Can Bureaucrats Really Be Paid Like Ceos? Substitution Between Incentives and Resources Among School Administrators in China

Abstract Unlike performance incentives for private sector managers, little is known about performance incentives for managers in public sector bureaucracies. Through a randomized trial in rural China, we study performance incentives rewarding school administrators for reducing student anemia—as well as complementarity between incentives and orthogonally assigned discretionary resources. Large (but not small) incentives and unrestricted grants both reduced anemia, but incentives were more cost-effective. Although unrestricted grants and small incentives do not interact, grants fully crowd-out the effect of larger incentives. Our findings suggest that performance incentives can be effective in bureaucratic environments, but they are not complementary to discretionary resources.

NOTES. Table uses sample of children testing anemic at baseline. Children are considered anemic if they have an altitude-adjusted hemoglobin concentration below 120 g/L (per WHO guidelines). The dependent variable in columns 1 and 2 is a dummy variable indicating missing hemoglobin measurements at endline. The dependent variable in columns 3 and 4 is a dummy variable indicating missing household forms at endline conditional on a child's hemoglobin measurement being non-missing. In addition to what is shown regressions also control for county and randomization strata fixed effects. Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference are shown in curly brackets. Adjusted pvalues were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1%. based on adjusted p-values.   (12) (controlling for baseline hemoglobin concentration, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1%. based on adjusted p-values. Panel B shows unadjusted and adjusted p-values from tests between coefficients.   (12) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). The dependent variable in each regression is a summary index constructed using the GLS weighting procedure in Anderson (2008). Estimates for the individual components of each index are shown in Appendix Tables 5 and 6. Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values.   Table uses sample of children testing anemic at baseline. Children are considered anemic if they have an altitude-adjusted hemoglobin concentration below 120 g/L (per WHO guidelines). Rows 1-5 show estimated coefficients for treatment group indicators and interactions obtained by estimating equation (12) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference across all tests corresponding to each index are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values.

Supplements
Food Food at School Food at Home   (12) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference across all tests corresponding to each index are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values.
(1) B1: Small Incentive (2) B2: Large Incentive (3) B3: Large Grant NOTES. Table uses sample of children testing anemic at baseline. Children are considered anemic if they have an altitude-adjusted hemoglobin concentration below 120 g/L (per WHO guidelines). Rows 1-5 show estimated coefficients for treatment group indicators and interactions obtained by estimating equation (12) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference across all tests corresponding to each index are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values.
(2) B2: Large Incentive   (12) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference across all tests corresponding to each index are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values.  (2) B2: Large Incentive  NOTES. Table uses sample of children testing anemic at baseline. Children are considered anemic if they have an altitude-adjusted hemoglobin concentration below 120 g/L (per WHO guidelines). Rows 1-5 show estimated coefficients for treatment group indicators and interactions obtained by estimating equation (19) (controlling for the baseline value of the dependent variable, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). The dependent variable in each regression is a summary index constructed using the GLS weighting procedure in Anderson (2008). Estimates for the individual components of each index are shown in Appendix Tables 5 and 6. Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference are shown in curly brackets. Adjusted p-values were constructed using the free step-down resampling method of Westfall and Young (1993) (2008) with individual responses corresponding to each scale combined into indices using the GLS weighting procedure in Anderson (2008). Administrator scores are categorized as low if scores are below the median in the sample at baseline and high if they are above the median. Heterogeneous effects are estimated using equation (19) and interacting treatment dummies with indicators for high baseline prosocial or intrinsic motivation (and controlling for baseline hemoglobin concentration, student age, student grade, student sex, number of students in the school, whether the school has a canteen, student teacher ratio, distance to the furthest village served, percent of boarding students, whether the school has implemented the "Free Lunch" policy, county dummy variables, and dummy variables for randomization strata). Standard errors are shown in parentheses, unadjusted p-values are shown in square brackets and p-values adjusted for multiple inference are shown in curly brackets. Adjusted pvalues were constructed using the free step-down resampling method of Westfall and Young (1993) with 10,000 iterations. *, **, and *** indicate significance at 10%, 5% and 1% based on adjusted p-values. For simplicity of notation, we suppress the arguments of the functions in the following, so that f iij refers to f iij (e, A) for i, j = {e, A}, v 0 refers to v 0 (e), S 0 refers to S 0 (G A) and similar for the second and third order derivatives of v and S.
The second order conditions are: And di↵erentiating both equations with respect to t, we obtain: U ee @e @t pplying Cramer's Rule to the above system of equations, we obtain: Equivalently but di↵erentiating B.3 and B.4 with respect to G, we obtain: U Ae @e @G And correspondingly, using Cramer's Rule: Note that the numerator of @A @G can be written as: anemia, their grades in school have been shown to increase by up to five points! This is the difference between a C+ (75 points out of a 100) and a B (80 points or more)! There is some Good News, however. All of these problems can be easily fixed, just by treating anemia in your school.
Researchers have found that around one-third of children in your area have anemia. If you suspect that some students in your school might be anemic, don't worryanemia is easy to treat! To fight and prevent anemia-and improve student performance in class-you can just make a few simple changes in your students' diet to make sure they are getting enough iron and eating a balanced diet.
What foods have iron? MEAT. Chicken, beef, lamb and pork-all animal meats are great sources of iron. Eggs and milk are healthy foods, but they don't have any iron! Some other foods have iron too, but it is harder for children to absorb iron from non-meat sources. Meat is the best source of iron. Other sources of iron are pumpkin or squash seeds, potato skins, tofu and soy products, peanuts, and beans. If you want to add these foods to your students' diet, make sure you also add lots of fruits and vegetables so that the children can absorb the iron into their bodies. Vegetables like green peppers, chili peppers, and red dates are all good sources of vitamin C, which children need to absorb iron. Did you know that adding a red date or other source of vitamin C to your students' meal can boost their rate of iron absorption by as much as 300%?
Eating a balanced diet each day is important for overcoming anemia, so feed your child lots of meat and vegetables in addition to staples like plain rice, noodles, and buns.
Eating staples may make children feel full, but they don't contain any iron and have few In addition to supplementing school meals with meat, vegetables and fruit, there are other things you can do to reduce anemia in your schools. Some of these options might be cheaper or easier than buying meat everyday, and can be just as effective!
One option is to give students a daily multivitamin with iron. A single vitamin tablet has as much iron as a whole dish of pork! Vitamins are easy to dispense -other principals in your area have tried it and found that students in their schools became more energetic and performed better in school. A daily multivitamin helps prevent anemia and especially helps children who are already anemic and most at-risk for learning problems.
Follow the recommended dosages for children on the bottle, and be sure to provide clean drinking water for students to use when taking vitamins. Many manufacturers now make chewable vitamins for children. These are easy to take, don't require any water, and kids often enjoy the taste! Vitamins may not be readily available at your local grocery store, but we can help you purchase them easily.
Another way to reduce anemia in your school is to cook school meals with foods that are fortified with iron. Some special types of soy sauce and flour have extra iron added.
These are cheap, and taste the same as regular soy sauce and flour. Fortified foods are often difficult to find at your local grocery store, but we can help you purchase them easily.
You may also choose to tell parents about anemia and why it is important. Tell them that their children might not be doing as well in school as they should be, and share the strategies you have learned to reduce anemia. Explain the importance of a balanced diet. Persistent efforts from teachers and principals to talk to parents is very important, and there are many successful examples of this. Most parents are very concerned about the well-being of their children. However, it is not an easy task to convey key nutrition and health information to parents! Giving parents pamphlets or letters about anemia and how to prevent anemia is not always effective. Some parents may not receive the information, some may not understand it, and some may think they do not have the economic means to change their dietary habits.
Instead, try holding meetings at school with parents to communicate information directly. Frequently follow up to check in about progress and to answer any questions parents might have.
Remember that the best strategy for reducing anemia will depend on local conditions at your school-and no one knows your school better than you do! Can you think of other ways to reduce anemia in your school? Be creative! You know your school and students best, so you may be able to think of even better strategies for keeping your students healthy.
Why is it so important to address the anemia problem in your school? Well, researchers in Shaanxi, Ningxia, Gansu and Qinghai found that anemic children who took a daily multivitamin with iron saw a big improvement in their test scores. This was