Why Counties Spend Less on Development and What we Could do About It

Devolution was a force for good, and while most of us spend invaluable time looking at and even critiquing what the national government does daily, only some are usually interested in what happens in the counties. But the counties are the grassroots, and that’s where many things matter, including how we vote. Most counties in Kenya are rural and ethnically homogenous, meaning tribalism usually doesn’t matter when electing the MCA, MP, Governor, Senator, and Women’s rep. That leaves room for voters to evaluate their options objectively without tribal bias clouding their views.

Before we get ahead of ourselves, here are my observations of what is happening with county finances. First, counties spend too little on development, yet all the signals show they could spend more. Devolution promised to take development to the grassroots, meaning if counties are spending too little on it, we might not be living the dream. Second, counties spend too much on personnel emoluments, or simply salaries and benefits. Nothing wrong with paying people their fair dues. However, this is where we need to be more reasonable. The grassroots would transform tremendously if we could find ways to manage these two problems. Third, the Public Finance Management Act is probably the source of these problems.

The Problem in Development and Personnel Expenditures

A little context is needed at this point. First, the total amount of money received by the counties in the Financial Year 2022/2023 was Ksh 466.01 billion, of which Ksh 370 billion was from the equitable share, Ksh 37.8 billion from own source revenue, Ksh. 42.03 billion cash balance, and Ksh. 15.17 billion from development partners. Overall, the counties spent 83.3% of that amount, or Ksh 428.9 billion. Ksh 330.92 billion went to recurrent expenditure (operations & maintenance + personnel emoluments), and Ksh. 97.98 billion (22.8%) on development. For recurrent expenditure, Ksh. 195.09 billion (45.5%) was spent on personnel emoluments, and Ksh 135.83 billion was spent on operations and maintenance. You could cringe at this point and wonder why close to half of the county government’s expenditure goes to salaries.

More context: The total number of people employed by the county governments in 2019 was 190,949. Assuming an equal number of employees in each county, that’s like 4,063 employees. Obviously, some counties have more personnel than others, and since 2019 more people have been employed. However, assuming nothing changes, that’s 1,021,686 for each employee annually or 85,141 monthly. It could be lower (though not by much) if additional people have been employed since 2019. Other than County Executives and MCAs, I cannot tell who these other employees are, what they do, or what their job group is.

What We Learn from County Finances

I’ll use box plots to communicate the two claims I made above. Nothing big about box plots other than they can tell us how much a majority (50%) of the counties are spending as represented by the colored box (see below) and how much below or above the interquartile range some counties spend, represented by the lines or whiskers and consist of 25% of the counties. The end of the line gives us the maximum and minimum expenditures for each category, excluding outliers. Counties with outlier expenditures fall outside the maximum and minimum values. From the chart below, they are the tiny dots above each plot. The line in the middle of the box is the average amount spent by the counties.

The chart below shows expenditure on development, operations & maintenance (O&M henceforth), and personnel emoluments. Total expenditure consists of O&M plus development plus personnel emoluments, and total recurrent expenditure consists of O&M plus personnel emoluments only. From their means seen below, it’s evident counties spend the lowest on development, then operations, and a significantly higher amount on personnel emoluments (henceforth salaries or wages). Nairobi is the outlier county in all the sections. Its total expenditure is around Ksh 32 billion, and the total recurrent expenditure is Ksh. 28.6 billion. Even though it’s an outlier in development expenditure, it only spends Ksh 4.6 billion on development, which isn’t far from what Mandera spends Ksh 3.7 billion, Kilifi Ksh. 3.4 billion, and Kakamega 3.6 billion. Nairobi also spends Ksh 12.1 billion on salaries, which is an outlier value. The other county with an outlier expenditure on salaries is Kiambu County, spending Ksh. 7.7 billion. From the box plot, most counties spend between 3 and 4.7 billion on personnel emoluments and between 1.5 and 2.6 billion on development. What should we conclude:

1. Nairobi overspends on personnel emoluments and O&M, but its development expenditure is almost equal to that of other counties. This is an interesting paradox because if we treat Nairobi as a unique county with significant spending on everything, we should at least see an exceptional value for development expenditure. But we see something else. The city is conservative with its spending on development but too liberal with other expenses.
2. Most counties spend one to two times more on salaries than development. In most cases, even if you sum development expenditure with O&M, the expenditure on salaries would still be higher. For example, Mombasa County’s spending on both is 5.5 billion, lower than the expenditure on wages at 6.9 billion.

The plot above tells us a lot about the outliers like Nairobi. In the next chart, I removed Nairobi so we can clearly see the other counties. Look at how large the colored box is and how long the whiskers are, showing us the range of the values and how spread out they are. We notice two things: First, the box is tiny for development but larger for personnel expenditure, meaning when it comes to development, spending in all counties does not vary much, and the standard deviation is lower. On the contrary, spending on personnel emoluments varies greatly, and the standard deviation is larger. In fact, even after removing Nairobi, we still have outliers for salaries and O&M but none for development. What do we learn in this instance:

1. A county being large, small, rich, poor, and populous does not matter for development. Regardless of a county’s unique features, most counties will still spend roughly equal amounts on development as other counties. Even the whiskers aren’t very long, further cementing this view. This is interesting because larger, most populous counties, or counties with special needs, are, in theory, required to spend more on development due to their unique issues. Yet they do not do so. On the other hand, counties spend varying amounts on personnel emoluments, which many would argue is because some counties are larger and more populous. Doesn’t this beg the question? Why should these factors matter for personnel emoluments and not development expenditure?
2. Like development, expenditures on operations and maintenance aren’t spread out a lot, and most counties spend, on average, 2.3 billion. This is interesting because, once again, counties in Kenya differ significantly in their characteristics, and you wouldn’t expect lower variability in their operational expenses. An interesting fact is that Turkana and Narok are the two outlier counties in the chart in the middle. You’d expect populous counties like Kiambu and Nakuru to spend a lot on operations. However, they only spend 3 billion and 3.4 billion, respectively.
3. Expenditure on personnel emoluments varies a lot, and the box is large, meaning that even around the mean, values are still widely spread out. The more prolonged whisker (look keenly at the third plot. You notice the bottom whisker is shorter than the one at the top) suggests that the larger spread is further biased toward higher values, indicating that most counties are conservative in spending on development and operations, but quite liberal on personnel emoluments.

We also have a third set of box plots, communicating the same thing but differently. This time around, we are looking at absorption rates, the amount of money counties spend vis a viz what they had budgeted for. In most cases, counties are usually quite ambitious with their budgets, but how closely they commit to these budgets usually tells a different story. The absorption rate tells us how serious a county was in meeting its budget.

The first box plot, for example, shows us the development absorption rate, with a mean of around 64%. The second is the overall absorption rate, which is the percentage of the total budget that was met (around 85%). The third is the recurrent expenditure absorption rate, i.e., the percentage of the overall budget on recurrent expenditure (salaries and O&M) that was met (around 94%). Before highlighting what stands out, notice that the scale goes beyond 100%, meaning some aspects of the budget were oversubscribed. What do we learn?

1. Once again, there’s something weird going on with development expenditure since no county fully met its development budget. West Pokot leads with an 89% absorption rate, a county few would expect. Most counties have an absorption rate of less than 75%, a quarter have absorption rates below 55%, and the minimum value excluding outliers is 36%. Despite their ambitious development budgets, most counties do not meet even half of it.
2. Notice that we still have outliers on the development plot, albeit at the bottom, not at the top. These counties absorbed less than 36% of their budgets, including Kiambu 26.4% and Kisii 13.9%. The better way to put it is that out of the Ksh 100 budgeted for development, Kiambu spent 26 shillings, Kisii 14 shillings, and Nairobi 50 shillings. You can now imagine the effect in the counties if they all met at least 90% of the budgets.
3. The third observation is still within the development plot. Unlike previous charts, when it comes to absorption rates, there’s now more variability in how counties absorb their development budgets than they do for other expenses. The bottom whisker is also longer this time round than the top, telling us that while counties are committed to meeting their budgets on operations and personnel emoluments (see how thin the blue plot is), they don’t have the same commitment to development expenditure. As a result, we see more spread-out values for the development absorption rate but not for the recurrent absorption rate.
4. The recurrent absorption rate is oversubscribed, with Mombasa at 104.9%, Baringo at 100%, Laikipia at 99.4%, and Nairobi at 94.3%. This is interesting because recurrent expenditure includes the money counties need for operations and payment of salaries. Most counties meet their budgets for these expenses because I assume vehicles must be fueled, doctors paid, medicine bought, and utility bills paid. Yet, we do not see a similar zeal to meet the development expenditure.
5. The natural question to ask at this point is, why can’t counties meet their development budgets even though they meet and exceed their budgets on operations and payment of salaries? My partial answer, which might also be the whole answer unless I am missing something, is that counties cannot meet their development budgets unless they somehow cut down on their expenditure on personnel emoluments.

Why Counties Cannot Meet Development Budgets

Laws can sometimes be the source of perverse incentives. The theoretical reason for the failure of most counties to meet their county budgets can be found in the Public Finance Act, which outlines how county funds are to be used. The law makes two mistakes, which I believe is the first time anyone has raised. Please link me to any article that raises the same issues. Here’s how the PFM says the county funds are to be used:

Over the medium term, a minimum of thirty percent of the county government’s budget shall be allocated to the development expenditure.

Then,

The country government’s expenditure on wages and benefits for its public officers shall not exceed a percentage of the county government’s total revenue as prescribed by the County Executive member for finance in regulations and approved by the County Assembly.

The first law is problematic because it incentivizes counties only to allocate 30% of their expenditure on development. In reality, most counties could allocate more if there wasn’t this hard figure telling them how much they should allocate. This problem is quite prominent in economics, as we see with minimum wages and price controls. If you tell someone they need to pay their employees a minimum of Ksh 15,000 per month, they will pay 15,000 unless there’s a good reason to pay more, which many never have.

Another problem with the first law is hyperbolic discounting, which I highlighted in the Coordination Problem. It occurs when counties, for reasons both good and bad, fail to meet the 30% threshold. Failure to meet this threshold becomes an immediate reward to the county executives, pursued at the expense of the county’s future development (Remember, we saw counties spent more than was budgeted on recurrent expenditure). In this case, if a county allocates less than 30% to development, the difference will likely go to salaries and operations (but we’ve already seen they don’t spend a lot on operations either, which leaves us with salaries), which counts as an immediate reward for the county officials. An effect of hyperbolic discounting is that counties that fail to meet the 30% threshold will continue failing to meet it because they love to spend the extra on other things. The chart below is proof that the first law needs modification.

Notice all the counties allocate development just around the 30% mark required by law. More effective would be to cap wages and not development.

The first law explains why counties budget very little on development even though they could allocate more. The second law tells us why counties spend more on salaries even though they could spend less and why development spending cannot keep up. Unlike the first law, the second law does not explicitly state the maximum or minimum amount counties should spend on wages. Instead, it leaves it to the county finance executive and the county assembly to decide how much should be spent on wages. The gross failure here is that the law failed to impose a robust Schelling fence that openly states the maximum percentage that should go on wages. This lack of a fence makes hyperbolic discounting even more attractive, leading counties to spend more on salaries and less on development.

A better law would remove the minimum threshold for development and impose a maximum wage limit. The chart below shows that most counties spend 77.2% of their total expenditure on personnel emoluments and operations and only 22.8% on development. A further breakdown shows that 45.5% is spent on wages and 31.8% on operations. If the law had instead said counties could not spend more than 30% or 35% on wages, they would be forced to spend 65%+ on development and operations, with a bias on development. From the chart below, imagine what would happen if development expenditure in most counties moved from 22.8% to 35% and above. There would be a real effect on the ground, and the presence of the county government would be felt. An unintended consequence of caping the expenditure on wages by assigning a percentage to be spent is that it pushes counties to collect more revenue if they intend to increase their wages. The percentage is a floating metric whose absolute value increases or decreases depending on how much money the county gets. To get more in absolute wages, even as the percentage remains the same, counties would need to collect more revenue or demand more from the national government.

Or probably this is where the 30% mark needed to be considering we’ve seen budgetary allocation doesn’t necessary mean counties will meet those budgets. Caping the development expenditure as a share of the total, would have made more sense.

What about county employees?

If we were to cap how much counties can spend on wages and benefits and remove the lower minimum for development expenditure, some people would then ask what happens to county government employees, who are currently responsible for higher expenditures on wages. A cap would mean wages are reduced or some employees are laid off. Other than some white color professions like doctors, who are few and could be paid whatever they want, most county functions, like collecting garbage and keeping clean environments, do not require specialized personnel. It wouldn’t hurt to pay them the market rate. We don’t see many county employees dealing with everyday problems like garbage collection because more expenditure on wages goes toward MCA sitting allowances (which is unnecessary) and irregular payments to unknown entities. Restricting wages means a lot more low-cost workers are hired. Because counties have enough for operations and maintenance, they’ll have more than enough money to deal with the essential functions of the county. Read the auditor general’s report for details on how counties misappropriate salaries.