Efficiency of Italian opera houses: A stochastic frontier production function approach Sabrina Auci –University of Palermo Antonio Cognata – University of Palermo 18th International Conference on Cultural Economics, June, 2014, Montréal, Canada
Building blocks Literature Italian opera sector Data description Empirical model Estimation results Conclusions 2
Literature (1) Opera is the ‘queen’ of performing arts for the costliness of its production for the share of public subsidies Understanding the production process, measuring efficiency and identifying possible sources of inefficiency is important, especially in times of cost- pressure on public budgets 3
Literature (2) Cost disease: no or few productivity improvements No profit maximization Constrained output maximization: quality is important There is asymmetric information between opera house managers and Government on production costs Managers seek large budgets through subsidies and do not follow cost-minimizing behavior 4
Literature (3) Performing Arts Models Author(s)YearFocusArtformOutputFindings Globerman & Book 1974CostOrchestra & TheatrePerformancesScale economies Gapinski1980ProductionTheatre, Opera, Orchestra, Ballet AttendancesDeclining inputs marginal products Lange & Luksetich1993CostOrchestraAttendancesScale diseconomies Felton1994Cost & DemandOrchestraAttendances & Performances Cost disease Taalas1997CostTheatreAttendancesScale economies & allocative inefficiencies Fazioli & Filippini1997CostTheatrePerformancesScale & scope economies Marco-Serrano2006ProductionTheatreAttendances & Performances Managerial inefficiencies Last & Wetzel2011Production & CostTheatreSupplied TicketsNo cost- minimizing behavior Zieba2011ProductionTheatre, Orchestra, Opera, Musical, Ballet Attendances & Supplied Tickets Public funds have positive effects on quality 5
Italian Opera Sector (1) Foundations: 13 Opera Houses 1 Symphony Orchestra 1Turin (Regio) 2Genoa (Carlo Felice) 3Milan (La Scala) 4Verona (Arena) 5Venice (Fenice) 6Trieste (Verdi) 7Bologna (Comunale) 8Florence (Maggio) 9Rome (Opera) 10Naples (S. Carlo) 11Bari (Petruzzelli) 12Palermo (Massimo) 13Cagliari (Lirico) 14Rome (S. Cecilia Orch) 6
Italian Opera Sector (2) Heavily dependent on public funding Of total revenue (average ) 66.7%public funds 8.4%private fund 15.7%box office 9.2%other sources of revenue Very high uncertainty on the amount of subsidies each year 7
Italian Opera Sector (3) Very strong local-politics influence: The Mayor is the President of the Board of the Opera House and chooses the management Low productivity: 63 opera performances per house per year (average ) Very high level of debt 8
Italian Opera Sector (4) International Comparison N˚of Performances – 2009 Staatsoper (Wien)241 MET Opera (New York)225 Opernhaus (Zurich)191 Opéra (Paris)176 Royal Opera House (London)151 Bayerische Staatoper (Munich) 126 Teatro Real (Madrid)120 La Scala (Milan) 96 9
Italian Opera Sector (5) Total Debts (€ Real Values) 10
Italian Opera Sector (6) Subsidies to Italian Opera Houses – 2012 €411 mln Government budget for the performing arts €193 mln (47%) share for the major opera houses €330 mln total of all public funds to major opera houses (including regions, counties and cities) 11
Data set: 12 major italian opera houses From 2001 to 2010 Financial and accounting data from annual reports Box office figures from EDT/CIDIM (Annual Opera Report) Political Influence: Mayor party – left/right Superintendant professional background: manager/artist/other Quality: ‘Oscar’ from the Italian Musical Critics Association Geographical heterogeneity: North/South of Italy 12
Empirical model (1) Very unlikely two different opera houses produce a similar output with the same cost structure The difference among opera houses is best explained through efficiency analysis To understand the relationship between output and input we estimate a production function using the stochastic frontier approach (SFA) 13
Empirical model (2) The methodology allows to distinguish between production inputs and efficiency/inefficiency factors, to estimate distances from the efficient frontier, differentiating between systematic component and noise error The idea is that the maximum output frontier for a given set of inputs is assumed to be stochastic An additional error term is introduced to represent technical inefficiency 14
Empirical model (3) The Battese and Coelli (1995) specification is a Stochastic Frontier model in which specific opera houses effects are assumed to be distributed as truncated normal random variables where the unobserved random noise is divided in two components: first, v it, a set of random variables with normally distributed error terms, and … (1) 15
Empirical model (4) … second, an independent component, u it, a set of non- negative random variables, capturing the effects of technical inefficiency, with a zero-truncated normal distribution (only positive values) (2) 16
Model specification (1) We estimate a log-linear Cobb-Douglas production function with the following specification: 17
Model specification (2) VarDefinition Y1Y1 Performances Y2Y2 Ticket supplied (Ticket on offer) K Non-personnel Expenses (Operating expenses) L_totPersonnel Cost L_artGuest Artist Cost d_yearTime Dummies Production function variables (outputs/inputs) 18
Model specification (3) Two alternative definitions of Artistic Output are considered: Ticket supplied (or Ticket on offer): which measures the “produced output” in contrast to sold output (Throsby, 1994; Last and Wetzel, 2010) Performances: number of all staged performances (Globerman and Book, 1974; Lange et al. 1985) which however in Italy is constrained by the “stagione” system of production Since our focus is on production, we do not consider Theatre Attendance (or Tickets Sold) that instead refers to the demand side, capturing the “cultural experience” of visitors (Gapinski, 1984; Taalas, 1997) 19
Model specification (4) … and the following error function (inefficiency term) as: 20
Model specification (5) VarDefinition PubFundShare of Public Subsidies on Total Revenue CostPTotal Production Cost on Total Revenue n_opera Total number of operas performed in a “stagione” d_polPolitical Dummy (1=left and 0=right) d_prize Prize (‘oscar’) Dummy (0=no prize; 1= one prize; 2= two prizes …) d_sovrin Superintendent (background) Dummy (0=manager 1=other 2=artist) d_NordGeographycal Dummy: 0=center-south and 1=north Inefficiency function variables 21
Output descriptive analysis (1) Tickets supplied 22
Output descriptive analysis (2) Performances 23
Inputs descriptive analysis (1) Personnel Costs: Employees & Guest Artists 24
Inputs descriptive analysis (2) Capital: Operating Costs 25
Summary Results (1) Ticket Supplied Performances expected sign actual sign expected sign actual sign K L_tot L_art PubFund CostP n_opera d_pol d_prize d_sovrint d_nord
Summary Results (2) Note: * significant at 10%; ** significant at 5%; *** significant at 1% Ticket suppliedPerformances coeff.t-ratiocoeff.t-ratio Constant *** K0.46 *** L_tot1.14 *** L_art0.53 * *** 3.58 Constant *** PubFund1.88 *** ** 1.97 CostP3.41 *** n_opera-0.07 *** *** d_pol *** 2.92 d_prize0.17 *** d_sovrint-0.19 *** d_nord sigma-squared0.17 *** *** 4.76 Gamma1.00 *** log likelihood function Obs117 27
Summary Results (3) The input coefficients of the Cobb-Douglas production function perform relatively well for all output specifications We expect that public subsidies would increase inefficiency – since they prompt higher input requirements. Actually, in our results the share of public subsidy on total revenue has the expected effect As expected the ratio of Total Production Costs on Total Revenue has a positive effect on output: Italian opera houses clearly do not follow a cost-minimizing behavior and are rather knee to break budget constraints 28
Total production costs on total revenue 29
Summary Results (4) Signs for n_opera are negative: an increase in the number of performed operas improves efficiency (increases output) The local-political dummy has always a positive sign: going from right to left of the political spectrum seems to reduce output and to increase inefficiency 30
Summary Results (5) The prize (‘oscar’) dummy respects the expected sign on both output definitions confirming the well known trade-off between quantity and quality of production The Superintendent (General Manager) dummy has a mixed effect. When considering tickets supplied the sign is negative and significant and seems that managers are not better than artists in running an Opera house (?!). However, when considering the number of performances, the sign turns out to be positive (but not significant) 31
Technical Efficiency Ranking (1) TICKET on OFFER SFA modelDEA model with variable return of scale id_teatrimeanid_teatrimean 1VERONA0.741VERONA1.00 2TRIESTE0.312TRIESTE0.96 3GENOVA0.283GENOVA0.95 4CAGLIARI0.2110CAGLIARI0.89 5BOLOGNA0.185BOLOGNA0.91 6TORINO0.1811TORINO0.88 7NAPOLI0.179NAPOLI0.90 8FIRENZE0.136FIRENZE0.91 9PALERMO0.137PALERMO VENEZIA0.1112VENEZIA ROMA0.114ROMA MILANO0.048MILANO
Technical Efficiency Ranking (2) PERFORMANCES SFA modelDEA model with variable return of scale id_teatrimeanid_teatrimean 1MILANO0.975MILANO0.94 2ROMA0.962ROMA0.96 3TORINO0.941TORINO0.98 4TRIESTE0.933TRIESTE0.95 5VERONA0.906VERONA0.93 6BOLOGNA0.884BOLOGNA0.95 7PALERMO0.837PALERMO0.92 8VENEZIA0.838VENEZIA0.89 9FIRENZE0.6910FIRENZE GENOVA0.659GENOVA CAGLIARI0.5911CAGLIARI NAPOLI0.5812NAPOLI
Technical Efficiency Ranking (3) The technical efficiency scores of SFA and DEA show similar rankings among opera houses As expected, Verona is the best opera house in term of Tickets on offer, while Milano is the best when output is measured by performances 34
Conclusions Our analysis confirms the general findings that performing arts have space for improvements in the use of resources; and this is expecially true in the heavy subsidized Italian opera sector Improvements may be realized through an incentive-driven funding system that favors cost-minimizing behavior without reducing quality 35
Thank you very much for your attention Suggestions are welcome 36