Why not fully spend a conditional block grant?

For an overview of conditional intergovernmental grants and data on their importance see, e.g., Oates (1999), Bergvall et al. (2006), Huber and Runkel (2006), and Boadway and Shah (2007).

Others study situations in which subnational governments reject a grant because their policy goals interfere with the conditions of the grant (e.g., a grant for abstinence-only education). Another related question is whether subnational governments apply and lobby for grants. See, for example, Chernick (1979), Stein (1979, (1981), Foster (1981), Rich (1989), Schneider and Ji (1990), Knight (2002), Bickers and Stein (2004), Volden (2007), Craw (2010), and Doan and McFarlane (2012). In our setting, however, policy goals of the national and subnational government do not fundamentally differ and the subnational government automatically receives the grant.

See, e.g., Divosa (2006), Van Geuns et al. (2006), and Berkhout and Mallee (2008).

Stein (1979, (1981) mentions three types of costs that subnational governments possibly incur in case they are seeking and receiving a grant: costs of applying for the grant, costs of implementing the grant, and opportunity costs (the costs of curtailing other expenditures due to the costs of application and implementation). Stein hypothesizes that if these costs are high, subnational governments seek grants less frequently. We focus on the costs of using the grant (i.e., the costs of implementing the grant and the opportunity costs) as municipalities do not have to apply for the grant that we study.

In the case of federal funded research grants in the USA, there exists a similar distinction between direct costs (e.g., personnel and equipment during the project) and indirect costs (e.g., university libraries and building maintenance). Direct costs are easy to assign to a specific project, indirect costs of a project are difficult to measure (see, e.g., May and Sarson 1999). In spite of this, federal funded research grants (partly) cover indirect costs. However, it will always be difficult to determine what are legitimate and illegitimate indirect costs. Possibly, this is a reason that many other research grants do not cover indirect costs. In these cases, universities have to pay indirect costs from their own resources. Indirect costs are about 30 % of the total costs of research (Goldman et al. 2000).

Chernick (1979, (1981) also argues that a conditional block grant can implicitly be a matching grant. In his model, recipients apply for a conditional block grant and the donor awards grants to those recipients who are willing to finance a larger part of the project themselves. Therefore, the donor actually provides a matching grant. Chernick also empirically tests this model using the HUD Basic Water and Sewer grant and finds a positive relationship between the size of the grant that recipients receive and their own contribution. Thus, next to implicit costs of using a grant, there can also exist implicit costs of applying for a grant (see also Footnote 4). Also Stein (1979) shows the importance of costs of applying since he finds that communities that apply more often for some specific matching grants have more active and developed planning structures (e.g., a planning body and a high number of planners). For similar evidence see, for example, Collins and Gerber (2006, (2008) and Hall (2008a, (2008b), and also Okten and Weisbrod (2000), Reinhardt (2006), and Lindsay et al. (2014).

Moreover, Becker et al. (2013) show that regions that receive European grants need a certain level of institutional quality and education to absorb grants effectively.

In a similar vein, Cashel-Cordo and Craig (1990) argue that also certain forms of foreign aid and IMF programs look like conditional block grants, but in fact contain an implicit matching component because of, for example, required policy changes that generate additional expenditures for the recipient. See also McGuire (1978), Chubb (1985), Brennan and Pincus (1990), and Singhal (2008) for a discussion on explicit and implicit conditions of grants.

The Canadian government also provides a grant that complements other federal research grants by covering the indirect costs of research (Government of Canada 2010). For the situation in the USA, see, e.g., the references in Footnote 5.

To put this amount in context, in 2008, the total unconditional block grant that municipalities receive for social assistance benefits for people who are younger than 65 equals 3.7 billion euros and the Dutch GDP is 596 billion euros (source: Ministry of Social Affairs and Employment and Statistics Netherlands).

The current system exists since 2004. See, e.g., Ministry of Social Affairs and Employment (2006, (2008) for background information.

Many employees who lose their job can claim unemployment insurance first (the maximum duration of this benefit is individual-specific and varies between 3 months and 5 years during the sample period). Therefore, relatively many social assistance recipients are already unemployed for a longer period.

For an overview of the literature on evaluations of active labor market policies see, e.g., Card et al. (2010) and Kluve (2010). Wunsch (2013) and Lindsay et al. (2014) illustrate the diversity of the used methodologies.

Next to the standard grant from the national government, a municipality may receive additional conditional grants for welfare-to-work programs (e.g., via the European Social Fund). Van Geuns et al. (2006, p. 22) estimate that the size of these grants is 5–6 % of the grant a municipality receives from the national government and that 40–50 % of the municipalities receive such a grant. In our analysis, we assume that the grant of the national government is the only grant a municipality receives (since this is the only grant that we observe).

Each month, the national government transfers one-twelfth of the annual grant to a special checking account. A municipality can transfer money from this account to its regular accounts. The special checking account cannot have a negative balance. If a municipality borrows, then it has to temporarily finance this amount. In our analysis, we assume that a municipality does not receive interest on its savings and that it does not pay interest on its borrowings.

See, e.g., Kok et al. (2006, p. 17), Kuhry and Jonker (2007, p. 89), and Cebeon (2010) for a discussion and estimations of indirect costs (mostly via surveys and balance sheets of a selection of municipalities).

An alternative assumption in this section could be that the national government ignores the indirect costs. This assumption is less realistic and leads sooner to the conclusion that underspending exists. We will come back to this point in Footnote 21.

Survey evidence confirms that municipalities first focus on social assistance recipients with relatively high revenues (Divosa 2006, pp. 15 and 37).

Alternatively, we could assume that the indirect costs depend on the level of U. However, it is a priori not clear what this relation would be. In practice, it seems that some types of indirect costs are higher if a municipality already has spent a lot on actual programs (e.g., it is harder to find a new appropriate participant). Other types of indirect costs, however, may be lower in this case (e.g., civil servants have more experience in the procurement of programs). Furthermore, a more complex cost function substantially increases the required number of computations in the next section. For these reasons, we assume that indirect costs are a constant fraction of U. Another possibility is the existence of (indirect) fixed costs. These do not impact the conclusions in this section. In the empirical part of the paper, one of our estimates is robust to the existence of (indirect) fixed costs.

We could also explicitly divide the indirect costs of programs into two types: indirect costs that also exist if the municipality uses its own resources instead of the grant (e.g., costs of selecting program participants) and indirect costs that only exist if the municipality uses the grant (e.g., administrative costs of proving that money is spent on the purpose of the grant). In this section, we assume for simplicity that all indirect costs are of the first type (i.e., the indirect costs are not higher if the municipality uses the grant than if it uses its own resources). In practice, the first type seems to be much larger than the second type. In the empirical part of the paper, we implicitly include both types of costs.

If we would assume that the national government ignores the indirect costs (see Footnote 17) while these costs do exist, then the national government would underestimate the costs of programs and provide a larger grant. In that case, the municipality already would not fully use the grant at a lower level of x (\(\tilde{x}\) would be lower). Hence, the assumption that the national government ignores the indirect costs leads sooner to the conclusion that underspending exists than the assumption that its estimate of the indirect costs is correct.

The national government can compensate the crowding out of expenditures that the municipality would do if there would not be a grant (\(\left( \beta /(1+x)\right) -1\)) by decreasing the unconditional block grant for general expenditures (see Sect. 2) by this amount.

Possibly, municipalities also spend money from their own resources on actual welfare-to-work programs. We do not observe these possible additional expenditures. However, for a municipality there seems no reason to use its own resources if it has savings or if it can still borrow. Table 1 shows that most municipalities have savings. Moreover, in Sect. 3 we show that in theory a municipality never wants to spend more than the grant. Therefore, we assume that municipalities do not spend money from their own resources on actual programs (compare this assumption with Moffitt 1984).

The annual amounts that a municipality spends, saves, and returns do not necessarily add up to the grant it receives in that year since it may have savings from the past and/or it may borrow from the national government. Hence, in a certain year, a municipality can spend more than the grant it receives. Moreover, a municipality can return more than the grant if its grant decreases compared to the previous year. In that case, it can save less than in the previous year and therefore it is possible that it returns both the new grant and a part of its savings from the past.

Most municipalities have some savings from 2003 (the final year of the previous system), but these are for most municipalities much lower than the maximum allowed amount of savings in 2004.

On the one hand, the total revenues might be higher since a municipality may also provide additional income support (e.g., local tax waivers, discounts on local services, and additional benefits for certain expenditures). We do not take this support into account since it is often for everyone with a low income and a person may still receive it after leaving social assistance. Moreover, the extent of this support is relatively small. On the other hand, the total revenues might be lower since a municipality with fewer than 40,000 inhabitants (of all ages, not only between 20 and 65) might receive in the future a lower unconditional block grant for social assistance benefits from the national government if people leave social assistance. Based on the annual outflow during 2004–2006, the national government can (partially) lower the grant after 3 years and based on the annual outflow during 2007–2008 after 2 years. The exact decrease depends on the number of inhabitants and the interplay with other rules. The national government does not lower the individual grant for a municipality with 40,000 or more inhabitants (see, e.g., Ministry of Social Affairs and Employment 2006; Toolsema and Allers 2014). For simplicity, we do not take this difference between municipalities into account. Non-monetary communal objectives, such as stimulating social inclusion or avoiding paternalism, may also impact the total revenues.

We assume that indirect costs are constant over time to limit the required number of computations. However, we allow indirect costs to vary over time in Sect. 4.4.1. The total costs do not include the costs of actual programs (see Footnote 24).

In addition to implementation costs, the estimated total indirect costs also include opportunity costs. Opportunity costs are the costs of curtailing other expenditures to pay for implementation costs.

For a municipality with a relatively short sample period, the assumption on the value of the savings in the final period is more important for determining its optimal decision.

We only interpret the average value of \(x_{i}^{e}\). Intuitively, we set the marginal costs equal to the marginal revenues for each municipality. However, the effectiveness of spending (\(\beta _{t}\)) is an average revenue over all municipalities. As a result, we underestimate revenues, and therefore indirect costs, for some municipalities, while we overestimate them for others. Therefore, we focus on the average indirect costs. Since the effectiveness of spending has to change over time to explain saving by municipalities, we cannot estimate the effectiveness of spending for individual municipalities (moreover, we have only 5 years).

If we only consider municipalities that have one optimal path that minimizes the expression in Eq. (6), then the estimated indirect costs are in general lower, but do not differ much. For example, in case of \(y=0.5\), \(\bar{x}^{e}\) equals 0.87 (this value is the average over 293 municipalities).

A back-of-the-envelope calculation shows that this estimate is in the same range as the part of the general unconditional block grant that the national government labels for the indirect costs of social assistance and welfare-to-work programs (1.5 billion euros in 2008, see Sect. 2). According to our estimation, in 2008, municipalities need in total about 1.4 billion euros to fully spend the grant of 1.6 billion euros (\(=0.9 \cdot 1.6\)).

The estimated \(x_{i,t}\) is robust to the existence of additional indirect fixed costs since these costs would not have an impact on the marginal profit of spending.

We use groups because we cannot estimate the effectiveness of spending (and thus indirect costs) for individual municipalities (see Footnote 31).

Although large municipalities spend on average 104 % of the grant in 2008, still many large municipalities do not fully use the grant in that year and 23 % even return money to the national government (see also Footnote 25).

The distribution of \(x_{i}^{e}\) shows that the set of possible values of \(x_{i}\) that we use in the main analysis is appropriate for both groups.