a point on the scale with "1" indicating there are many vendors in the market and "5" indicating there are very few for a given service. Such a measure has two advantages over the case study and general sufficiency methods. First, it allows researchers to directly link levels of competition to specific services in a large number of jurisdictions. Second, it accounts for matters of degree as opposed to simply dichotomizing competitive and noncompetitive markets.
However, it also has a notable weakness. It introduces subjectivity bias, since what constitutes "many" or "very few" could vary depending on the respondent's disposition regarding contracting. Privatization advocates might argue that having two bidders is sufficient to rate the environment as reasonably competitive, others might have higher thresholds. As such, the scale scores can mean different things across the cases.

The Use of Proxies
Another common technique used to account for competition is the adoption of proxies. This is the most frequently used method in large-N studies at the local level. The most regularly used proxies are metro status (Brown & Potoski, 2003;Greene, 1996;S. Lamothe et al., 2008;Levin & Tadelis, 2010) and population (Brown & Potoski, 2003;Carr, LeRoux, & Shrestha, 2009;Kodrzycki, 1998;Levin & Tadelis, 2010). The logic for both is essentially the same-larger jurisdictions should have better developed markets with more vendors, and thus, should be more competitive (Ferris & Graddy, 1986). Specifically, it is posited that large metro areas should be relatively more likely to contract for services since they are more likely to reap the benefits of market discipline. Another proxy that has been employed is the total number of private firms in the county (S. Lamothe et al., 2008). The assumption in this case is that the existence of greater numbers of firms indicates the likelihood of greater competition.
All proxies discussed above suffer from the same weakness. They are global and are unable to differentiate by service, which assumes the same levels of market development across services within a jurisdiction. While it is likely true that small and/or rural areas have weaker markets generally, it seems probable that even the largest jurisdictions would suffer from noncompetitiveness for some services. These proxies simply cannot address this issue.

Recollections Regarding the Number of Responders
The measures of competitiveness, up to this point, have either suffered from being somewhat subjective or global in nature. However, scholars have attempted to address these issues by employing survey designs that explicitly ask respondents to make their best estimates as to the number of bids received during the most recent procurement effort for a given service (Bel & Fageda, 2011;Fernandez, 2007;Hefetz & Warner, 2012). Such a measure is very useful in that it directly ties market characteristics to the relevant jurisdiction-service pair, which allows for a much stronger test of the relationship between competition and contracting behavior. In addition, it seems to address the subjectivity issue noted regarding the M. Lamothe (2012a, 2012b) measure by using specific reference to the number of vendors when asking questions as to the extent of competition.
However, there are some concerns here as well. One issue is that, to date, most such efforts ask respondents to estimate the number of bidders for a large number of services at one time (both Fernandez, 2007, andWarner, 2012, use the 67 services found in the ICMA surveys 1 ). It may be asking a bit much of respondents to accurately remember such a volume of information without reviewing actual bidding documents, especially when some of the procurement activities may have taken place years ago. For this reason, the accuracy of such measures may be in question.

Counting the Number of Responders
The most straightforward method of assessing the competitiveness of the service market is to ascertain how many potential vendors actually responded to a given competition (e.g., Savas, 2002;Schlesinger et al., 1986). This is done by reviewing competitive tendering documents to examine exactly how many entities participated. Such a technique seems to offer the most precise measure of competition since it utilizes the historical record to establish exactly how a procurement process played out. As such, it would seem superior to the techniques discussed to this point, at least with respect to accurately gauging competitiveness.
Unfortunately, gathering such information is a challenging undertaking and requires gaining access to and examining a large number of contract documents. The level of resources required for such an endeavor is often prohibitive. Hence, researchers utilizing this technique tend to restrict their analyses to a relatively small number of competitions within a limited geographical boundary (Savas, 2002, looked at three social services in New York City and Schlesinger et al., 1986, focused on mental health care in Massachusetts). Therefore, while count measures do appear superior as true indicators of the market environment, the challenges involved with gathering them seems to make them unlikely candidates for use in large-N, generalizable studies.
Furthermore, as constructed to date, they may also overestimate the true competitiveness of the market. This is so because, in any solicitation, there is a significant probability that some responders are completely ill-prepared to undertake the given task and have no chance of winning the competition. The key to unleashing the purported benefits of competition is that winners should perform well (i.e., high quality, low price) because they fear replacement by rivals (Savas, 2000). However, nonviable vendors or notably inferior ones do not serve this function. A strong argument can be made that such bidders should not be counted since they do not assist the contracting government in leveraging the benefits of competition (i.e., they do not serve as credible replacements). Thus, only acceptable bidders should be considered.

Acceptable Bidders
In this study, I gather information to construct a narrower, yet more rigorous measure of competition, and use it as a baseline to compare other measures from past studies. There are two noncontroversial ways in which bidders might be identified as lacking sufficient quality. First, there are commonly administrative deadlines, such as filing dates, which must be met for bidders to move forward in the process. Respondents who miss closing dates (also known as "fatal criteria") are generally dropped from the competition. It seems fairly straightforward that such organizations should not be counted among the bidding entities as they are never truly "in the game" even though they may show up on the original responder rolls and have their names carried forward for administrative reasons.
Second, contracting governments might set minimum bid scores that must be achieved for the bidder to be considered for awards. It seems a stretch to argue that vendors who do not reach minimum requirements and, therefore, are eliminated from competition, are likely candidates for enforcing market discipline. Likewise, these vendors should not be counted in an analysis of competition. For the sake of this research, that subset of responders who meet the fatal criteria and score at or above the minimum cut-off are termed acceptable bidders-a more stringent measure than the raw number of initial responders. While gathering information regarding acceptable bidders poses even greater challenges to researchers due to its intensive resource commitment, this type of measure serves two useful purposes: first, it provides an objective as well as conservative baseline measure for competition from which to evaluate the efficacy of other estimates. Second, it may offer a more realistic picture regarding the true competitiveness of the market for social service outsourcing. 2

The Florida DCF
The DCF is Florida's largest social service agency and offers a wide array of services to assist and protect vulnerable citizens. It divides its services into 10, sometimes overlapping, program areas which are reviewed in Table 1.

DCF Contracting
DCF contracts extensively for service delivery and has done so for many years. For example, a report written for the Florida legislature states that as of October 1, 2006, it held 966 contracts with an annualized value of $1.3 billion and a total (including multi-year commitments) value of $3.8 billion (Florida Legislature, 2006). However, due to the nature of these services, many of these contracts are statutorily exempted from competitive tendering processes that are required for other similarly priced contracts. 3 It is important to point out these exemptions remove the requirement for competitive tendering, but do not ban the procedure. In fact, DCF guidelines encourage competitive solicitation even for exempted services. Still, most such contracts are noncompetitively procured, although use of competitive procurement has increased recently (M. Lamothe & Lamothe, 2009). The most common competitive procurement techniques used by the Department are "Requests for Proposals" (RFPs) and "Invitations to Negotiate" (ITNs). Both methods evaluate bidders on specific and fairly sophisticated sets of criteria and are used when services are not easily described and, hence, not good candidates for unit price bidding (i.e., bid price alone, is not sufficient for choosing a vendor). Most commonly, a number of evaluators score the bidders independently and these scores are averaged to produce a composite, overall bid score. The potential vendors are then compared on these aggregated values. In general, the bidder with the highest score is awarded the contract although this need not be the case if two bidders are very close, but the lower one shows superiority on a pertinent value (e.g., ties to the community) not properly accounted for in the formal evaluation criteria.
While, as described above, exemptions are common, DCF nonetheless utilizes competitive techniques quite often. A snapshot taken from Florida's Accounting and Information data system (FLAIR-discussed in more detail below) shows that in 2008, 202 of DCF's 777 (26.0%) health and human service contracts were coded as utilizing competitive procedures. However, a closer examination of this database showed that the number "202" might be a bit inflated. For example, some of the contracts that were listed as pertaining to social services, where actually for technical support and a few were reported to me by their contract managers to have been noncompetitively procured. In addition, as expanded upon below, some actually represented grants. Regardless, a sizable portion of contracts utilize competitive procedures. This frequent use of competitive tendering combined with the sophisticated nature of the processes used (i.e., the use of RFPs or ITNs), makes DCF an ideal source of information regarding the competitiveness of a wide variety of social service contracts-not only can the number of bidders be identified but also the quality of the bids, as revealed by bid scores, can be examined and factored into the analysis.

DCF Grants
The DCF administers a number of grant programs designed to assist local agencies in pursuance of their efforts to address social problems. Such grants are usually awarded through competitive processes, which may consist of statewide or regional competitions. In many ways, the process of securing grant money is very similar to winning competitive contracts. Both require agencies to submit detailed applications to demonstrate they have the capacity and expertise to deliver the services for which they seek funding. As with contracts, these applications are reviewed by teams of public managers that assign scores to each application based on a variety of criteria. These scores are then used to rank candidates and make awards. In many cases, grants are also managed similarly to contracts in that they are given "contract" numbers, entered into the FLAIR database discussed above, and assigned to contract managers who monitor and oversee the grantees. Further like contracts, grants can have varied timeframes, with some rebid annually and others spread out over a number of years.
In this research, I have gathered data on two such competitive grant programs: Prevention Partnership grants (PPGs) and Emergency Shelter grants (ESGs). The Prevention Partnership Program is federally funded through the Substance Abuse Prevention and Treatment Block Grant and is administered in Florida by the DCF in collaboration with the Florida Department of Education and the Florida Department of Juvenile Justice (Florida Department of Children and Families, 2009). The goal is "to encourage school/community substance abuse prevention partnerships" by "fund[ing] effective programs and strategies that are relevant to community prevention needs" (Florida Department of Children and Families, 2009, p. 4). The program was begun in 2001 and the awards are for 3 years. "Applicants [currently] may request up to $150,000 to conduct direct and/or indirect prevention activities and prevention capacity-building activities" (Florida Department of Children and Families, 2009, p. 3). Only schools, school districts, or nonprofit organizations are eligible to receive funding-for-profit entities are expressly forbidden from winning PPGs (Florida Department of Children and Families, 2009). These grants are awarded regionally, resulting in the running of multiple concurrent competitions in award years.
ESGs are also federally funded, through the U.S. Department of Housing and Urban Development (HUD). HUD directly funds many cities and counties, but also sets aside money for the state to distribute to areas that have not received awards (Florida Department of Children and Families, 2010a). Florida does so through the use of statewide grant competitions. ESGs are annual grants (although the 2010 cycle utilized a 2-year window) and have two components. Facility grants can be used for emergency shelter renovation and the supply of essential services, such as assistance in obtaining permanent housing or counseling (under restrictions not germane to the current research effort; Florida Department of Children and Families, 2010a). Prevention grants aim "to serve persons at imminent risk of losing their permanent housing and becoming homeless" (Florida Department of Children and Families, 2010b, p. 3). Both types are currently capped at $100,000 and, as with PPGs, for-profit firms are ineligible. Rather, these grants are restricted to local government units and secular and religious nonprofit 501(c)(3) entities.

Managing Contracts Versus Managing Grants
In this study, I examine both contracts and grants in an effort to better understand the competitive nature of social service delivery in Florida. It is reasonable to question whether these two methods of funding service delivery are similar enough to warrant their inclusion in a single research design. Simply put, from a management perspective, the two are strikingly similar. As discussed above, the competitive procurement processes utilized are quite comparable. Furthermore, the management of contracts and grants are nearly identical. A passage from DCF's Guide to grant solicitations and awards highlights this point: Subsections 287.057 (15), and 402.7305, F.S., require the Department, for each contractual services contract, to designate a Department employee to be responsible for enforcing performance of the contract terms and conditions and serve as a liaison with the contractor. While there is no corresponding requirement in law that creates a similar requirement for grants, the Department's policy is that the fundamental premise that underlies the statutes is applicable for grants as well as procurement contracts. (Florida Department of Children and Families, 2010c, pp. 6-2) In a practical sense, this means that grant managers review and approve invoices as well as undertake accountability efforts, such as on-site visits, in the same manner as their contract manager brethren. There are a few differences worthy of mention; however, none are particularly relevant to accountability issues. Since grants tend to be narrower in substantive scope, they are more likely than contracts to be governed under a single funds disbursement regime. For example, funds for ESGs are disbursed on a cost reimbursement basis, while funds for a mental health contract may be paid out in a variety of ways (e.g., per unit cost, lump sum) for the different components of the contract. In addition, due to the way in which grant monies are allocated, DCF has more flexibility in "moving money around" with grants than with contracts. Other than these minor differences, contracts and grants appear to be treated nearly identically. 4

Data Collection and Measurement
Competition in this study is measured in a variety of ways so as to allow for a comparison of alternate methods against the baseline metric of the number of acceptable bidders, described above. Specifically, document-based counts of all responders and acceptable bidders, recollections regarding the number of bidders, and proxy measures are gathered. Furthermore, information on vendor performance is collected to allow for a test of the relationship between competition and performance.
I survey Florida Department of Children and Families contract officers to gather the information required to construct the count measures. These data include a mix of contract and grant competitions for social services. As Cooper (2003, p. 85) points out, there are three basic types of contracting officers. Procurement contracting officers (PCOs) solicit and select vendors, administrative contract officers (ACOs), manage contract execution, and termination contract officers (TCOs) ensure continuity of service delivery and assist in smoothing transitions when vendors are replaced. The officers surveyed for this project are ACOs, although they might also serve as TCOs, as necessary. ACOs, who are also referred to as "contract managers" in the Florida context, do not participate in procurement efforts. However, they are the ones who maintain the contract file during post-award periods, which gives them access to the bid records and other pertinent information.
In the spring of 2008, I obtained a snapshot from the Florida Accounting and Information Resource (FLAIR) database. This resource contains all of the contracts administered by DCF at that moment in time. From these data, I isolated all social service contracts (as opposed to contracts for technical support, for example) and further identified the subset of social service contracts that underwent competitive procurement efforts. As previously discussed, approximately 200 contracts qualified under both criteria. These contracts are administered by 88 5 managers indicating that, on average, each manager is responsible for about 2.3 contracts. On July 22, 2009, all managers were contacted with a request to fill out a survey and submit supporting documentation as they felt would be helpful. A follow-up effort was conducted during the winter of 2010. Phone contacts were also made to try and solicit more responses and to clarify existing responses, if needed. In the end, information on 49 contract procurements and 35 grant competitions were received from 22 DCF contract officers. From these data, I identify the number of responders, the number of acceptable bidders, as well as estimates of contractor performance. The number of responders includes all entities that responded to the RFP or ITN, regardless of how far they progressed in the process. Acceptable bidders are the subset of responders that met all required deadlines and whose bids scores were at or above the stated minimum acceptable values set for the competition.
In addition, I adapt Hefetz and Warner's (2012) measure as a third way to account for perceived competition. Specifically, I identify services from their city and county level research that are similar to the ones provided by DCF and assign their estimates regarding the number of bidders to these state-level services to see whether they are reasonable approximations of the levels of competitiveness found through my DCF data gathering. 6 The final competition data collected are two proxy measures. The first is metro status which, as reviewed previously, is very commonly used in the literature to account for competitiveness of the contracting environment. The second is a population measure, which shares theoretical underpinnings with the metro status variable. Both tap into the idea that more densely populated areas likely offer more vendor options. However, applying them to the current research requires some adjustment because, unlike the local level studies from which they are borrowed, DCF service delivery is a multi-jurisdictional affair. The primary DCF service delivery unit is the circuit. There are 20 such circuits, which align with Florida's judicial circuits, and each contains between one and seven counties, depending on how populous the counties are. 7 As such, adjustments must be made. Metro status is defined for each procurement competition by first assigning each county in its circuit of origin a value such that 1 = a county that contains a core metropolitan statistical area (MSA) city, 2 = a county in an MSA that does not contain a core city, and 3 = a county not located in an MSA. The metro measure is aggregated over each circuit by taking the weighted (by population) average of these scores for all counties in the circuit and rounding to an integer value. In the end, only two contracts and one grant competition code "3," so they are combined with "2s" to form a core versus noncore dichotomy for the sake of analysis. 8 Population density is also calculated for each circuit and serves as the second proxy measure of competition for each contract or grant. 9 In addition to the above data gathering efforts, I also briefly interviewed two senior managers in DCF's contract office (the staff director and the senior policy writer) in an effort to elaborate on some of my findings regarding competition (discussed below). The interview, which was done over the phone, took place on November 17, 2011, and lasted about 35 min.
Performance is ascertained by asking the contract managers to rate the performance of the winning vendors on a 5-point Likert-type scale ranging from performing "very poorly" to "very well." Contract managers serve as the point of contact between the vendor and the department and collect and organize all relevant information regarding vendor performance. For example, if a monitoring effort uncovers problems that warrant corrective action, the manager is the one to record the problems in the contract file, contact all relevant parties, and coordinate remediation efforts. As such, they are in a good position to see the "big picture" regarding overall performance and should be able to assign useful ratings. Furthermore, as mentioned above, contract managers do not generally participate in contract award decision making. This is important because, if they were major players in the process, how well the contractor is doing might be seen as a reflection on the managers' own job performance (since they chose them). Such a situation would likely lead to a bias toward positive vendor evaluations that is, fortunately, avoided in the current circumstance. 10 In the end, however, managers seem reasonably satisfied with their vendors in that only 3.6% of cases scored one or two on Performance. Therefore, these categories were collapsed into the third grouping resulting in a three-category, ordered measure of performance such that "1" = "poor/adequate," "2" = "well," and "3" = "very well." Unfortunately, performance estimates are not available for many of the competitions. Generally speaking, data for grant competitions were gathered from aggregate reports that made collecting information on specific vendor performance difficult (i.e., I could identify the bidding agencies, their scores, and the like, but generally could not match them to managers to offer their opinions on the quality of service delivery). And, a number of contract managers chose not to provide performance measures. In the end, I have performance information for 31 competitions (25 contracts and 6 grants). In addition to the data collected above, I also have information on 52 noncompetitive DCF social service contracts. These data were collected in a previous effort focused on noncompetitive procurement but are compatible with the current data in that an identical contractor performance measure was gathered. 11 Hence, these noncompetitive cases can be used in conjunction with the current data to examine the relationship between competition and performance. If market theory is correct, competitively procured vendors should outperform noncompetitive ones and performance should improve as the level of competition increases (or at least until competitive thresholds are reached).

Findings
This section reports the results from the comparisons of the different competition measures and the analysis of the relationship between competition and performance as well as a discussion of the quality of losing bidders. The overall findings suggest that the recollection-based count measure reasonably approximates the actual document-based count measures in terms of estimating the number of initial responders. The measures do not converge as well when competition is defined in a more stringent manner (i.e., counting the number of "acceptable" bidders). Proxies do not seem to perform as effectively as the recollection-based measure in assessing competition although they are able to provide a rough feel for the existence of competition. Substantively, the findings of this study confirm the common presence of weak competition across different human services. None of the service areas examined (both contracts and grants) exceeded the minimum competitive threshold of three (keeping in mind that the literature indicates that four or even five bidders might be preferable). Lack of competition deepens when accounting for the viability and quality of bidders, as measured by the number of acceptable bidders. However, losing bidders do appear to be of similar quality as the winners. Finally, performance is neither meaningfully related to competitive sourcing nor significantly associated with the extent of competition. In the following, I review the findings in details. Table 2 shows how three competition measures compare (two document-based count methods and the recollection-based estimate of the number of bidders). While there appears to be some variation by service type, the table clearly indicates that by no means, based on any measure, is competition robust. Concerning the 49 contracts for which I have data, the average number of initial responders (third column) is a less than adequate 1.73. The far right column contains the averages for the services (1.88) as collected by Hefetz and Warner (2012). The extent of competitiveness predicted by these perception-based measures appears to align reasonably well with the initial responder counts. A notable finding is that this measure and the actual counts of initial responders appear to overestimate the true competitive nature of the bidding process as assessed by the number of acceptable bids (1.29 from the fourth column). The difference between the initial and acceptable bidders, 0.44, is statistically significant (p = .001) and also of substantive import-losing nearly half a bidder, on average, is impactful when so few bidders are available in the first place. Behind the averages, is some information worthy of further discussion. Of the 49 contract competitions, 24 (or 49.0%) reported only one responder and 34 (or 70.8%) 12 had only one acceptable bid. Furthermore, only 5 (10.2%) had three or more responders. However, when only acceptable bidders are included, just 3 cases (6.3%) achieve the minimum competitive threshold and none are above it.

Grants: Responders Per Award and Acceptable Bidders
The results for grant competitions are very similar (see Table 3). Once again, competitive thresholds are not achieved, or really even approached, and accounting for acceptability leads to a significantly lower number of bidders (difference = 0.34, p = .000). While grants might appear a bit more competitive than contracts, with on average 0.11 more responders and 0.21 more acceptable bidders per award, the differences are not statistically significant (p = .625 and .234, respectively). Looking inside the numbers, 7 of the 35 grant competitions (20.0%) received only a single responder per award, but this number increases notably when only acceptable bidders are counted-18 or 51.4% of cases, averaged one bidder under the more selective standard. The above adds further support to the common finding in the literature that competition in social service markets tends to be anemic. In addition, the importance of accounting for the acceptability of bidders is demonstrated-simple responder counts tend to overestimate market potential.

Proxy Measures
I also look at proxy measures of competition. As mentioned above, due to the paucity of nonMSA contracts and grants, the metro status variable is collapsed to a dichotomous measure-core versus noncore. This allows me to assess the efficacy of metro status through the use of difference of means tests. The result, found in Table 4, indicate that core and noncore contracts and grants may be differentiable in terms of the raw number of potential bidders that respond to solicitations. The relationship is in the posited direction in that core areas appear to have slightly higher numbers of responders than their noncore peers. Specifically, 1.97 bidders, on average, respond in core areas while only 1.55 bidders respond in noncore areas. The difference, 0.42, is marginally significant with a p value of .079. However, the different types of locales are not differentiable when the more stringent measure of acceptable bidders is employed.  Note. Six contracts and nine grant competitions were conducted on a statewide basis and are, therefore, not included in this analysis.  (17) 90.0 (12) Note. Since grant competitions generally have multiple winners, the lowest winning bid score is used to calculate the values reported in this table.
In addition, I examine how well the other proxy, population density, accounts for competition, through the use of correlation analysis. Bivariate correlations between population density and the two count measures lead to results somewhat similar to those attained using metro status. However, while in the posited directions, neither correlation is statistically significant, but population density's relationship with the number of responders is stronger than with acceptable bidders (r = .163, p = .180 and r = .013, p = .914, respectively). Thus, while metro status, and to a lesser degree population density, might be reasonable proxies for the number of responders, neither seems to adequately reflect the more rigorous measure.

Quality of Losing Bidders
Beyond the simple number of bidders, it is also imperative that the losers are "up to the task" of enforcing market discipline. That is, they must be capable of serving as credible replacements. Generally speaking, the universe of social service contractors in any given area is small and they tend to know one another. As such, winners should be able to fairly accurately gauge the quality of their competitors. Furthermore, the contracting process in Florida is very transparent. Contracting documents are public records and award announcements very often include the scores for all bidders. In instances in which there is an oversight in this regard, these documents are subject to Florida's sunshine laws and would be readily available through a request to the contracting agency. 13 Hence, incumbents are likely to have a pretty good feel for the security of their position, which could impact behavior. With this in mind, I turn to an examination of the losing acceptable bidders. Specifically, I compare the bid scores of the first, second, and thirdplace losers, when available, to those of the winners. The results for this analysis are found in Table 5. The second column compares the winning score to the maximum possible score. 14 Generally, we see that DCF appears to set rather high standards that vendors find difficult to reach. The average winner, for the 73 competitions for which I have relevant data, receives only 80.0% of the possible points. Contract winners seem to do a bit better than grant winners in this regard; but overall, we see what would traditionally translate to "C+" to "B−" grades.
The staff director for DCF contract personnel, whom I interviewed, offers some interesting insights concerning the apparent discrepancy between DCF maximum scores and vendor achievement in this regard. His impression is that this finding does not indicate DCF standards are too high nor that vendors are generally lacking in quality. Rather, he argues that two factors are likely at play. First, DCF tends to rate bidders on a variety of 3-point scale items and aggregate the results. As such measures are rather rough, any subtractions can lead to large reductions in overall score. Furthermore, the items are commonly weighted as well, which can further exaggerate the impact of any deductions. Second, most of the vendors submitting bids are small nonprofits that lack management capacity and, therefore, do not tend to present themselves well in such competitions (i.e., they do not put together polished bid packages). They are also unlikely to utilize consultants to assist them in preparing solicitation responses. Hence, as the staff director puts it, you might be dealing with an excellent service delivery organization that simply does not look that good on paper.
These concerns are becoming more relevant, he adds, as for-profits enter the market. They tend to either use consultants or have the internal capacity to compile very professional proposals, even when they are not particularly effective service deliverers. To address these inconsistencies between for-profit and nonprofit bidding tendencies, DCF has started to adjust their bidding procedures to include bidder presentations which better allow DCF to determine the true potential of responding entities.
Moving to the third column (Table 5), we see that the losers with the highest scores appear to be competitive to winners, especially with regard to grant competitions. Overall, the scores for first losers average 95.1% of the winning score. There is only a slight drop-off when we move to the second losers. While there are no third losers regarding contracts, because of the nature of grant competitions (i.e., multiple awards per competition), it is much more common to be able to identify three (or more) nonwinning grant bidders. And, these vendors tend to stack up pretty well, with their scores averaging 90.0% of the lowest winning bid score. So overall, when there are losers (but remember, this is commonly not the case-a bit over 60% of the competitions in this analysis attracted only a single acceptable bidder per award), they tend to be pretty similar to the winners in terms of their bid scores.

Competition and Performance
Although the results to this point indicate low levels of competition for DCF social service contracts and grants, might there be something about the competitive procedures themselves that can lead to benefits? The most common expectation in this regard is that competitive procurement should lead to cost savings. However, this is not an appropriate metric in the current circumstance. DCF tends to set the value of social service contracts ahead of time (i.e., how much they will spend) and compares vendors on what they offer for the given sum of money. Florida is not unique in this respect- Schlesinger et al. (1986) discuss similar practices in Massachusetts. Furthermore, cost savings are not applicable to grant situations where the goal is to get predetermined funds to qualified vendors. For these reasons, I examine whether there is evidence that the use of competitive processes might be able to weed out weak vendors and, therefore, promote better performance than might be expected when awards are made noncompetitively.
To accomplish this, I run a series of ordered logit models in which Performance (described above) is regressed on different measures of competition, after controlling for other possible influences. The measures of competition are as follows: (a) Competitive process is dichotomous and codes "1" if competitive procurement procedures were used, (b) Number of bidders identifies the number of responders for the given solicitation, and (c) Acceptable bidders accounts for the number of bidders who met the fatal criteria and minimum bid scores. To protect the results from potentially spurious relationships, a number of control variables are utilized.
Vendor ownership is often posited to be related to performance. For example, privatization advocates, such as Savas (2002), argue that for-profits should outperform nonprofits and other government vendors because the profit motive gives them incentives to deliver high quality services at low costs. Weisbrod (1977), on the other hand, emphasizes the goal congruence found between nonprofit vendors and their contracting governments to assert that nonprofits might outperform their for-profit peers. With this in mind, For-profit and Nonprofit account for the type of vendor (other government contractors serve as the reference category 15 ). In addition, I account for the management capacity of the district by borrowing M. Lamothe and Lamothe's (2010) measure, which is dichotomous with "1" identifying high capacity districts. 16 The expectation is that higher capacity districts should be better able to manage the entire delivery process (i.e., vendor selection and contract execution) and achieve better results. Furthermore, I control for two district characteristics, population density and per capita personal income, which might influence contracting outcomes. Finally, I include district dummies to protect against distinctlevel idiosyncrasies that might play a role, but are not captured in my other independent variables. 17 The descriptive statistics for all of the variables included in the performance model can be found in Table 6 and descriptions and source information can be found in the appendix (Table  A1).
The results of the analysis are presented in Table 7. 18,19 In none of the three models is the variable capturing competition statistically significant. Rather, things like vendor type, population density, and the district in which the contract resides influence perceptions of vendor performance. For example, other government contractors seem to outperform their for-profit peers. While this finding is in-line with the results of previous research (M. Lamothe & Lamothe, 2010), it is important to point out that the paucity of government vendors (n=5) in the current data set requires us to not make too much of this result. In addition, more densely populated areas  a As "performance" is a 3-point ordered variable, the models are estimated using ordered logit. b 28 competitive and 51 noncompetitive contracts (27 competitive for "acceptable bidders" model). *p < .10. **p < .05. ***p < .01, two-tailed.
appear to be associated with poorer performance, on average. Why this would be the case is unclear, but it might be symptomatic of the general challenges associated with delivery of social services in urban areas. That is, urban areas often represent more difficult environments with regards to crime and poverty than their suburban and rural peers. However, the main driver of differences in performance ratings appears to be district-level idiosyncrasies, with Districts 2, 3, and 4 rating their vendors more poorly than the reference districts.

Discussion and Conclusion
In this article, I offer a more rigorous test of competition in the social services than has been done in the past. While I find some evidence of convergence for past measures, the concern is that they all seem to overestimate competition when compared to the number of acceptable bidders. As such, competition in the social services may be even more restrained than previously believed. Below, I discuss several implications of my analysis for the contracting literature. This research demonstrates the efficacy of the more precise measure of competition and the importance of ensuring that public managers are aware of the difference between simple responders and acceptable bidders when answering questions on competitiveness. On the other hand, by looking at the bidder quality closely, I find evidence that if at least two acceptable bidders can be located, the loser may well be of high enough quality to serve as a credible replacement for the incumbent should it not perform well. This is potentially good news in that such competitors are more likely than nonviable options to enforce market discipline. However, the relatively high probability that competitive thresholds will not be reached is a continuing concern as anticompetitive behaviors are more likely in such environments. These findings highlight the possible value of managed competition (Savas, 2000) where public agencies compete with private vendors to win contracts. Public bidders can serve to increase the competitiveness of the process and in-house capacity adds valuable information that should assist contracting governments to better understand production costs and best practices (Miranda & Lerner, 1995;) allowing them to be smart buyers (Kettl, 1993). As such, in-house capacity may also lower the substantial time and resources that might be needed to recruit multiple vendors in an effort to manufacture competition (Johnston & Girth, 2012).
I also explore the link between competition and performance and do not find evidence of a relationship. While this result is not unique (e.g., M. Lamothe & Lamothe, 2010), it is contrary to the assertions of contracting theory and calls into question the efficacy of efforts to increase competition. Combined with the above result, the message to public managers is that contracting for social services must be done with care. The advice of scholars like Cooper (2003) and Sclar (2000) that "management matters," must be taken very seriously. Governments without the capacity to nurture and maintain strong working relationships with their vendors, above and beyond simple monitoring, to include assisting them in improving service delivery as necessary, will likely not be able to get the best deal for the public (Cooper, 2003) regardless of the presence of competitive sourcing. Of course, it is important to point out that some scholars argue that competition can actually diminish performance in social service contracting (Johnston et al., 2004;Milward & Provan, 2000). This is so because it can lead to service disruptions for clients as they are forced to switch providers when contracts change. Competition can also make provider networks unstable as members are replaced. This can disrupt the coordination needed to maximize network outputs and lead to learning curve costs as new vendors try to "get up to speed" regarding their network obligations. From this perspective, a lack of competition is not nearly such a concern. However, the need for governmental capacity is not diminished.
As with all research, this study has a number of weaknesses that need to be addressed. Even though this article accounts for a more expansive set of social services than has been done in the past, it is still limited to a single state, Florida, and the number of competitions, 84, is fewer than ideal. Future efforts, accounting for other jurisdictions and possibly greater numbers of procurement efforts, should be undertaken to verify the findings presented here. A closer examination of the link between competition and performance is also warranted. There is a general paucity of such studies that focus on social service contracting and nearly none, to my knowledge, beyond the current analysis, that link performance to something beyond a dichotomous measure identifying competitive processes (i.e., no measures of the numbers of bidders). While this study makes such an effort, there is certainly room for improvement. To begin, a more expansive measure of performance would be helpful. As previously discussed, this could be challenging in that if data are gathered on a variety of services, constructing compatible measures would be difficult. However, using a method similar to that employed by Romzek and Johnston (2005) in which they interview a variety of actors and review pertinent documents to rate relative performance, might be appropriate. In addition, a greater number of data points (the current research has performance information for only 28 competitive awards) would strengthen the generalizability of any findings. Notwithstanding these limitations, I feel this article makes a meaningful contribution to the contracting literature through its more rigorous examination of competition and hope it spurs further interest and efforts moving forward.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes
1. It is possible that Bel and Fageda (2011) avoided this problem as they focus their efforts on a single service (solid waste collection). However, their data are drawn from a larger survey on the production of local services more generally, so it is unclear how many services respondents were asked to provide information for. 2. A reviewer raised two interesting points on competition worthy of discussion. First, this study does not explicitly account for managed competition in which in-house entities bid on contracts. While a generally valid concern, the Department of Children and Families (DCF) simply does not utilize this approach for the services in question. While doing so might be wise as it ensures an additional bidder, political circumstances in many instances preclude this option (i.e., DCF is under a lot of pressure from the legislature to contract with the private sector). Second, maintenance of competitive markets could be enhanced by breaking contracts into multiple components or a variety of small contracts in close geographic proximity so as to ensure vendors fairly regular opportunities to win contracts. Such actions would serve to keep potential vendors in the market, as losing bids do not necessarily imply sustained droughts from government funding. It could also enhance performance since vendors would have constant incentives to burnish their reputations to increase their chances for success during the myriad contracting opportunities. In the current instance, my data do not allow me to account for such activity, but it seems unlikely that DCF is employing these strategies since it has been showing a tendency toward consolidating contracts, both in terms of geographic area and contract length, over the past decade or so (M. Lamothe & Lamothe, 2009). 3. Specifically, contracts in excess of $25,000 are legally required to undergo competitive procurement unless an exemption applies. See 2010 Florida Statutes, Title XIX, Chapter 287.057 for a discussion and listing of exempted services. 4. The information in this paragraph was gleaned from discussions and email correspondences with a variety of DCF personnel who participate in contract and grant management. 5. Just as a number of the contracts in FLAIR were misidentified as either being health and human services or having been competitively procured, as discussed above, the information on some of the contract managers also appears to have been incorrect or out of date. Two initial contact emails "bounced back" indicating the person was no longer at DCF. Efforts to call managers who had not responded to my initial electronic inquiry found 11 managers not listed in DCF's directory, again indicating they might no longer be employed with the agency. Therefore, overall, the actual number of contract managers appears to be closer to 75 than 88. 6. Specifically, I use Hefetz and Warner's (2012) mental health and child welfare program categories for my mental health and family safety service areas, respectively. I combined two of their categories for each of my other measures (hospital and public health for health services and elderly and mental health programs for adult protective/substance abuse services). 7. Prior to 2008, DCF was organized into 14 districts which contained between 1 and 14 counties. In addition, to the current circuit scheme, DCF has six larger administrative units, regions, which consists of between two and five circuits. As the data collection for the current project took place during the reorganizational transition from districts to circuits (and their related regions), the contracts and grants included in this analysis are generally a mix of circuit and district-based. For further information on the 2008-2009 transition, readers are directed to Florida Senate (2009). 8. This is very similar to the coding scheme used by Hefetz and Warner (2007). 9. The source for the coding of the metro status variable was the U.S. Census Bureau (1999). Population density was derived from University of Florida (2005). 10. Of course, a convincing argument could be made that such a subjective measure is inferior to more objective measures such as client outcomes. While this is likely true, due to the nature of the project, available data do not allow for the use of objective measures. Generally speaking, no such information exists in anything resembling a collectible form. Furthermore, I look at a variety of different types of services and constructing objective measures that are comparable across the services would be a