Procuring Firm Growth:

Transcription

1 Procuring Firm Growth: The Effects of Government Purchases on Firm Dynamics Claudio Ferraz PUC-Rio Frederico Finan UC-Berkeley Dimitri Szerman CPI/PUC-Rio November 2014

2 Motivation Government purchases have been widely used as a tool to foster economic activity and increase employment

3 Motivation Government purchases have been widely used as a tool to foster economic activity and increase employment Significant share of this spending is done through public procurement OECD countries spend on average 13% of GDP on public procurement

4 Motivation Government purchases have been widely used as a tool to foster economic activity and increase employment Significant share of this spending is done through public procurement OECD countries spend on average 13% of GDP on public procurement Public procurement has become a policy instrument for encouraging firm growth and innovation, and strengthening industrial capabilities

5 Motivation Government purchases have been widely used as a tool to foster economic activity and increase employment Significant share of this spending is done through public procurement OECD countries spend on average 13% of GDP on public procurement Public procurement has become a policy instrument for encouraging firm growth and innovation, and strengthening industrial capabilities Whether public procurement is effective in stimulating economic activity ultimately depends on whether government contracts help firms overcome barriers to growth

6 Motivation Empirical literature on public procurement offers few insights into this question

7 Motivation Empirical literature on public procurement offers few insights into this question Difficult to estimate these effects because firms are not awarded government contracts at random

8 Motivation Empirical literature on public procurement offers few insights into this question Difficult to estimate these effects because firms are not awarded government contracts at random Governments target specific types of firms More productive firms when picking winners Less productive firms when political connections are important

9 Motivation Empirical literature on public procurement offers few insights into this question Difficult to estimate these effects because firms are not awarded government contracts at random Governments target specific types of firms More productive firms when picking winners Less productive firms when political connections are important Winning a government contract is likely to be correlated with unobserved productivity shocks

10 In this paper We examine whether firms that win government procurement contracts in Brazil grow more compared to firms that compete but win less contracts We assemble a large and novel data set combining matched employer-employee data for the universe of formal firms in Brazil universe of federal government procurement contracts over the period of 2004 to 2010 We estimate these effects for over 47,000 firms that participated in over 6.5 million lots auctioned off by the federal government during this period

11 Our approach To estimate these effects, we propose a new approach Auctions end at random Auctions end unexpectedly within 30 minutes based on a random draw from U[0, 30] No proxy-bidding Firms bid manually and routinely outbid each other several times until auction ends (at random) Close auctions We observe the entire distribution of bids Identify auctions in which the difference between the winning and second-place bid is less than 0.001% of the contract amount

12 Our findings Highlight the importance of demand shocks for firm growth Winning a government contract increases firm growth by 2.3 percentage points These effects persist for at least 10 quarters Average length of a contract is about a year Winning a government contract affects firm behavior in future auctions Barely winning firms participate in 30 percent more auctions over the next three months 20 percent more (within a 30 day window) a year later 93 percent of the growth in new hires come from firms hiring unemployed or informal workers Growth also comes from less job destruction

13 Contributions Recent models of firm dynamics emphasize heterogeneity in demand factors as an explanation for differential growth paths of heterogenous firms e.g. Fishman and Rob (2005), Dinlersoz and Yorukoglu (2010), Gourio and Rudanko (2011) Drozd and Nosal (2012), Perla (2013) Empirical evidence on the importance of demand for firm growth is limited (e.g. Syverson (2011)) Recent contributions use firm-level price data to separately identify shocks to demand and productivity requires functional-form assumptions need to account for endogeneity of input factors Examine the effects of exporting for firm growth and productivity combines demand shock with product quality upgrading

14 Contributions Demand factors might also explain why firms in developing countries stagnate as they age e.g. Bartelsman, Haltiwanger, and Scarpetta 2013; Hsieh and Klenow 2014 Effects of winning a contract are much more pronounced among younger firms (even conditional on firm size) Barriers to firm growth Access to market e.g. Campbell and Hopenhayn 2005, Combes et al 2012, Holmes and Stevens 2012 Access to credit e.g. Levine 2005, Guiso et al 2004, Manova 2011 Labor regulations e.g. Pierre and Scarpetta 2007, Levy 2010, Hseih and Olken 2014

15 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

16 Public Procurement Auctions in Brazil Brazil s federal government has used reverse auctions as a procurement method for off-the-shelf goods since 2001 As of 2005, it is mandatory for federal agencies to procure off-the-shelf goods through these auctions, and to conduct them online on ComprasNet Around 2200 public bodies throughout Brazil list around 1 million lots every year on ComprasNet In 2012, 0.76 percent of Brazil s GDP or R$ 33.6 billion worth of contracts were awarded through ComprasNet auctions

17 Public Procurement Auctions in Brazil Over 65,000 firms have placed bids in the ComprasNet To participate, firms must first register online as a vendor The process is streamlined and simple so as to encourage participation, especially among small firms Most documents supporting a firm s bid are submitted after winning an auction, which again lowers the cost of participating

18 ComprasNet Data Table 1: 20 Most Frequent Product Categories in ComprasNet: (1) (2) (3) (4) (5) (6) # of Lots % of Total Median Lot Estimate % of Total Value Purchased # of 4-digit # of 6-di Books, Maps, Other Publications 502, Medical & Veterinary Equip 446, , 149 Laboratory Equipment 398, , 009 Office Supplies and Devices 312, , 754 Subsistence (Food) 247, , , 383 IT E&S 238, , , 010 Vehicular Equipment Components 234, Electrical/Electronic Equip Com 151, , 751 Construction & Building Materia 121, , , 283 Total 4,163, e e , 819

19 A typical participant

20 Public Procurement Auctions in Brazil An auction starts with a public body defining lots it needs to procure A lot consists of some indivisible quantity of an off-the-shelf good or service The public body provides a reservation price for each lot Average of at least three quotes obtained through market research The public body advertises the tender at least 8 days before the session and publishes a tender document online

21 Auction Mechanism Two features of ComprasNet auctions are central to our empirical strategy. 1. Within time limits, these auctions end at random 2. These auctions do not contain a proxy bidding system

22 Bidding Timeline Outcry bidding starts Auctioneer announces t2 Start of End Phase Close t0 t1 t2 t3 Random Phase

23 Random auction duration Random Close Auctioneer will announce the start of phase II (random period) A random number is then drawn from U[0, 30] indicating how many minutes is left in the auction Information is unknown to the firms as well as the auctioneer

24 Random auction duration Random Close Auctioneer will announce the start of phase II (random period) A random number is then drawn from U[0, 30] indicating how many minutes is left in the auction Information is unknown to the firms as well as the auctioneer Quasi-random close Auctioneer will announce the start of phase II (random period) A number is drawn from U[5, 30] A number is drawn from U[0, 2] after every bid if the duration of phase II is less than 28 minutes after the 28 th minute the auction ends randomly Overall time is still capped at 30 minutes Information is unknown to the firms as well as the auctioneer

25 Random auction duration Ending Rule Quasi-Hard close II III IV Random-close I 27/Jun/04 15/Apr/06 25/Sep/06 13/Sep/10 16/Dec/10 31/Dec/10 Notes: All changes between 15/Apr/2006 and 16/Dec/2010 were unannounced both to bidders and auctioneers. The period between 15/Apr/2006 and 25/Sep/2006 was similar to the quasi-hard close algorithm described in the paper, but had minor differences. We therefore do not use data from that period.

26 Distribution of durations Period I Period II Density Minutes

27 Absence of a proxy bidding system Proxy bidding (e.g. available in ebay) allows bidders to submit their reservation prices and have the system automatically place new bids on their behalf as soon as they are outbid For ComprasNet, if firms wish to lower their bids, they must enter it manually on the auction page Minimum bid decrement is R$0.01 Throughout the auction, firms (and the auctioneer) only learn the currently low bid, but neither the identity of the firms nor the history of bids

28 Bidding phase

29 Random ending phase

30 Identity of winner disclosed

31 Identity of winner disclosed

32 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

33 Empirical Strategy We are interested in estimating the effects of a government-induced demand shock on firm growth g it = f (X i, U i, S it, ɛ it ) g it - growth rate of firm i in period t X i - firm s observable characteristics U i - firm s unobservable characteristics S it - demand for the firm products, firm sales, or purchased orders ε it - unobserved shock (e.g. changes in TFP)

34 Empirical Strategy Assuming an additive and linear model: ν it = U i + ɛ it S it = G it + P it g it = β 0 + β 1 S it + δx i + ν it ν it = U i + β 1 P it + ɛ it g it = β 0 + β 1 G it + δx i + ν it

35 Potential Sources of Bias g it = β 0 + β 1 G it + δx i + ν it corr(g it, P it ) < 0 If private-sector contracts crowd-out government contracts, then we will under-estimate β 1 corr(g it, ɛ it ) > 0 If government contracts are awarded to the most productive firms, then firm productivity shocks will lead to over-estimate of β 1

36 Research Design We exploit two sources of exogenous variation 1. Random auction duration 2. Close auctions

37 Close auctions Conditional on participation, winning a close auction is as good as random Firms do not know exactly when the auction will end Assumption firms with similar bids prior to the auction end are similar on average in their productive factors Share of close auctions a firm wins in a particular period represents an exogenous demand shock to the firm How do we define close?

38 Defining a Close Auction Table 4: Runnerups and Winners: Sample Balance Runnerups Winners Mean Std. Dev Mean Std. Dev p-value Sample: 251,126 auctions with 2 active bidders in last 30 seconds; bid difference <0.005 Number of Employees in previous quarter Growth rate in previous quarter Growth rate in previous 12 months Average real wages in previous quarter Employees Schooling in previous quarter Accumulated win rate Bidder in same city as Auction Bidder registred as SME

39 Defining a Close Auction Table 5: Runnerups and Winners: Sample Balance Runnerups Winners Mean Std. Dev Mean Std. Dev p-value Sample: 103,648 auctions with 2 active bidders in last 12 seconds; bid difference <0.005 Number of Employees in previous quarter Growth rate in previous quarter Growth rate in previous 12 months Average real wages in previous quarter Employees Schooling in previous quarter Accumulated win rate Bidder in same city as Auction Bidder registred as SME

40 Defining a Close Auction Table 6: Runnerups and Winners: Sample Balance Runnerups Winners Mean Std. Dev Mean Std. Dev p-value Sample: 135,865 auctions with 2 active bidders in last 30 seconds; bid difference <0.001 Number of Employees in previous quarter Growth rate in previous quarter Growth rate in previous 12 months Average real wages in previous quarter Employees Schooling in previous quarter Accumulated win rate Bidder in same city as Auction Bidder registred as SME

41 Close Auctions g it = αd it + η i + δ t + X itγ + ε it g it = E it E it 1 0.5(E it +E it 1 ) D it - total value of government contracts a firm won in the auctions in period t (quarter) X it - time-varying attributes of the firm or participation

42 Close Auctions g it = αd it + η i + δ t + X itγ + ε it g it = E it E it 1 0.5(E it +E it 1 ) D it - total value of government contracts a firm won in the auctions in period t (quarter) X it - time-varying attributes of the firm or participation Instrument: total value of close auctions firm i won in period t divided by the total value of auctions that firm participated in that period

43 Close Auctions g it = αd it + η i + δ t + X itγ + ε it g it = E it E it 1 0.5(E it +E it 1 ) D it - total value of government contracts a firm won in the auctions in period t (quarter) X it - time-varying attributes of the firm or participation Instrument: total value of close auctions firm i won in period t divided by the total value of auctions that firm participated in that period Important Caveat Conditional on participation treatment affects future participation We take care of participation with a selection model

44 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

45 Data 1. Universe of federal procurement auctions from Employer-employee matched data from the universe of formal firms in Brazil from

46 ComprasNet For each lot, we extract from the web pages the following information: reservation price name and tax revenue number of firms participating in the auction time stamps for each auction event purchasing unit running the auction From the Ministry of Planning, Budget, and Management, we collect additional information on lots (e.g. product classification codes) bidders purchasing units

47 ComprasNet Data: Descriptive Statistics Table 2: Sample Descriptive Statistics: Auctions All auctions Close auctions Mean Std. Dev. Mean Std. Dev. Reserve price Winning bid x Ranked2 Ranked1 Ranked Winning bid/reserve Number of Bidders Auction duration Total (minutes) Random phase (minutes) Number of bids Total In random phase In random phase, placed by winner In random phase, placed by runner-up Rank of first bid placed Winner Runner-up Number of outbids in random phase Number of auctions 4,291, ,642 Notes: Table shows summary statistics for auctions held by federal purchasing units between June 2004 and December 2010 in

48 Brazil s employer-employee matched data (RAIS) An administrative data set covering all (formal) firms and workers in Brazil. We have information on wages, education, gender, and age of every employee in all firms for At the end of each year, firms give a monthly breakdown of the status of each of their employees. We construct quarterly measures of firm growth in terms of number of employees.

49 RAIS Data: Descriptive Statistics Table 1: Sample Descriptive Statistics Mean Std. Dev. Number of employees Total Temporary Growth Number of new hires Total From other firms From same municipality Employees Characteristics Average monthly wages Average hourly wages Average years of schooling Wage bill (R$ 1,000) All auctions... Participated Won Winnings (R$ 1,000) Close auctions...

50 Employment size by firm age Employment Share < >=40

51 Firm Size over the life-cycle by regions Employment < >=40 Age Below Median GDP pc Above Median GDP pc

52 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

53 Participation - 30 days

54 Participation - 1 year

55 Win rates - 30 days

56 Win rates - 1 year

57 Amount of Contracts Won - 30 days

58 Amount of Contracts Won - 1 year

59 Amount of Contracts Won - 1 year

60 Future Auction Outcomes We also find that winning firms expand their markets bid for contracts outside of their home municipalities no evidence that winning firms bid for larger lots some evidence winning firms bid for different products (but effects are small)

61 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

62 Event-study analysis, one year

63 Event-study analysis, one year

64 Event-study analysis, two-years

65 Effects of Winning a Government Contract on Firm Growth Dependent variable Firm Growth OLS OLS Reduced-form Reduced-form (1) (2) (3) (4) Won a contract [ ] Value of contracts [ ] % of close auctions won [ ] % Amount won in close auction [ ] R N 949, , , ,252 Mean dep var

66 Effects of Winning a Government Contract on Firm Growth Dependent variable Firm Growth Won a contract Value of Contracts Won IV IV First-Stage First-Stage (1) (2) (3) (4) Won a contract [ ] Value of contracts [ ] % of close auctions won [ ] % Amount won in close auction [ ] R N 949, , , ,562 Mean dep var

67 Effects of Winning a Government Contract: Selection Employment Growth (1) (2) (3) (4) (5) Panel A: Won a contract [ ] [ ] [ ] [ ] [ ] Mill's Ratio [ ] [ ] [ ] Panel B: Value of Contracts won (logs) [ ] [ ] [ ] [ ] [ ] Mill's Ratio [ ] [ ] [ ] Number of Auctions Participated N N 5th 5th 5th Cummulative Win Rate (t-1) Y N N Y Y N 949, , , , ,068 Mean dep var

68 Long-term effects Effects of Winning a Contract Quarters

69 Long-term effects Effects of Winning a Contract Quarters

70 Decomposing the growth effects Growth Layoffs New Hires New Hires Same Municipality From other Not from other Unemployment firms firms First Job (1) (2) (3) (4) (5) (6) (7) (8) Won a contract [0.0021] [0.0138] [0.0033] [0.0011] [0.0028] [0.0020] [0.0013] [0.0016] N 935, , , , , , , ,051

71 Local Spillovers? Growth Wages OLS IV OLS IV (1) (2) (3) (4) Total Amount Won (logs) [ ] [ ] [ ] [ ] R N 46,725 46,725 48,060 48,060 Mean dep var Sample of non-participating firms within the same municipality

72 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

73 Effects of winning by age of firm Treatment Effects < >=40 Age

74 Effects of Winning on Firm Growth by Firm Characteristics Employment Growth Manufacturing Non-Manufacturing SME Non SME (1) (2) (3) (4) Winning a contract [ ] [ ] [ ] [ ] N 169, , , ,064 Mean dep var

75 Outline 1. Background on the program 2. Empirical Strategy 3. Data 4. Effects of winning procurement contracts on future contracts 5. Effects of winning procurement contracts on firm growth dynamic effects formal sector employment 6. Effects of winning procurement contracts by sector firm size age 7. Exploring barriers to firm growth financial constraints labor regulations market access 8. Concluding remarks and future work

76 Barriers to Firm Growth Literature has proposed various potential barriers to firm growth Credit constraints - complement or substitute? Firms might need liquidity in order to respond to this positive demand shocks Government demand shocks might allow firms to use a governmental contract as collateral so it may bootstrap their way into more growth Remoteness Literature has documented a positive relationship between firm size and market size marketplace may break this link depending on the type of good Firms who are more distant from larger markets may benefit relatively more from winning a governmental contract Labor Regulations Labor regulations are typically seen as an impediment to firm growth

77 Finance Municipal Characteristic: Below the median Above the median Below the median Above the median Age of firm Less 10 years At least 10 years Less 10 years At least 10 years (1) (2) (3) (4) (5) (6) Panel A: Access to Credit Won a contract [ ] [ ] [ ] [ ] [ ] [ ] N 540, , , , , ,386 Mean dep var

78 Labor regulation Municipal Characteristic: Below the median Above the median Below the median Above the median Age of firm Less 10 years At least 10 years Less 10 years At least 10 years (1) (2) (3) (4) (5) (6) Regulations Won a contract [ ] [ ] [ ] [ ] [ ] [ ]

79 Remoteness Municipal Characteristic: Below the median Above the median Below the median Above the median Age of firm Less 10 years At least 10 years Less 10 years At least 10 years (1) (2) (3) (4) (5) (6) Panel C: Access to market Distance to Municipality with population <50,000 Won a contract [ ] [ ] [ ] [ ] [ ] [ ] Distance to Municipality with population <100,000 Won a contract [ ] [ ] [ ] [ ] [ ] [ ] Distance to Municipality with population <300,000 Won a contract [ ] [ ] [ ] [ ] [ ] [ ]

80 Conclusions Governments throughout the world are significant purchasers of goods and services We exploit close auctions to isolate the causal effects of winning government contracts on firm growth We find that winning a government contract does lead to firm growth, and these effects persist beyond the length of the contract. The effects are more pronounced for younger firms and for firms located in municipalities with high access to credit and more relaxed labor regulations. Increases in firm size come from a combination of employment creation and formalization of workers

81 Management upgrading Dep variable: Has a Manager (0/1) (1) (2) (3) (4) (5) Won contract [ ] [ ] [ ] [ ]** [ ]** Sample All Firms Firm size 1st Tercile Firm size 2nd Tercile Firm size 3rd Tercile Firms that ever had a manager Observations 948, , , ,265 70,365 Mean dep var

82 Management upgrading Dep variable: Quality of manager Age Years of Schooling Tenure Wage (6) (7) (8) (9) Won contract [ ]** [ ] [ ]** [ ]** Sample Conditional on having a manager Observations 32,944 32,735 32,944 32,758 Mean dep var

83 Manipulation Test