The relation between organization size and
labor productivity
In the Dutch healthcare sector
Iwana Suzanne Mietus
Student number 10294805Supervisor: Prof. dr. C.M. (Matthijs) van Veelen Second supervisor: Prof. dr. E.J.S. (Erik) Plug Date: February 2nd, 2016
Bachelor’s Thesis
Economics & Business, track Economics
Table of contents
Statement of Originality p. 2
§1. Introduction p. 3
§2. Theory p. 4
§2.1 Diminishing marginal returns p. 4
§2.2 Returns to scale p. 5
§2.3 Economies of scale p. 5
§3. Methodology p. 7
§3.1 Data p. 7
§3.2 Model & statistical approach p. 8
§4. Results p. 10
§4.1 Descriptive results p. 10
§4.2 General model p. 12
§4.3 Small, medium and large size organizations p. 13
§4.4 Mental, disability and elderly care p. 14
§5. Conclusion p. 15
§6. Discussion p. 16
Reference list p. 18
Statement of Originality
This document is written by Student Iwana Mietus who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
§1. Introduction
In the Netherlands, the size of organizations in the public and semi-public sector has increased over the last thirty years, and these organizations are the largest within Europe (CPB, 2013). One would expect that larger organizations are performing better because of economies of scale. This is defined as: “Situation in which output can be doubled for less than a doubling cost” (Pindyck & Rubinfeld, 2009, p. 246). Economies of scale is a topic from the microeconomic field of research. Economies of scale occur due to specialization of employees, flexibility of variable inputs and cheaper inputs as a result of more bargaining power (Pindyck & Rubinfeld, 2009, pp. 245-246). Furthermore, there exist larger departments and more formalized processes (Mintzberg, 1983). A larger firm is also expected to have more market power, which leads to higher profit margins (CPB Notitie, 2013).
However, the relevant literature also shows a law of diminishing marginal returns, which is defined as: “Principle that as the use of an input increases with other inputs fixed, the resulting additions to output will eventually decrease” (Pindyck & Rubinfeld, 2009, p. 202). Larger organizations become less efficient due to bureaucracy and indecisiveness for example (Mintzberg, 1983, pp. 124-126). This means that there possibly exists an optimal size for healthcare organizations and that they can grow too big. The idea of an optimal size in healthcare organizations is supported by Tarcan et al., who did research on the optimal size of Turkish hospitals (2015).
To my best knowledge, the relation between size and productivity levels in the Dutch healthcare sector has not been studied yet. The Netherlands has to deal with an aging population, longer average life expectancy and budget cuts in healthcare (CPB, 2013). Because demand for care increases and the related budget decreases, it is important to organize the Dutch healthcare organizations in the most efficient way. Whether economies of scale exist can be measured by a rise or fall in productivity levels. This leads to the main research question:
What is the relation between the size of healthcare organizations in the Netherlands and the productivity level of employees?
The hypothesis from basic theory is that there exist economies of scale followed by diseconomies of scale when an organization grows too big. Results of empirical studies in other sectors are ambiguous with respect to economies or diseconomies of scale (Hughes and Mester, 2013; Kasman, 2012; Davies and Tracey, 2014). There exists a lacuna in the knowledge about economies of scale in general and in healthcare organizations in the Netherlands, and therefore it is scientifically relevant to do research in this field.
The structure of my thesis is as follows. In paragraph 2, I provide an overview of the theory with respect to economies of scale. Thereafter, in paragraph 3, I outline the methodology by describing the data. Furthermore, I show the equations that will be estimated by Ordinary Least Squares. In paragraph 4, I show results of the general model and individual type of organization models. These are categorized in size: small, medium and large size organizations, and in type: mental, elderly and disability care. Finally, in paragraph 5, I conclude with a summary of my main results. The conclusion is followed by a discussion.
§2. Theory
To examine the relation between size and productivity, I describe the relevant theory first. To start off, the relation between production and extra input from labor in microeconomic theory will be described. Followed by the concepts ‘returns to scale’ and ‘economies of scale’.
§2.1 Diminishing marginal returns
In general it is expected that the level of production is higher when the number of employees rises. The reason for that is that every employee adds some production. When the number of employees is low, an additional employee adds more than the ones that already work. An example is illustrated in Figure 1, where production of jackets per day is on the y-axis and units of labor per day is on the x-axis.
The first unit of labor produces one jacket. When a second employee is added, the two employees together do not produce two, but three jackets. A smoother production process can explain this. Employees can help each other and divide tasks. Therefore, they can both work faster and produce more: this is called increasing marginal returns. When extra units of labor are added, the marginal returns per unit can decrease. In this example, the fifth employee possibly needs to wait for scissors until the rest is done. Therefore he
works slower and can produce less: this is called
diminishing marginal returns. When an eighth employee is added, a negative marginal return has been rised. This can be explained by the possibility that an additional employee makes is difficult to organize the schedules or makes the working space overcrowded, which causes distractions. This principle is called the ‘Law of diminishing marginal returns’ and means: “Principle that as the use of an input increases with other inputs fixed, the resulting additions to output will eventually decrease” (Pindyck & Rubinfeld, 2009, p. 202). In our example this means that while the units of labor rise, the number of scissors, sewing machines and other capital do not
rise.
The law of diminishing marginal returns holds in the short term, because at least one input factor is stable. This is shown in the graph beneath. The first production line (O1) is related to the graph above. The lines O2 and O3 are new time periods with technological progress or capital growth, so that employees become more productive.
Figure 1. Diminishing marginal returns
In the long term, it is shown that marginal returns still diminish, but the optimal size moves to the right. The optimum in the first period is on point A with 6 employees, because an additional employee does not generate extra returns. In that case, the cost will rise and the returns will not rise, which is not beneficial for the firm. In the second period (O2) it is beneficial to hire a 7th employee (point B), which is the optimum. In the final period, the optimum is at 8 employees (point C) (Pindyck & Rubinfeld, 2009, pp. 201-203).
§2.2 Returns to scale
An important concept that relates to the diminishing marginal returns, is returns to scale. This is defined as: “Rate at which output increases as inputs are increased proportionally” (Pindyck & Rubinfeld, 2009, p. 2015). Three types of returns to scale exist: increasing, constant and decreasing returns to scale. When a double input creates more than a double output, there exist increasing returns to scale. This is shown in the Figure 3 in the movement from point A to point B. When capital and labor are both doubled from 1 to 2, the production rises from 1 to 3. If output doubles when input is doubled, the returns to scale are constant. This is shown in the movement from point B to point C. When input is doubled, but output is less than doubled, decreasing returns to scale exist. This is shown in the movement from point C to point D. Decreasing returns to scale is present in some companies that produce on a large scale, because this can lead to a lower productivity of labor and capital. It is important to note that returns to scale highly differ between different companies and sectors. Capital-intensive sectors usually have higher returns to scale than service sectors (Pindyck & Rubinfeld, 2009, pp. 215-216).
§2.3 Economies of scale
Related to the concept ‘returns to scale’ is the concept ‘economies of scale’, defined as: “Situation in which output can be doubled for less than a doubling cost” (Pindyck & Rubinfeld, 2009, p. 246). Its counterpart, ‘diseconomies of scale’, is the situation in which double production requires more than double costs. In a study on household economies of scale for example, it is shown how couples save “one-third of their total expenditures” in comparison to single households (Browning, Chiappori and Lewbel, 2013). Similarily, research in banking shows that large banks can also experience scale economies (Hughes and Mester, 2013). Kasman (2012) supports this and claims that there are no diseconomies
of scale for larger banks. This could mean that there is no law of diminishing marginal returns, or that the optimal size lies above the largest bank in the sample. Other research regarding the banking sector shows that too-big-to-fail banks do not show evidence of scale economies, while small banks do (Davies and Tracey, 2014).
Increasing returns to scale can be seen as a part of economies of scale, but economies of scale are more generic. In general, a company experiences economies of scale in low production levels and diseconomies of scale in higher production levels. When production rises and an organization becomes larger, there are several reasons to expect the average cost will fall and several reasons to expect the average cost will rise (Pindyck & Rubinfeld, 2009, pp. 245-246). When production increases it can be expected that average costs fall because of three reasons:
1. Employees can specialize in the parts of the production process where they are most productive.
2. Scale can provide flexibility, because the combination of inputs varies and can be used more effectively.
3. Purchasing products in a larger scale is usually cheaper because the organization has more bargaining power.
The Dutch Central Planning Bureau, or ‘Centraal Planbureau’ (CPB), adds that a larger scale in the primary process increases the level of experience and expertise for employees (Notitie, 2013). According to Tarcan et al. (2015), there exist economies of scale in Turkish hospitals. Hughes and Mester (2013) provide evidence that larger banks experience larger economies of scale not because of too-big-to-fail advantages that large banks have, but because of technological advantages like diversification and the spreading of information costs. Kasman (2002) describes how all size groups of banks show statistically significant economies of scale. This is particularly the case for small and medium-size banks. Therefore he concludes that the average bank in his sample has not reached the optimal size yet and should become larger, either by increasing the scale or by merging. Kasman confirms the technological advantages of larger banks.
On the other hand, there are also reasons to believe that the average production cost will rise when production rises.
1. In short term capital and space are limited which decreases efficiency.
2. Managing a larger organization is more complex and less efficient as the number of tasks rises.
3. When an organization becomes very large, it is not cheaper to purchase more products because scarcity pushes the price upwards.
Other effects mentioned by classic organizational scientist Henry Mintzberg (1983, pp. 124-126) are:
1. A larger organization is more specialized, departments are more differentiated and the administrative component is further developed. Both a rise and a fall in cost can be the consequence. Specialization is usually associated with lower costs, but more administration shows a more complex and probably less efficient organization. 2. Departments within larger organization are bigger on average than within smaller
organizations. This can lead to lower costs because there are more employees per manager.
3. The processes in a larger organization are more formalized. This can be efficient, but also bureaucratic and less effective.
§3. Methodology
In this paragraph the methods used in this thesis are explained. First the data will be described and the variables are defined. Subsequently the model and statistical approach will be outlined.
§3.1 Data
The data are retrieved from the Berenschot Benchmark Care database. Every year the benchmarking department of Berenschot, in 2015 ranked as the best consulting firm of the Netherlands according to research from ‘Management Team’ (website www.mt.nl), invites all healthcare organizations in the Netherlands to join their benchmark. Healthcare organizations can join if they pay the participation fee. Every participating healthcare organization fills in a detailed questionnaire. Berenschot uses these data to compare organizations with similar care products (elderly, mental and disability care) and similar size on several indicators, like overhead percentage and overhead cost, and the productivity ratio. This database includes 145 unique healthcare organizations over the last five years. When an organization has participated in the benchmark more than once over the last five years, the most recent data are used. The definitions of the variables that are relevant in this study are given in Table 1.
Variable Definition
Productivity The income of an organization divided by the cost of healthcare in the primary process
Size
FTE: the number of full time equivalents in the organization Employees: the number of people working in the organization LogFTE: the natural logarithm of FTE
Parttime The difference between employees and FTE, divided by the number of full time equivalents
Overhead Formation: the percentage overhead formation as part of the total formation
Cost: the percentage overhead cost as part of the total cost Year The year in which the organization provided there data
FTEsmall Dummy variable with value 1 for health organizations with less than 500 FTE and 0 for all other cases
FTElarge Dummy variable with value 1 for health organizations with more than 1000 FTE and 0 for all other cases
Table 1. Definition variables
The dependent variable is Productivity, measured by the revenue of an organization divided by the cost of healthcare in the primary process. A healthcare organization realizes income by delivering care to patients. The patient’s healthcare insurance companies pay for this care. Hence, the organization’s income is a measure of the value of production in a healthcare organization. It is not clear whether productivity level has a positive or negative effect on the quality of healthcare and the satisfaction level of patients.
The main independent variable is Size, measured by variables FTE and Employees, both as indicators of organization size. The first one, FTE, refers to the number of full time equivalents within an organization. The number of employees1 is higher than the number of full time equivalents, as some employees work part time and no employees work more than full time. Still, the two variables are highly correlated, which would cause multicollinearity. A third measurement is the natural logarithm of FTE. Because FTE and productivity have a very different range and functional form, the effects can be better estimated by using the natural logarithm of FTE.
The organizations that participate in the Berenschot benchmark research are not a random sample, which could cause a selection bias. What they usually have in common is that these organizations are interested in their overhead, costs and productivity. However, their results show that some organizations are interested in these topics because they perform very well, and others because they do not perform well at all. Furthermore, the organizations are of a wide range of sizes. The specific range can be seen in the descriptive statistics section in this thesis. Organizations from all 12 provinces of the Netherlands and from urban as well as non-urban areas participate in the benchmark. Therefore, this research sample is valid and representative for other healthcare organizations in the Netherlands and healthcare organizations in countries similar to the Netherlands.
§3.2 Model & statistical approach
In order to measure the relation between organization size and productivity, four regression equations are estimated in this study by the Ordinary Least Squares (OLS) method. The first one is the general model:
(1) 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 = 𝛼₀ + 𝛼₁𝐿𝑜𝑔𝐹𝑇𝐸 + 𝛼₂𝑌𝑒𝑎𝑟 + 𝛼₃𝑃𝑎𝑟𝑡𝑡𝑖𝑚𝑒 + 𝛼₄𝑂𝑣𝑒𝑟ℎ𝑒𝑎𝑑 + ℇ
Variations are shown where the least significant independent variables are taken out of the model. The parameters α₀ indicates the constant of the model. Parameter α₁ indicates the effect of the independent variable LogFTE as an indicator for organization size on productivity. In this model the logarithm of FTE is used, because productivity is an indicator with a small range and FTE is an indicator with a large range. The logarithmic model’s functional form has a better fit with the data than the linear model. The other parameters: α₂, α₃ and α₄ indicate the effect of the control variables which are described below. The parameter Ɛ is the error term. The model is tested for heteroscedasticity with the robust standard error option in Stata. It has to be noted that it is beyond the scope of this research to study the causality of the relationship between organization size and productivity.
Next to the general model, a model is estimated where organization size is categorized in small organizations (with less than 500 FTE), medium size organizations (between 500 and 1000FTE), and large organizations (more than 1000 FTE).
(2) 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 = 𝛽₀ + 𝛽₁𝐹𝑇𝐸𝑠𝑚𝑎𝑙𝑙 + 𝛽₂𝐹𝑇𝐸𝑙𝑎𝑟𝑔𝑒 + 𝛽₃𝑌𝑒𝑎𝑟 + 𝛽₄𝑃𝑎𝑟𝑡𝑡𝑖𝑚𝑒 + 𝛽₅𝑂𝑣𝑒𝑟ℎ𝑒𝑎𝑑 +ℇ
1
In the second model, the same variations as in equation (1) are shown. The parameter β₀ indicates the constant of the model. The parameters β₁, β₂, indicate the effect of the independent dummy variables for small and large organizations on productivity. Parameters β₃, β₄ and β₅ indicate the effect of the control variables and parameter Ɛ is the error term.
The first two models are estimated using the whole sample of 145 healthcare organizations. The third model is estimated when the sample is cut into three categories: organizations smaller than 500 FTE, between 500 and 1000 FTE, and organizations larger than 1000 FTE.
(3) 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 ( 𝑠𝑚𝑎𝑙𝑙, 𝑚𝑒𝑑𝑖𝑢𝑚, 𝑙𝑎𝑟𝑔𝑒) = Ѳ₀ + Ѳ₁𝐿𝑜𝑔𝐹𝑇𝐸 + Ѳ₂𝑌𝑒𝑎𝑟 + ℇ
Next to a categorization in size, three types of healthcare organizations are identified in the Berenschot benchmark database: elderly care, disability care and mental care. To estimate the effect of size on productivity a third model is estimated using three different samples: mental care, elderly care and disability care. This model is shown in equation (4).
(4) 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 (𝑚𝑒𝑛𝑡𝑎𝑙, 𝑑𝑖𝑠𝑎𝑏𝑖𝑙𝑖𝑡𝑦, 𝑒𝑙𝑑𝑒𝑟𝑙𝑦) = ɣ₀ + ɣ₁𝐿𝑜𝑔𝐹𝑇𝐸 + ɣ₂𝑌𝑒𝑎𝑟 + ℇ
Model (3) and (4) are simplified versions of equation (1). In these models, Ѳ₀ and ɣ₀ are the constants. The parameters Ѳ₁ and Ѳ₂ estimate the effect of LogFTE and Year on productivity for small, medium and large size organizations. The parameters ɣ₁ and ɣ₂ estimate the effect of LogFTE and Year on productivity for mental, elderly and disability care organizations. The parameter Ɛ is the error term.
The first control variable that is added in all models is Year, which shows the year that an organization provided the data. The possible outcomes are: 2010, 2011, 2012, 2013, and 2014. When an organization participated in the Berenschot benchmark more than once in the last five years, the most recent observation is used. Because there have not been big changes in the last five years in this sector, it is expected that there is no effect in productivity over the years.
With the two variables FTE and Employee a new variable is created: Parttime. This variable indicates the level of part time employees, calculated as: the difference between number of employees and FTE, per formation place. The variable is created to control whether the difference in full timers and part timers has an effect on the productivity, and if it changes the effect of size on productivity. In the Netherlands, the percentage of part time employees is relatively large (Wielers and Raven, 2013).
The variable Overhead is also controlled for in the model. The percentage overhead formation and overhead cost of total formation and cost are added as control variables. These two are highly correlated of course; so only overhead formation is used in the empirical model.
Lastly, two dummy variables are created to control for categorized organization size. The first, FTEsmall, is a dummy variable with value 1 for health organizations with less than 500 FTE and 0 for all other cases. The second, FTElarge, is a dummy variable with value 1 for health organizations with more than 1000 FTE and 0 for all other cases. When both dummies are zero, the health organization has a medium size between 500 and 1000 FTE. The categorization is based on the three categories Berenschot uses in their benchmark database and research.
§4. Results
After the description of the data, model and statistical approach in the last paragraph, the results will now be presented. First, the descriptive statistics of the chosen variables for the total sample and different types of healthcare will be shown. This is followed by an overview of percentage small, medium and large organizations per healthcare type. Thereafter the general model estimated by the first equation is given. Then, the model estimated by the second equation with dummy variables for organization size is shown. Lastly, separate simple models from equation (3) for small, medium and large organizations and from equation (4) for mental, elderly and disability care are shown.
§4.1 Descriptive results
Before presenting the estimation results, the descriptive results can be seen in Table 2. In the first column the relevant variables are shown, followed by the number of observations (N), mean, standard deviation, minimum and maximum value. The results are presented per variable. First, the results from the total sample can be seen, thereafter the results are specified for the three different types of healthcare: mental care, disability care and elderly care.
It is interesting to note the large differences in minimum and maximum for FTE and employees. Furthermore the large difference in mean for FTE and employees is noteworthy. As written in the methodology section, the variable LogFTE is created to have a better fit with the model. The effect on productivity can also be better estimated because the variables are more similar in range.
With regard to the differences between types of healthcare, it is interesting to note the higher productivity in mental care. This is the case even though the range is wider than in disability and elderly care. Furthermore it is shown that mental care organizations are (approximately 300 FTE) larger on average than disability and elderly care organizations. The average of the part time indicator, employees minus FTE divided by FTE, is the smallest in mental care organizations (0,31), middle score in disability care organizations (0,57) and the largest in elderly care organizations (0,79). The high level of part time employees in elderly care could by explained by the high level of small contracts in home care and in cleaning assistance for elderly patients (Berenschot Benchmark Care). The percentage overhead formation and cost is higher in mental care, and similar in disability and elderly care. This could be the case because mental care treatments are oftentimes short, whereas elderly and disability care treatments are usually for a longer term (Berenschot Benchmark Care). In general it can be concluded that the descriptive statistics of mental care organizations deviate from that of disability and elderly care organizations.
Variable N Mean Std. Dev Min Max Productivity 145 1,97 0,22 0,81 2,80 Mental care 26 2,04 0,36 0,81 2,80 Disability care 58 1,96 0,18 1,45 2,38 Elderly care 61 1,95 0,19 1,34 2,42 FTE 145 1300,89 1218,96 77,25 8912,95 Mental care 26 1547,31 1238,94 247,15 6399 Disability care 58 1252,57 1361,53 103,21 8912,95 Elderly care 61 1241,80 1064,57 77,25 4929,30 Employees 138 2036,95 1980,51 130 13569,5 Mental care 24 2096,05 1600,45 362 7682 Disability care 55 1867,75 2090,96 166 13569,5 Elderly care 59 2233,80 2027,90 130 10377 Parttime 137 0,62 0,31 0,02 2,39 Mental care 24 0,31 0,13 0,02 0,56 Disability care 54 0,57 0,25 0,09 1,83 Elderly care 59 0,79 0,30 0,22 2,39 Overhead formation 145 0,15 0,04 0,04 0,27 Mental care 26 0,21 0,03 0,15 0,27 Disability care 58 0,14 0,03 0,04 0,21 Elderly care 61 0,13 0,02 0,07 0,20 Overhead cost 145 0,18 0,04 0,08 0,28 Mental care 26 0,22 0,03 0,17 0,28 Disability care 58 0,17 0,03 0,08 0,25 Elderly care 61 0,16 0,03 0,11 0,25 Year 145 2012,8 1,35 2010 2014 Mental care 26 2012,7 1,44 2010 2014 Disability care 58 2012,9 1,35 2010 2014 Elderly care 61 2012,7 1,33 2010 2014
Table 2. Descriptive results.
To take the difference in organization size between healthcare types into consideration, the percentages small, medium and large size organizations are shown in Table 3. The total numbers in absolute terms are also shown in the table. Most organizations are large, but in mental care, the percentage of large organizations is the largest in comparison to the other healthcare types. These findings are in line with the results in Table 2. The percentages of small, medium and large size organizations are similar for disability and elderly care organizations.
Organization type Small Medium Large Total number
Mental care 23% 8% 69% 26 (100%)
Disability care 24% 31% 45% 58 (100%)
Elderly care 16% 36% 48% 61 (100%)
Total number 30 42 73 145
Table 3. Size per healthcare type.
§4.2 General model
The results of the estimation of the model are shown in Table 4. Both the whole model and the most relevant variations estimated with OLS can be seen. When studying Model 1, a statistically significant (on a 1 per cent level) negative effect (-0,191) of LogFTE on Productivity is shown. It is interesting to note that in all models, the effect of LogFTE is approximately the same negative effect on a 1 per cent significance level. This is the case regardless of the control variables that are added. This means that in larger organizations, employees are less productive than in smaller organizations. However, it has to be noted that this is not necessarily a causal relationship.
When studying the Model 2, 3 and 4, a statistically significant negative effect of the part time indicator (difference in FTE and Employees per FTE) on productivity can be seen on a 5 per cent and 10 per cent level. When more employees work part time in a healthcare organization, productivity falls. This could be explained by for example the extra time needed to transfer work between colleagues. Furthermore, a weekly meeting of an hour has a twice as large impact on an employee that works 16 hours than on an employee that works 32 hours. All other control variables have insignificant effects on productivity.
In all the models, R-squared is low, probably because of the relatively low number of observations. In Model 1 and 2 (0,097 and 0,098 respectively) it is lower than in Model 3 and 4 (0,172 and 0,179 respectively). This can be explained by the significant influence of the difference in FTE and employees in the last two models.
Variable Model 1 Model 2 Model 3 Model 4
LogFTE -0,191*** -0,191*** -0,230*** -0,232*** (0,063) (0,063) (0,064) (0,064) Overhead formation 0,492 (0,581) Parttime -0,182** -0,152* (0,078) (0,079) Year -0,005 0,001 0,003 (0,014) (0,014) (0,014) Constant 2,538*** 12,122 0,135 -2,422 (0,187) (27,286) (27,377) (28,789) N 145 145 137 137 R-squared 0,097 0,098 0,172 0,179
* Significant at 10 per cent level, ** Significant at 5 per cent level, ***Significant at a 1 per cent level
§4.3 Small, medium and large size organizations
The results of equation (2) estimated by OLS can be seen in Table 5. In this model, three categories are created: a small organization (with less than 500 FTE), a medium size organization (between 500 and 1000 FTE) and a large organization (more than 1000 FTE). Four models are estimated, with different control variables added.
All models show a relatively stable positive significant effect of small organizations and a relatively stable negative insignificant effect of large organizations on productivity. The positive effect of size for small organizations ranges from 0,099 to 0,135 and is significant on a 5% level in Model 3 and 4, and significant on a 10 per cent level in Model 1 and 2. This means that the productivity level is higher in small organizations than in medium and large organizations. It also suggests that the productivity level in large organizations is lower than in small and medium size organizations, but this effect cannot be precisely estimated and is not statistically significant. The results can indicate that a large number of healthcare organizations exceed the optimal size if there exists an optimum. This is in line with the theory of economies of scale and the law of diminishing marginal returns as discussed in paragraph 2.
Just like in the first model, the part time indicator (difference between FTE and employees divided by FTE) results in a negative effect. However, this effect is only significant on a 10 per cent level. In all the models, R-squared is low. This means that there are other factors that explain the variance of productivity that are not in this model. Usually R-squared is lower in studies with a lower number of observations.
Variable Model 1 Model 2 Model 3 Model 4
FTEsmall 0,099* 0,100* 0,135** 0,130** (0,058) (0,058) (0,055) (0,055) FTElarge -0,029 -0,028 -0,034 -0,038 (0,035) (0,035) (0,034) (0,032) Overhead formation 0,247 (0,632) Parttime -0,168* -0,153* (0,091) (0,90) Year -0,006 -0,001 -0,000 (0,014) (0,014) (0,015) Constant 1,964*** 14,527 4,297 2,914 (0,023) (27,831) (28,383) (30,166) N 145 145 137 137 R-squared 0,047 0,049 0,115 0,116
* Significant at 10 per cent level, ** Significant at 5 per cent level, ***Significant at a 1 per cent level
Table 5. Results total sample with categorized size, dependent variable Productivity.
The relation between size and productivity per size category
The results of equation (3) for small, medium and large size organizations estimated by OLS can be seen in Table 6. In all samples there exists a negative effect of size on productivity. For small and medium size organizations, this effect is not statistically significant. It is
possible that the reason for this insignificancy is the relatively low number of observations in these samples, which makes the estimator less precise.
In large organizations, meaning organizations larger than 1000 FTE, the negative effect of size on productivity is statistically significant. When the control variable Year is added, the effect of LogFTE on productivity stays approximately the same. It is difficult to conclude a causal relationship between size and productivity, as the model is simple and the explained variance is relatively low (R-squared is 15 per cent in the large sample). The negative effect of size on productivity in large organizations could suggest that large organizations exceed the optimal size in terms of FTE. In other words, it would not be beneficial for the productivity level of a healthcare organization to grow when the number of FTE is already higher than 1000. This would be a case of diseconomies of scale and could be explained by bureaucracy, the high level of complexity in larger organizations, or higher prices due to a large demand as described in paragraph 2. The negative result is in line with the hypothesis that there exists a law of diminishing marginal returns. This leads to a decreasing productivity when an organization grows too large. However, also other factors could influence the productivity level that are not included in this model, and a causal relationship cannot be concluded from this model.
Variable Small (1) Small (2) Medium (1) Medium (2) Large (1) Large (2) LogFTE -0,182 -0,186 -0,148 -0,189 -0,433** -0,431** (0,187) (0,188) (0,275) (0,267) (0,202) (0,202) Year 0,020 -0,018 -0,004 (0,037) (0,019) (0,020) Constant 2,507*** -37,596 2,385*** 39,145 3,345*** 10,617 (0,439) 73,391 (0,777) (37,827) (0,651) 39,729 N 30 30 42 42 73 73 R-squared 0,018 0,27 0,008 0,036 0,154 0,155
* Significant at 10 per cent level, ** Significant at 5 per cent level, ***Significant at a 1 per cent level
Table 6. Results small, medium and large size, dependent variable Productivity.
§4.4 Mental, disability and elderly care
The results of equation (4) for mental, disability and elderly care estimated by OLS can be seen in Table 7. The models are estimated for the three categories: health organizations that provide mental, disability or elderly care. In all models the effect of size on productivity is negative. In the case of mental care, this effect is significant on a 1 per cent level and relatively high in comparison to all other models in this study: between -0,706 and -0,711. This indicates that productivity falls when a mental care organization employs more employees. It is not studied why this is the case and whether a causal relationship exists or not. R-squared is relatively high in mental care: around 48% of the variation in productivity is explained by the LogFTE indicator for size. This is especially high when it is taken into account that there are only 26 observations in mental care. The significant negative effect and high R-squared could be explained by the high percentage of large organizations in mental care (69%). In the previous paragraph it is shown that there only exists a significant negative effect of size on productivity for large organizations.
In disability care, the effect of size on productivity is not statistically significant. In elderly care, the effect of size on productivity is negative and significant on a 5 per cent level. R-squared is much lower in these models than in the models for mental care, even though the number of observations is doubled. In all models, the control variable Year does not have a significant effect on productivity and does not dramatically change the effect of size on productivity. Variable Mental care (1) Mental care (2) Disability care (1) Disability care (2) Elderly care (1) Elderly care (2) LogFTE -0,706*** -0,711*** -0,060 -0,063 -0,157** -0,163** (0,227) (0,229) (0,069) (0,068) (0,071) (0,070) Year -0,011 -0,026 0,012 (0,036) (0,020) (0,017) Constant 4,202*** 26,143 2,139*** 55,227 2,410*** -22,471 (0,681) (72,838) (0,206) (39,613) (0,217) (34,402) N 26 26 58 58 61 61 R-squared 0,479 0,481 0,017 0,057 0,088 0,096
* Significant at 10 per cent level, ** Significant at 5 per cent level, ***Significant at a 1 per cent level
Table 7. Results mental, disability and elderly care, dependent variable Productivity.
§5. Conclusion
In this thesis I estimated the relation between organization size and productivity in the Dutch healthcare sector. The data are retrieved from the Berenschot Benchmark Care database. I estimated the effect of healthcare organization size on productivity with the Ordinary Least Squares method, and controlled for percentage overhead formation, a part-time indicator (difference in FTE and employee per FTE), and the year the organization participated in the benchmark.
The main finding is a statistically significant negative effect of size on productivity, which is stable when different control variables are added. This means that the larger an organization is in terms of FTE, the less productive employees are. This indicates that on average Dutch healthcare organizations are larger than the size in which employees achieve an optimal productivity level. However, it is not clear whether there exists a causal relationship between organization size and productivity, or whether other factors influence the effect. Next to size, the percentage of part time employees has a negative effect on productivity. When more employees work part time, the average productivity falls. This is especially the case in elderly care organizations and could for example be explained by the small contracts in home care and cleaning assistance for elderly patients.
After this general model, I estimated a model with dummy variables indicating the sizes ‘small’ (less than 500 FTE) and ‘large’ (more than 1000 FTE). The remaining category is ‘medium’ (between 500 and 1000 FTE). Results show that small organizations have a higher level of productivity than medium and large organizations. The three categories are also estimated in separate models. Only in the model for large organizations a statistically significant effect is found, which is negative. This indicates that organizations larger than 1000 FTE lose productivity if they grow in terms of FTE. This is a case of diseconomies of scale and corresponds to the law of diminishing marginal returns. It has not been studied whether this statistical effect exists because of a causal relationship between size and
productivity, or that other factors that both correlate with size and productivity cause this effect. If there would exist a causal relationship between size and productivity, the results would indicate that if there exists an optimal size, a large part of healthcare organizations in the Netherlands exceed this optimum. It is assumed that these results can be generalized to other healthcare organizations in the Netherlands and to healthcare organizations in similar countries.
In addition to a categorization in size, also a categorization in health care type is made. I estimated simple separate models for mental care, disability care and elderly care. The effect of size on productivity is negative in mental and elderly care. This indicates that employees in larger organizations are less productive. As mentioned before, it cannot be concluded from this study that this relationship is causal. The effect is significant on a 1 per cent level in mental care and the model has a high R-squared (0,48). It could be the case that the high percentage of large organizations in mental care causes the high R-squared. The results are not statistically significant in disability care.
To summarize, organization size has a negative relation with productivity levels in the Dutch healthcare sector. Whether the lower productivity in larger organizations is caused by its size or by other factors has not been studied in this thesis. The effect in mental care organizations shows the highest level of significance and a high R-squared. Employees in organizations smaller than 500 FTE have the highest productivity levels on average. When an organization is larger than 1000 FTE, an extra employee will cause a significant negative effect on productivity.
§6. Discussion
In the last paragraph I concluded how size has a negative relation with productivity in Dutch healthcare organizations. Now I will discuss my results and shortcomings of this study, followed by suggestions for further research.
First, it is beyond the scope of this research to study the causality of the relationship between organization size and productivity. Therefore it could be the case that other variables influence both size and productivity and cause the negative relation that results from the database. The causality of the negative relationship would be an interesting topic for further research.
Furthermore, the number of observations in the dataset is limited, and there exists a potential selection bias because healthcare organizations choose to participate in the benchmark. However, the participating organizations provide different healthcare types, have a wide range of financial performances and are located in different areas of the Netherlands. Furthermore, the models estimated with OLS are relatively simple. Therefore, it is assumed that the results are representative and valid for healthcare organizations in the Netherlands and for similar organizations in countries similar to the Netherlands. The Berenschot Benchmark Care database that is used is the best possible sample available at this moment.
In further research, it would be interesting to expand the dataset and use data of more healthcare organizations in the Netherlands. It would also be interesting to compare these results to findings in other types of healthcare, like hospital care. The findings could also be compared to similar healthcare organizations in other countries than the Netherlands. The comparison is especially relevant because healthcare organizations in the
Netherlands are on average the largest in Europe. Extra data could not be collected due to the limited time and scope of this thesis.
Next to expanding the dataset, it would also be intriguing to study why the negative relation between size and productivity exists. Does a causal effect of size on productivity exist? Or is there another factor that influences both size and productivity? And, why is the effect of size on productivity the largest in mental care? As productivity is measured by revenue divided by cost in the primary process; does revenue fall or does cost increase? And, why is the effect of size on productivity in disability care insignificant? Incorporating the further research questions, it would be pleasurable to find a more specific optimal organization size per organization type. Furthermore, the negative effect of the part-time indicator on productivity could be a fascinating topic for research. This is relevant because a high percentage of employees in the Netherlands work part time compared to other European countries. Lastly, it would be intriguing to study the effects of productivity on the quality of healthcare and satisfaction levels of patients. But for now, thought provoking results have been delivered that are relevant for current and future societal challenges, and for studying the lacuna in the knowledge about economies of scale in general and in healthcare organizations in the Netherlands.
Reference list
Aggarwal, R.K., Evans, M.E., Nanda, D. (2012) Nonprofit boards: Size, performance and managerial incentives. Journal of Accounting and Economics, Vol 53, Issues 1-2, pp. 466-487.
Blank, J., Haelermans, C., Koot, P., Putten-Rademaker, O. van (2008) Schaal en zorg. TU Delft IPSE Studies. Digitally retrieved through:
http://www.eumonitor.nl/9353000/1/j4nvgs5kjg27kof_j9vvik7m1c3gyxp/vhw4c948vpzx/f =/blg16409.pdf
Browning, M., Chiappori, P. Lewbel, A. (2013) Estimating Consumption Economies of Scale, Adult Equivalence scales, and Household Bargaining Power. Review of Economic Studies, Vol 80, 1267-1303.
Brummelman, E.G. (2008) Schaal en zorg: internationale vergelijking, in: Schaal en zorg, Achtergrondstudies, rvz, p. 7-50.
CIZ (2014) Trendrapportage 2013. Digitally retrieved through: https://www.ciz.nl/voor-professionals/Documents/CIZ_Trendraportage_2013.pdf
CPB (2013) Toekomst voor de zorg. Digitally retrieved through: http://www.cpb.nl/campagne/toekomst-voor-de-zorg CPB Notitie (2013) Schaalgrootte. Digitally retrieved through:
file:///Users/iwanamietus/Downloads/cpb-notitie-4nov2013-schaalgrootte.pdf
Davies, R. and Tracey, D. (2014) Too Big to Be Efficient? The Impact of Implicit Subsidies on Estimates of Scale Economies for Banks. Journal of Money, Credit and Banking, Volume 46, Issue s1, pp. 219-253.
Dziwornu, R.K. et al. (2013) Does Scale Matter in Profitability of Small Scale Broiler Agribusiness Production in Ghana? A Translog Profit Function Model. Journal of Economics and Sustainable Development, Vol. 4, No. 20, p. 128.
Dranove. D. (1998) Economies of scale in non-revenue producing cost centers: implications for hospital mergers. Journal of Health Economics Vol 17 pp. 69-83. Evanston, USA. Hartarska, V., Shen, X., Mersland, R. (2013) Scale economies and input price elasticities in
microfinance institutions. Journal of Banking and Finance, Volume 37, Issue 1, pp. 118-131.
Hughes, J.P. & Mester, L.J., (2013) Who said large banks don’t experience scale economies? Evidence from a risk-return-driven cost function. Journal of Financial Intermediation, Volume 22, Issue 4, pp. 559-585.
Kasman, A. (2012) Cost Efficiency, Scale Economies, and Technological Progress in Turkish Banking. Central Bank Review.
Mason, C.F. (2012) On equilibrium in resource markets with scale economies and stochastic prices. Journal of Environmental Economics and Management, Volume 64, Issue 3, pp. 288-300.
Mintzberg, H., (1983) Structures in Fives: Designing Effective Organizations, PrenticeHall. Mullineaux, D.J. (1978) Economies of Scale and Organizational Efficiency in Banking: A
Profit - Function Approach. The Journal of Finance, Vol. 33, No. 1, pp. 259-280. Pindyck, R.S. & Rubinfield, D.L. (2009) Microeconomics. Pearson International Edition, New Jersy, USA.
Raad voor de Volksgezondheid & Zorg (2008) Schaal en zorg. Digitally retrieved through:
http://www.rvz.net/uploads/docs/Achtergrondstudies_-_Schaal_en_zorg.pdf
Ruiz, J.L., Sirvent, I. (2012) Measuring scale effects in the allocative profit efficiency. Socio-Economic Planning Sciences, Vol. 46, Issue 3, pp. 242-246.
Spencer, G.M. (2011) Creative economies of scale: an agent-based model of creativity and agglomeration. Journal of Economic Geography, Volume 12, Issue 1, pp. 247-271. Tarcan, M., Celik, Y., Plante, C., Younis, M. (2015) An investigaion into the economies of scale in turkish ministry of health hospitals. Journal of Public Budgeting, Accounting and Financial Management, Vol. 27, Issue 1, pp. 98-116.
Thompson, P. (2012) The Relationship between Unit Cost and Cumulative Quantity and the Evidence for Organizational Learning-by-Doing. Journal of Economic Perspectives, 26(3), pp. 203-24.
Wielers, R. and Raven, D. (2013) Parttime work and work norms in the Netherlands. European Sociological Review, Oxford University Press.
Other sources
Berenschot Benchmark Care (2010-2014), a product of management consulting firm Berenschot B.V., located in Utrecht, the Netherlands.
Website mt.nl. Digitally retrieved through: http://www.mt.nl/663/88819/de-mt100-van-2015/dit-zijn-de-beste-bureaus-in-nederland-voor-organisatieadvies.html
