Assessing Costs of Using Local Foods in Independent Restaurants

Abstract

The purpose of this article was to assess process and production costs that a convenience sample of restaurants in the Midwestern region of the United States incurred using locally produced foods versus sourcing food products through national suppliers. Using a novel application of data envelopment analysis, results showed that surveyed restaurants could improve their time with the delivery process of local foods. While there was no statistically significant difference in production efficiencies measured by preparation time of menu items using local versus nonlocal ingredients, scale and establishment effects existed. This study fills a gap in foodservice literature by developing a unique analytical framework to assess multistage production cost.

KEYWORDS:

• restaurant
• local foods
• process-production costs

The phenomenon of eating locally grown food has recently been on the increase in the United States and is expected to continue increasing as consumers become more health conscious and environmentally aware (Center for Sustainable Systems, 2006; National Restaurant Association, 2007). While many restaurants are interested in using local suppliers, they are concerned about higher costs associated with purchasing locally (Strohbehn & Gregoire, 2002). The literature analyzing production efficiencies of foodservices, especially restaurants, remains limited. This article assesses successive process and production stage time and monetary costs of using local ingredients to prepare menu dishes, and compares these with costs of similar dishes prepared using ingredients purchased through nonlocal suppliers. This article is timely as many restaurants consider using locally grown foods as a differentiation strategy.

LITERATURE REVIEW

Local Foods

The concept of local foods is pluralistic in its definition (Holt & Amilien, 2007). In addition to referring to a geographic origin, it can also characterize the food by cultural subjectivity and social and political environment. For the purpose of this study, local food is defined as food grown or processed locally and purchased by restaurants from the local market or primarily through local producers. Despite an increasing use of local vendors by restaurants to purchase locally grown and produced foods, there is limited research in understanding and comparing cost and efficiency aspects of using locally grown and produced food from local food vendors versus national vendors in the context of purchasing, receiving, storage, production, and customer service in restaurants. Improved understanding of process costs could aid restaurants to continually increase production efficiencies with better vendor selection policies. Above all, restaurants will be able to make knowledgeable decisions about using local foods as a differentiation strategy (Gregoire & Strohbehn, 2002). Given that restaurants are highly sensitive to consumer preferences (consumer driven) and can emphasize convenience in preparation, delivery, and services from suppliers, knowledge of additional process and production costs could impact product and service quality and profitability (Capps & Park, 2003).

While there is no systematic research to assess whether local farmers prefer selling to small or national chain restaurants, anecdotal evidence suggests that local farmers prefer to use locally owned smaller restaurant establishments as another marketing source to promote their products (Sharma, 2006b). Small independently owned restaurant establishments remain an under-studied aspect of foodservice and hospitality literature. A primary reason for this literature gap is the challenges of collecting reliable data (Sharma, 2006a). Therefore, recent literature focusing on small restaurants is limited. Kwansa (1994) presented a synopsis of financing opportunities for small restaurants and hotels, emphasizing that banks and personal resources remain the leading methods of financing for such businesses. Highlighting the lack of a marketing plan focused specifically on restaurant entrepreneurs, Knychalska and Shaw (2002) demonstrated implementation of such a development framework. Thompson (2002) investigated the practicality of dedicating party-size tables in walk-in restaurants to avoid process delays of configuring large party tables. Even though systematic research is lacking, anecdotal evidence points to the fact that small restaurants can be an important source of entrepreneurial activity, employment generation, and economic linkages. From these perspectives, this article adds not only to small independently owned restaurant literature, but also fills a critical gap of how such restaurants can extend a critical economic role in the local economy through the purchase of local grown and produced foods.

Cost and Production Analysis in Restaurants

The literature on cost and profit analysis of restaurants and foodservice operations has continued to evolve since its early years. Various aspects of foodservice cost analysis research exist in the literature: relationship of food expenditure with energy prices (Arbel, 1983), indirect costs and menu pricing (Pavesic, 1990), general revenue and cost analysis (Malk & Schmidgal, 1995), principles of restaurant management (Smith, 1996), profit measurement of menu items (Chan & Au, 1998; Sullivan, 2003), costs associated with Hazard Analysis and Critical Control Points (HACCP; Cohen, Cukierman, & Schwartz, 2000), management of energy costs (Stipanuk, 2001), and more recently, activity-based costing (Raab & Mayer, 2003, 2007). Still, there is a lack of studies that analyze process and production cost elements at the individual menu item level across different restaurant establishments. Recent methodological developments such as the nonparametric Data Envelopment Analysis (DEA) present novel and interesting ways to analyze disaggregate and restricted data sets.

Production efficiency analysis literature using nonparametric techniques such as Data Envelopment Analysis (DEA) have recently been used among researchers (Cooper, Seiford, & Zhu, 2004). DEA uses a relative measure of efficiency of a unit compared to others in the group measured by the ratio of weighted sum of outputs to the weighted sum of inputs (Farrell, 1957; Farrell & Fieldhouse, 1962). Different sets of weights are applied to each unit based on their respective inputs and outputs. These sets of weights for each unit are determined by comparing the unit to other most efficient units in the group (Charnes, Cooper, & Rhodes, 1978). While there are certain limitations to the use of DEA, such as the results are sample specific, it still provides a powerful tool to analyze small sample data. Small business financial information is not easy to collect. From that perspective, too, this tool provides analytic flexibility to incorporate various inputs and outputs into the empirical models. As a consequence, its use in the analysis of hotel productivity analysis is gaining recognition. Parkan (1996) conducted a case study analysis to measure the operating performance of a hotel using 13-month financial data using DEA. This study compared DEA with other types of financial ratio analysis and found that DEA was able to provide a better understanding of operating inefficiencies. Hwang and Chang (2003) used DEA to evaluate managerial efficiency of 45 international tourist hotels in Taiwan. The results of this study showed that efficiency of international hotels in the country was being impacted by the internationalization of hotels. They also found a link between customer profile and management style of hotels with the management efficiency of these properties. Barros (2005) analyzed production efficiency of the Portuguese state-owned hotel chain company using DEA. The analysis was able to identify inefficient hotels versus the efficient ones and also slacks in inputs and outputs of the inefficient hotels to identify critical intervention points. FeiChing, WeiTing, and JuiKou (2006) used DEA to evaluate the cost efficiency of international tourist hotels in Taiwan. Their results showed a link between scale effects and technical inefficiencies, implying that scale of operations in the international hotels was too small to enable cost-savings associated with larger operations. Even though the use of DEA is gaining popularity among hospitality researchers, to the best of the authors' knowledge, this methodology has not been used in foodservice settings, especially to assess efficiency at the individual menu item level.

Further, there is a gap in the literature to assess process and production costs of menu items to evaluate their production efficiencies. In the context of using locally grown and produced food, restaurants and other foodservice establishments have suggested that these costs may be higher. If so, the use of local foods in menu item preparation may not be a self-sustaining strategy without other changes in practice. For this reason, this study is timely and fills a critical gap in foodservice process and production cost analysis literature. In view of this study's purpose, the following research questions were posed:

1. Is there a difference in time spent by independent local restaurants on sourcing and receiving locally grown and produced foods from local vendors versus those purchased through nonlocal vendors?

2. Is there a difference in restaurants' food costs for such locally sourced foods versus those purchased through nonlocal vendors by independent local restaurants?

3. Is it more efficient for local restaurants to produce menu items using local foods versus those prepared using nonlocal foods?

METHODS

Participants

As stated earlier, this study defined local food as food grown or processed locally and purchased by restaurants from the local market or primarily through local producers. This definition is somewhat restrictive as it refers only to the geographic aspects of food. At the same time, this definition is relatively broad as it includes locally farmed, grown, and processed foods within a 50-mile radius of the participating restaurants. Distance has been used previously as a definition variable to identify food by its local origins (Amilien, Holt, Montagne, & Téchoueyres, 2007). However, the actual magnitude of this distance is relatively subjective and context specific. After a preliminary review of purchasing practices of a selected sample of restaurant participants, researchers decided that defining food bought within a 50-mile radius of the establishment would be appropriate for this study. In addition to this criterion, management was also asked to confirm that the selected menu items for this study were promoted either verbally by waitstaff, on the menu within brief descriptions of the respective dishes, and/or on signage posted within the restaurant, as dishes using local ingredients. This criterion was used based on the assumption that when prompted, customers would consider purchasing the local dishes thereby increasing their sales (Sharma, 2006b).

Data for the study were gathered through face-to-face structured interviews with 10 restaurateurs (representing small and independently owned establishments), owners, or general managers/head chefs in a Midwestern state in the United States. All 10 respondents were the chief decision makers for each of the restaurant operations. The restaurants were selected to maximize the diversity of the establishments by the type of menu offering, size, location, type of preparation-service mix, and affiliation. Attempts were made to represent restaurants that offered American-style menus, more elaborate European-style cooking, and ethnic menu variety, such as Mexican-style menus. The participating restaurants also included a mix of white tablecloth, diner style, casual family seating, upscale café, and quick-food service restaurants. Size also varied across the restaurants with a maximum seating of 250 and a low of 85. Average seating across all 10 restaurants was 136 seats with a standard deviation of 47 (34%). These restaurants also represented varied location, from rural small town to urban city and suburban townships. While eight restaurants were privately owned sole proprietorships, one was a privately owned partnership and another a franchise but management owned. Three of the 10 restaurants were part of a larger establishment such as a hotel, a golf-club, or a privately owned clubhouse.

Data and Measures

Six menu items at each restaurant, representing starters, main dishes, side orders, and desserts were identified. Of these dishes, three were identified that used locally grown and produced primary ingredients and the other three nonlocal primary ingredients. The definition of primary ingredient was identified by the chef/management for each selected menu item. The menu item production stages were defined as processes (sourcing, purchasing, production, and service). While there is no prior literature in foodservice, value chain analysis proposed by Porter (1985) follows a similar approach. Successive tasks at each process stage were then identified. Resources required for each task were measured as time costs and total item cost for the primary ingredient; these were compared between the two sets of menu item data representing local versus nonlocal purchasing practices using the DEA approach. For instance, sourcing and purchasing costs were measured as the time required for completing each of these processes for the primary ingredient of the menu item. The total food cost of the item was calculated by multiplying the food cost percentage of each item and its retail price. Other variables that were used in the analysis included portion size of the dish and the weight of the primary ingredient. Item contribution margin was calculated by taking the difference between published retail price of the menu item and total cost of the item.

Data Analysis

DEA is a nonparametric, linear programming-based technique to measure the relative efficiency of units in the presence of multiple inputs and outputs. Typically, the ratio of an output to a corresponding input can be used to measure production efficiency. However, such a technique is inadequate if multiple inputs and outputs are present. Farrell (1957) and Farrell and Fieldhouse (1962) provided a solution to this problem by proposing a relative measure of efficiency of a unit compared to others in the group as measured by the ratio of weighted sum of outputs to the weighted sum of inputs. This raised the obvious question of whether or not common weights should be applied across all units, given that each unit may value its inputs and outputs differently. Charnes, Cooper, and Rhodes (1978) suggested that each unit should be allowed to apply different sets of weights to their respective inputs and outputs. These sets of weights would be determined by comparing the unit to other most efficient units in the group. The fractional model of such weighted inputs and outputs can be presented as follows:

such that,

Here, h ₀ is the production efficiency of unit j in context of y outputs and x inputs. The outputs are assigned weights u and the inputs are assigned weights v. h ₀ for each unit is less than or equal to 1 for unit j. If it is equal to 1 then the unit is said to be efficient; otherwise, if h ₀ is less than 1 then it is inefficient. This efficiency or inefficiency can be presented as a percentage relative to a 100% or fully efficient. The unit itself is labeled within the DEA approach as a decision-making unit (DMU) signifying that management is able to make a production decision to vary inputs and outputs of that unit. The efficiency scores of each DMU are calculated relative to other DMUs in the group that are fully efficient. These fully efficient DMUs are said to be on the efficiency frontier. Among the important properties of DEA efficiency models are that these can be either input oriented or output oriented. Further, such models can assume constant returns to scale (CRS) or variable returns to scale (VRS). Input-oriented models maximize efficiency of the DMU based on minimizing the inputs used for a given level of output. Similarly, output-oriented models maximize efficiency based on maximizing the outputs given a certain level of inputs. In constant returns to scale (CRS) models, input changes lead to a proportional change in outputs. However, variable returns to scale (VRS) models relax these assumptions so that scale effects can be assessed. That is, if VRS models result in increasing a DMU's efficiency higher than the corresponding CRS models, a scale effect is said to exist, implying that given the same level of inputs, the DMU could increase production and therefore increase the efficient use of its inputs.

In this study, individual menu items represented DMUs. This was a novel use of DMUs in the foodservice setting. DEA models were constructed using process times, total item cost, and primary ingredient weight as individual inputs and corresponding value of menu items measured as item portion size, published retail price, and item contribution margin as the individual outputs. This approach is consistent with prior literature using DEA (Kumbhakar, 2002). Therefore, the model consisted of multiple inputs versus multiple outputs measured on different scales consistent with DEA modeling requirements. Data were analyzed using the two key types of data envelopment models; the first assumes a constant returns to scale (CRS) and the second assumes a variable returns to scale (VRS). To improve the study's rigor, and given that no one particular approach has been recommended as the better one for foodservice research, both models were used to analyze the data wherever possible. DEA analysis outputs included efficiency scores of menu items and slack scores that represented the magnitude of inefficiency in particular inputs. Slack scores in DEA analysis represent the amount of each input that needs to be reduced to increase efficiency of the DMU. Slack percentages used in this analysis were calculated by taking the slack scores of each input and dividing these by the actual amount of that input used. As the purpose of this study was to evaluate the efficiency of inputs, only input-oriented models were utilized. Once the output of DEA models was generated as efficiency scores and slack scores, these scores were converted into percentages and then analyzed statistically using techniques such as mean difference tests and analysis of variance to evaluate whether there were significant differences in various types of study groups.

To ensure anonymity of participating restaurants, they were coded as Restaurant 1, Restaurant 2, etc. Menu items were coded as L for those using primary local ingredients and N for those using nonlocal primary ingredients. Following the letters L and N were two-digit sequences, the first signifying the sequence of that type of dish from a particular restaurant and the second signifying the restaurant. Therefore, L17 was the first local dish item from Restaurant 7. Types of menu items were not analyzed separately. All restaurants except Restaurant 5 were able to provide complete information as requested by the researchers. During the study Restaurant 5 changed ownership. Although the previous management had been committed to providing information, the ownership change did not allow for that establishment to complete its participation. As a result, this restaurant was dropped from the final analysis of data. This represented a 90% success rate of gathering a complete data set from participating establishments.

RESULTS

Input-oriented constant returns to scale (CRS) and variable returns to scale (VRS) model efficiencies are presented in Table 1. The desired efficiency level is 100%. If the efficiency level is 100% then the DMU is on the efficient frontier maximizing the use of its inputs to produce the identified outputs. Lower efficiency levels are interpreted as inefficient compared to those on the frontier. Input-oriented efficiency means that given the level of outputs, the DMU is either using its inputs efficiently or not. If the efficiency levels are lower than 100%, then the use of inputs to produce the given levels of outputs is inefficient. Therefore, all efficiency scores lower than 100% are characterized as lower in efficiency. The CRS efficiency levels were produced through a constant returns to scale model. The VRS results were produced through a variable returns to scale model. All the inefficient menu items through the CRS model showed an improvement in VRS efficiency levels. These findings clearly suggest a scale effect, implying that if restaurants would increase production of the inefficient menu items, the input usage efficiencies would improve. There was no statistical difference in change of efficiency levels from CRS to VRS between the set of local and nonlocal menu items.

TABLE 1 Overall Input-oriented Efficiency Scores Constant Returns to Scale (CRS) and Variable Returns to Scale (VRS) for Menu Items Prepared Using Local and Nonlocal Primary Ingredients

Restaurants	Local				Nonlocal
	DMU Code	Dish type	Input-oriented CRS efficiency	Input-oriented VRS efficiency	DMU code	Dish type	Input-orientedCRS efficiency	Input-orientedVRS efficiency
1	L11	Entrée	92.17%	100.00%	N11	Entrée	100.00%	100.00%
1	L21	Starter/side	100.00%	100.00%	N21	Dessert	100.00%	100.00%
1	L31	Starter/side	100.00%	100.00%	N31	Starter/side	95.66%	100.00%
2	L12	Entrée	100.00%	100.00%	N12	Entrée	100.00%	100.00%
2	L22	Starter/side	100.00%	100.00%	N22	Starter/side	100.00%	100.00%
2	L32	Dessert	100.00%	100.00%	N32	Dessert	100.00%	100.00%
3	L13	Entrée	100.00%	100.00%	N13	Entrée	100.00%	100.00%
3	L23	Starter/side	77.69%	100.00%	N23	Starter/side	100.00%	100.00%
3	L33	Dessert	91.84%	100.00%	N33	Dessert	100.00%	100.00%
4	L14	Entrée	100.00%	100.00%	N14	Entrée	87.43%	89.77%
4	L24	Entrée	100.00%	100.00%	N24	Starter/side	100.00%	100.00%
4	L34	Dessert	100.00%	100.00%	N34	Dessert	99.62%	100.00%
6	L16	Entrée	100.00%	100.00%	N16	Entrée	100.00%	100.00%
6	L26	Entrée	100.00%	100.00%	N26	Starter/side	100.00%	100.00%
6	L36	Entrée	100.00%	100.00%	N36	Starter/side	100.00%	100.00%
7	L17	Entrée	100.00%	100.00%	N17	Entrée	100.00%	100.00%
7	L27	Entrée	100.00%	100.00%	N27	Entrée	100.00%	100.00%
7	L37	Entrée	100.00%	100.00%	N37	Entrée	100.00%	100.00%
8	L18	Entrée	54.95%	57.69%	N18	Entrée	90.21%	99.21%
8	L28	Entrée	100.00%	100.00%	N28	Entrée	100.00%	100.00%
8	L38	Dessert	100.00%	100.00%	N38	Starter/side	100.00%	100.00%
9	L19	Entrée	100.00%	100.00%	N19	Entrée	100.00%	100.00%
9	L29	Starter/side	75.00%	87.50%	N29	Starter/side	100.00%	100.00%
9	L39	Dessert	68.00%	100.00%	N39	Entrée	100.00%	100.00%
10	L110	Entrée	94.30%	100.00%	N110	Entrée	78.80%	79.51%
10	L210	Entrée	77.84%	100.00%	N210	Entrée	62.74%	100.00%
10	L310	Dessert	50.64%	93.89%	N310	Dessert	83.33%	94.44%
	Mean		91.94%	97.74%			96.71%	98.63%
	SD		14.56%	8.43%			8.55%	4.41%

Analysis was also conducted to assess whether certain types of dishes performed better in efficiency than others. As the number of sides and starters were few, the two categories were combined for this analysis. Further, as the purpose of this study was not to assess differences at the type of dish level, this combination of categories did not lead to any significant loss in result information.

Average efficiency of Starter/side (CRS = 96%; VRS = 99%) was marginally better than that of Entrées (CRS = 95%; VRS = 98%) and Desserts (CRS = 91%; VRS = 98%); however, these differences were not statistically significant (CRS: p = 0.62; VRS: p = 0.74). While the average overall CRS and VRS efficiency scores were higher for nonlocal dishes (CRS = 96.71%; VRS = 98.63%; Table 1), the differences across menu items were not statistically significant, even at p < 0.10. Restaurants 2, 6, and 7 had all their menu items on the CRS frontier (Table 1). Restaurant 2 was an independently owned restaurant located in a hotel, Restaurant 6 was also an independently owned establishment located in a small downtown location, and Restaurant 7 was an operation part of a private club operation. Therefore, operationally, there were few commonalities among the three. While chefs at Restaurants 2 and 7 were professionally trained graduates of culinary programs, Restaurant 6's owner/chef had not acquired any formal food production/service training. One aspect that would have been common among the three establishments was that each of these filled a critical gap through their product/service offering in the local market. Restaurant 2 was an upscale, white table-cloth establishment serving exclusive European-style meals. Restaurant 6 was a quick-service establishment serving exclusive Mexican cuisine. And Restaurant 7 was a foodservice operation that was part of a private club exclusively for members. Therefore, each of these three restaurants had created a unique clientele.

At the restaurant level there was difference of efficiency improvements from the CRS to VRS models (Table 2). Restaurant 10 showed the highest improvements, suggesting that this establishment would gain most from an increase in production of these menu items given the same level of inputs.

TABLE 2 Constant Returns to Scale (CRS) and Variable Returns to Scale (VRS) Efficiency Differences at Restaurant Level

Restaurant		N	(CRS Eff. – VRS Eff.) Mean	Std. deviation	Std. error
1		6	2.03%	3.33%	1.36%
2		6	0.00%	0.00%	0.00%
3		6	5.08%	9.05%	3.70%
4		6	0.45%	0.94%	0.38%
6		6	0.00%	0.00%	0.00%
7		6	0.00%	0.00%	0.00%
8		6	1.96%	3.62%	1.48%
9		6	7.42%	13.04%	5.32%
10		6	20.03%	17.32%	7.07%
Total		54	4.11%	9.63%	1.31%
Model	Fixed effects			8.01%	1.09%
	Random effects				2.17%
	Sum of squares	df	Mean square	F	Sig.
Between groups	0.203	8	0.025	3.960	0.001
Within groups	0.288	45	0.006
Total	0.492	53

CRS model slack percentages were calculated using the absolute slack levels and original input levels. These slack percentages were then analyzed to identify statistical differences between local and nonlocal items and within the 10 restaurant establishments. Statistical comparisons between local and nonlocal menu items are presented in Table 3. There was a significant difference in the time to deliver (p < 0.05) between local and nonlocal menu item ingredients (Table 3). The nonlocal menu ingredients for the surveyed restaurants had the time of delivery inefficiency of only 4.52%, whereas this inefficiency for local menu ingredients was over 20%. Preparation time slack percentage difference between the local and nonlocal menu items was also significantly different at the p < 0.05 level (Table 3). Surprisingly, it took longer to prepare nonlocal menu items compared to the local menu items. Further, results suggest that other input process and outputs were not significantly different between the types of menu items.

TABLE 3 Mean Difference of Constant Returns to Scale (CRS) Model Slack Percentages^(a)—Local Versus Nonlocal Menu Items

Inputs and outputs	Local/nonlocaldish	N	Mean of slacks	Std. deviation	Levene's test for equality of variances		t test for equality of means
					F	Sig.	T	df	Sig. (2–tailed)
Total item food cost	Local dish	26	0.16%	0.82%	11.93	0.00	− 1.58	48	0.12
	Nonlocal dish	24	1.98%	5.80%			− 1.52	24	0.14
Total sourcing time	Local dish	21	9.74%	17.91%	11.54	0.00	− 1.35	40	0.19
	Nonlocal dish	21	20.18%	30.70%			− 1.35	32	0.19
Time to delivery∗	Local dish	24	21.19%	28.61%	19.54	0.00	2.46	46	0.02
	Nonlocal dish	24	4.52%	16.83%			2.46	37	0.02
Receiving time	Local dish	21	14.92%	29.16%	3.85	0.06	0.972	38	0.34
	Nonlocal dish	19	7.28%	18.86%			0.993	35	0.33
Time to storage	Local dish	21	8.44%	21.27%	3.52	0.07	0.907	43	0.37
	Nonlocal dish	24	3.51%	14.99%			0.887	35	0.38
Preparation time∗	Local dish	21	4.69%	14.83%	24.19	0.00	− 2.22	40	0.03
	Nonlocal dish	21	21.36%	31.09%			− 2.22	29	0.04
Primary ingredient weight	Local dish	27	7.97%	15.12%	0.52	0.47	0.628	52	0.53
	Nonlocal dish	27	5.54%	13.26%			0.628	51	0.53
Retail price	Local dish	27	0.73%	2.85%	2.85	0.10	− 0.83	52	0.41
	Nonlocal dish	27	3.22%	15.24%			− 0.83	28	0.41
Total item Portion size	Local dish	27	9.49%	17.96%	7.62	0.01	1.599	52	0.12
	Nonlocal dish	27	3.42%	8.19%			1.599	36	0.12
Item contribution margin	Local dish	27	0.00%	0.00(b)	5.39	0.02	− 1.31	52	0.20
	Nonlocal dish	27	0.00%	0.00(b)			− 1.31	43	0.20
^a.Slack percentages represent the amount of each input reduction required to increase efficiency of the menu item divided by the actual amount of input used.
^b.t cannot be computed because the standard deviations of both groups are 0.
^c.∗Significantly different at p < 0.05.

Slack percentages were also analyzed at the restaurant level. Selected results of these analyses showed that among the nine completed responses from surveyed restaurants, there was a significant difference in receiving times of menu ingredients (p < 0.10), portion sizes (p < 0.05), and CRS efficiency levels (p < 0.00; see Tables 4 and 5). The lowest CRS efficiency was of Restaurant 10 menu items. This restaurant also had the highest inefficiency in portion sizes and among the higher level of inefficiency in receiving times.

TABLE 4 Constant Returns to Scale (CRS) Slack Percentages at Restaurant^(a) Level

	Restaurant	N	Mean	Std. deviation	Std. error
Receiving time	1.00	6	4.38%	6.82%	2.78%
	2.00	6	25.82%	40.00%	16.33%
	3.00	6	0.00%	0.00%	0.00%
	4.00	1	0.00%
	6.00	0
	7.00	3	0.00%	0.00%	0.00%
	8.00	6	0.00%	0.00%	0.00%
	9.00	6	35.54%	39.75%	16.23%
	10.00	6	9.53%	12.04%	4.92%
	Total	40	11.29%	24.80%	3.92%
Portion size	1.00	6	1.73%	3.96%	1.62%
	2.00	6	0.00%	0.00%	0.00%
	3.00	6	20.18%	31.64%	12.92%
	4.00	6	1.18%	1.82%	0.74%
	6.00	6	0.00%	0.00%	0.00%
	7.00	6	0.00%	0.00%	0.00%
	8.00	6	8.33%	12.94%	5.28%
	9.00	6	7.29%	13.57%	5.54%
	10.00	6	19.38%	10.55%	4.31%
	Total	54	6.46%	14.16%	1.93%
CRS efficiency	1.00	6	97.97%	3.33%	1.36%
	2.00	6	100.00%	0.00%	0.00%
	3.00	6	94.92%	9.05%	3.70%
	4.00	6	97.84%	5.10%	2.08%
	6.00	6	100.00%	0.00%	0.00%
	7.00	6	100.00%	0.00%	0.00%
	8.00	6	90.86%	18.02%	7.36%
	9.00	6	90.50%	14.88%	6.08%
	10.00	6	74.61%	15.54%	6.34%
	Total	54	94.08%	12.10%	1.65%
^a.Slack percentages represent the amount of each input reduction required to increase efficiency of the menu item divided by the actual amount of input used.

TABLE 5 Constant Returns to Scale (CRS) Slack Percentages^(a)—Restaurant Level Analysis of Variance

	Sum of squares	df	Mean square	F	Sig.
Receiving time
Between groups	0.714	7	0.102	1.936	0.096
Within groups	1.686	32	0.053
Total	2.400	39
Portion size
Between groups	0.321	8	0.040	2.436	0.028
Within groups	0.741	45	0.016
Total	1.062	53
CRS efficiency
Between groups	0.322	8	0.040	4.000	0.001
Within groups	0.453	45	0.010
Total	0.776	53
^a.Slack percentages represent the amount of each input reduction required to increase efficiency of the menu item divided by the actual amount of input used.

DISCUSSIONS AND IMPLICATIONS

The purpose of this study was to assess whether use of locally grown and produced primary ingredients in menu items costs independent restaurants more than using nonlocal ingredients. In response to the research questions posed for this study the following can be concluded based on the results of data gathered from nine independently owned restaurants in a Midwestern U.S. state: There does appear to be a statistically significant difference in the delivery times of primary local ingredients versus those purchased nonlocally; however, there was no statistically significant difference in the sourcing time of local and nonlocal ingredients; there appeared to be no statistically significant difference in the food cost of local and nonlocal primary ingredients used to prepare the selected dishes for this study; and there does appear to be an establishment effect suggesting that certain restaurants were found to be producing menu items using local ingredients more efficiently than others. Statistically significant differences were found in receiving times and portion sizes of efficient versus inefficient restaurants.

Implications

In view of these results, it appears that while not all production costs of locally grown and produced foods are significantly different from menu items prepared using nonlocal ingredients, delivery time costs may be higher for local ingredients. This implies that local vendor selection and the management of these relationships are critical variables for restaurants to control. This would ensure that local producers/suppliers adhere to delivery requirements and minimize time inefficiencies. Relationship building will also prevent the necessity of changing suppliers often, thereby minimizing high transfer costs of time required to establish new suppliers. On the other hand, while sourcing time difference was not statistically significant, the average time required to source local ingredients was lower. Qualitative observations suggested that this was likely due to already established local vendors for the selected items. While there were no statistically significant differences in the overall CRS and VRS efficiency of local versus nonlocal ingredient menu items, there did appear to be statistically significant differences in the efficiency of restaurants. Qualitative analysis of the efficient restaurant suggests that the establishments that were able to create a unique market for their products were more efficient in their use of locally produced foods versus the others. This included carving out a niche market and also consciously promoting the use of local foods in the menu offerings. The more formal promotion of menu dishes included description and brief introduction of local foods used to prepare the dishes. However, there were also informal promotions in which the chef wrote on a blackboard posted at the reception or entrance that local foods were used for preparing certain dishes or a mention of “today's special” that used a local food item. Both these types of promotion were combined with the waitstaff presenting selected dishes that used locally produced foods.

Finally, given the consistent increase in efficiency scores from CRS to VRS models, there appears to be a scale effect. That is, if participating restaurants were to increase the level of production of the selected surveyed dishes, efficiency scores would improve for the inefficient menu items. Probably the most significant result that needs to be highlighted is that the actual costs as measured by food costs (a proxy for production costs within the DEA model) was not significantly different between locally grown and produced and other items. This is an important result for restaurants wanting to increase use of locally grown and produced foods. Another way to state these results would be that while it possibly won't cost more to use local foods, the dishes using such items will need to be promoted so that restaurants can increase the production volume of local food dishes to maximize on production efficiencies.

In a larger context, this research stresses that foodservice establishments need to assess the financial and economic viability of operating strategies. It also highlights a critical gap in the literature to evaluate various aspects of the operational value chain. Much research is required to enhance our understanding of how and why foodservice establishments are financially and economically successful. This research also proposes a viable study design and methodology to pursue other issues in foodservice operations management, such as operations process design and productivity changes.

Limitations

One of the challenges that researchers faced during this research was the lack of established operating procedures and routines at small independently owned restaurants. As a consequence, there is no systematic maintenance of operating decisions and the information that results from these decisions. Future research could explore these issues. Other issues that could be explored in future research are investigation of the cost and profit structure of different types of menu items, which were not analyzed separately in this study.

CONCLUSIONS

Results of this study suggest that restaurants must carefully select vendors when planning to use locally produced foods. While restaurants' production costs as measured by the cost of food purchased are not likely to be higher, it would be critical to assess market capacity and accordingly plan production levels. Most importantly, restaurants must strive to clearly differentiate their products in the marketplace, either by clearly communicating the use of local foods in their menus or by creating unique menu dishes. The result of this differentiation would be seen as increased volume, which will also influence the efficiency of using local ingredients in menu items. In addition to contributing to foodservice literature, this study is timely and practical as restaurant owners can utilize findings to improve profitability and productivity. Still, future studies could build on this current research due to its inability to generalize the results. For instance, further analysis could be done by more clearly characterizing restaurants and menu items and focusing on only certain types versus a convenient sample: for instance, casual dining versus white tablecloth, quality of service employees, and various production/preparation techniques.

This study was made possible by funding from the Leopold Center for Sustainable Agriculture at Iowa State University, Ames, Iowa