Why some restaurants charge couvert?
Evidence from Czech restaurants
Michal Kvasniˇcka Zuzana Kalm´arov´a
MUES, April 23, 2014
1 / 28
Content
• goal of the research
• literature review
• description of the data
• results of survey
• stylized facts
• theoretical model
• econometric model
• conclusion and discussion
2 / 28
What couvert is
Couvert is a fixed fee charged by some restaurants in addition to the
food price when a customers orders a main dish.
Also known as
• couvert (France, Czech Republic, Slovakia, English countries)
• pane e coperto (Italy)
• cubierto (Spanish speaking countries)
• table charge, cover charge (English speaking countries)
3 / 28
Goal of the research
Goal of the research: find out why some restaurants charge it while
others do not.
Hypotheses (non-exclusive):
• a cost-based practice: a lump-sum payment for the place setting,
bread, sauces, etc.
• a monopolistic practice: two-part tariff
• a way to select customers when the physical capacity of the
restaurant is lower than optimal
4 / 28
Previous literature
Unknown. Only entries in Wikipedia and tourists’ guides.
Closest are papers on restaurants rationing by Becker (1991),
Ungern-Sternberg (1991), and Bose (1996).
Literature on two-part tariff, originating from Oi (1971).
5 / 28
Data
Data gathered in 2011 in Brno restaurants.
A restaurant defined as an enterprise whose major business is to serve
hot food to customers which are provided with seats and are attended
by waiters.
Restaurant selection:
1 restaurants listed on e-restaurace.cz (1 140)
2 sub-sample of 284 restaurants that
• had their web sites
• had menu available on their web sites
• were listed in at least three other registers
3 stratified sample of 70 restaurants (15 with couvert, 55 without)
4 final sample of 66 restaurants (4 refused to cooperate)
5 sample of 64 restaurants for econometrics (2 leverage points)
6 / 28
Data (cont.)
Survey and hard data gathering.
Hard data:
• couvert in CZK
• restaurant’s average price in CZK
• extra payment for additional bread or sauce in CZK
• pizzeria (dummy)
• foreigners (dummy)
• capacity (number of seats inside)
• customers (number served daily)
• reservations (share of days when reservation needed)
7 / 28
Results of the survey (1)
Restaurants’ definition of couvert:
• fee for extra bread, sauces, dressings, and table setting
• extra lump-sum fee for food
Why they charge couvert:
• to cover the cost of extra bread, sauces, etc.
• it is a usual practice; should be charged
8 / 28
Results of the survey (2)
What restaurants provide in exchange for couvert:
• ketchup (90 % agreed)
• bread (82 %)
• dressing (80 %)
• serviettes (78 %)
• table setting (71 %)
• mustard (50 %)
Other motives to charge it:
• customers order more food if charged couvert (43 %)
• it is usual practice in the Czech republic (36 %)
• it discourages undesirable customers (13 %)
9 / 28
Results of the survey (3)
Couvert is charged for “every” main dish with exception of lunch
specials.
Lunch specials are exempt because
• less service is provided, extra bread etc. is charged separately
• highly elastic demand
• other restaurant do not charge for lunch specials
Many exceptions:
• when food shared by the customers
• when the food is local cheap food
• when the customers are perceived not to be willing to pay it
(for soup, salad; constitutional court judges)
10 / 28
Results of the survey (4)
How long restaurants charge couvert:
• since their inception (most of them)
• started two or three years after their inception (2)
Few restaurants stopped charging couvert because their owners felt
guilty and for other uncertain reasons.
Restaurants that do not charge couvert say:
• all costs are covered by the price of the food
• they charge extra fee for an extra portion of bread etc.
• customers do not understand what couvert is and get angry when
charged it
11 / 28
Data on couvert: frequency
selection sample
charging couvert 59 21 % 25 38 %
total 284 66
Restaurants often hide couvert both on their web sites and in their
menus.
12 / 28
Data on couvert: size
size of couvert
frequency
0 5 10 15 20 25
010203040
Revenue from couvert can finance one waiter in a median restaurant.
13 / 28
Data on couvert: variables
all w. couvert w./o. couvert
mean st. dev. mean st. dev. mean st. dev.
price 170.59 53.38 189.16 53.23 159.27 50.83
sauce 8.94 10.27 6.69 11.07 10.31 9.63
bread 4.35 6.10 3.76 5.67 4.71 6.39
pizzeria 0.17 0.38 0.24 0.44 0.12 0.33
capacity 87.27 59.22 93.00 79.14 83.78 43.68
customers 151.97 110.08 164.00 119.05 144.63 105.07
foreigners 0.21 0.41 0.20 0.41 0.22 0.42
reservations 0.31 0.31 0.41 0.32 0.25 0.29
Restaurant charging couvert are more expensive, charge less / less
often for extra bread and sauces, their customers more often need
reservations, and are more often pizzerias.
14 / 28
Data on couvert: correlations
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
couvert
price
sauce
pastry
pizzeria
capacity
customers
foreigners
waiters
reservations
k1
k2
couvert
price
sauce
pastry
pizzeria
capacity
customers
foreigners
waiters
reservations
k1
k2
15 / 28
Model: setup
Chamberlin’s monopolistic industry.
Total cost: C = cQ + rN
Individual inverse demand: p = θ − bq, θ ∼ U(0, ¯θ]
Assumptions:
• b > 0 same for all customers in all restaurants
• c = a(¯θ) : 0 < a(¯θ) < ¯θ, 0 < da(¯θ)/d ¯θ < 1
• r ≥ 0 same for each restaurant of a particular type, independent
from ¯θ; may be lowered if its part is charged separately
Restaurant’s profit:
π(x, p, A) =
¯θ
x
[(p − c)(θ − p)/b + (A − r)]dθ (1)
16 / 28
Model: Oi’s algorithm
π(x, p, A) =
¯θ
x
[(p − c)(θ − p)/b + (A − r)]dθ (1)
Optimization with unconstrained capacity is solved in three steps:
1 calculate optimal A for given p and x
2 plug the optimal A into (1) and calculate optimal p for a given x
3 plug the optimal A and p into (1), calculate optimal x, and then
calculate uncodintional optimal A and p
Constrained optimization is solved in two steps (x = ¯θ − k is fixed):
1 calculate optimal A for given p and x
2 plug the optimal A into (1) and calculate unconditional optimal p
and A
17 / 28
Model: optimal two-part tariff with sufficient capacity
Set A and p to maximize (1) s.t. CS(x) ≥ 0.
Proposition 1: The optimal two-part tariff has the following
properties:
1 A ≥ r and rises with r if it is optimal for the restaurant to
operate; A > 0 if r = 0, A > r if the restaurant operates with a
positive profit.
2 dA∗
/d ¯θ > 0, dp∗
/d ¯θ > 0
3 A∗
= f (p∗
, r) where df (p∗
, r)/dp∗
> 0
4 A∗
decreases for any given p∗
when r decreases
18 / 28
Model: optimal two-part tariff with insufficient capacity
Set A and p to maximize (1) s.t. x = ¯θ − k where k < N∗
= ¯θ − x∗
.
Proposition 2: The optimal two-part tariff with insufficient capacity
has the following properties:
1 A and p are independent from r
2 dA /d ¯θ > 0, dp /d ¯θ > 0
3 A = g(p , r) where dg(p , r)/dp > 0
4 A decreases for any given p when k increases
19 / 28
Model: cost-based couvert
Set p to maximize (1) s.t. A◦
= r and CS(x) ≥ 0
(sub-optimal two-part tariff).
Proposition 3: The optimal two-part tariff with insufficient capacity
has the following properties:
1 there is no relationship between A◦
and p◦
2 A◦
decreases for any given p◦
when r decreases
20 / 28
Identification strategy
Testable hypotheses:
1 cost-based couvert: decreases when the restaurant charges for
any place-setting item (which lowers r); does not depend on the
restaurant’s price or its capacity; is higher for the types of
restaurants which have higher r
2 monopolistic couvert not used to select customers: decreases
when the restaurants charges for any place-setting items;
increases in the restaurant’s price; does not depend on the
capacity; can be higher for types of restaurants with higher r
3 monopolistic couvert that is used to select customers: increases
in the restaurant’s price and its utilization rate (decreases in its
physical capacity); does not change when the restaurants charges
for any place-setting items and should be the same for all types
of restaurants (i.e. independent from r)
21 / 28
Identification strategy (cont.)
Hypotheses:
• H1 = cost-based couvert
• H2 = monopolistic two-part tariff but not selection tool
• H3 = monopolistic two-part tariff used as selection tool
Couvert
• decreases with decrease in r (pizzeria, bread, sauces)
⇒ H1 or H2, not H3
• rises with increase in price
⇒ H2 or H3, not H1
• depends on restaurant’s capacity
⇒ H3, not H1 or H2
22 / 28
Identification strategy: econometric models
Three possible ways to check the theory:
• whether couvert is charged or not
• LPM
• probit
• the size of the couvert
• tobit
The models are not that different:
• the latent variable is the same
• the models’ parameters in roughly fixed ratios
23 / 28
Identification strategy: the latent variable
Latent index ≡ the optimal value of couvert − a cost of charging it.
binary couvert =
1, if the latent index > 0,
0, if the latent index ≤ 0.
size of couvert =
the latent index, if the latent index > 0,
0, if the latent index ≤ 0.
(This part is missing in the theoretical model.
How should I incorporate it into the model?)
24 / 28
Identification strategy: definition of the capacity
Two empirical definitions of the capacity:
1 capacity = (1 − reservations)
2 capacity = (1 − reservations) × number of seats
The difference in r tested
• type of the restaurant: pizzeria dummy
• lowered when extra charge for extra bread (not for sauces—it was
never statistically significant and deteriorated the fit; perhaps
because of its high correlation with bread)
• no test whether the impact of bread depends on the capacity
(never statistically significant; perhaps too few observations to fit
interactions)
25 / 28
Econometric model: estimates
LPM1 LPM2 probit1 probit2 tobit1 tobit2
price 0.005∗∗∗
0.007∗∗∗
0.014∗∗∗
0.030∗∗∗
0.151∗∗∗
0.279∗∗∗
(0.001) (0.002) (0.005) (0.008) (0.049) (0.070)
bread −0.016∗
−0.021∗∗
−0.047 −0.076∗∗
−0.423 −0.638∗∗
(0.010) (0.009) (0.030) (0.033) (0.318) (0.308)
free cap. −0.003∗
−0.003∗∗
−0.010∗
−0.012∗∗
−0.097∗
−0.107∗∗
(0.001) (0.001) (0.005) (0.006) (0.055) (0.051)
foreign. −0.346∗∗
−1.650∗∗
−14.246∗∗
(0.151) (0.666) (6.009)
pizzeria 0.340∗∗
1.323∗∗
12.245∗∗∗
(0.145) (0.528) (4.578)
const. −0.139 −0.517∗
−1.967∗∗
−4.343∗∗∗
−20.363∗∗
−39.617∗∗∗
(0.259) (0.271) (0.878) (1.254) (9.185) (11.732)
26 / 28
Econometric model: relative importance of the hypotheses
Estimated by hierarchical partitioning on binary models:
model price pastry free capacity foreigners pizzeria
LPM2 53.33 8.94 15.22 8.29 14.22
probit2 45.96 10.62 17.30 11.38 14.73
Grouping:
• monopolistic: price, free capacity
• cost-based: bread, pizzeria
• uncertain: foreigners
model monopolistic cost-based uncertain
LPM2 68.56 23.16 8.29
probit2 63.26 25.36 11.38
27 / 28
Conclusion and discussion
Restaurants explain couvert as a cost-based practice.
Our econometric study provides an evidence that it is a monopolistic
two-part tariff and a selection mechanism too (perhaps more efficient
than queuing by Bose, 1996).
Other possible explanation: couvert is a hidden fee, i.e. “robbing”
(but less often charged to tourists).
28 / 28