Fiscal rules: the imitation game

ABSTRACT

De jure fiscal rules have known a rapidly increasing popularity worldwide. This paper aims at analysing their spatial diffusion in 108 countries over the period 2001–2015 using a pooled version of the Bayesian SAR probit model. Using two different types of weighting (geographic proximity and bilateral trade) and desegregating the results for specific rules, I find a significant and positive spatial lag in line with the imitation (strategic complementarity) hypothesis. Rational imitation, deriving from a race to fiscal credibility, is preferred over the blind imitation hypothesis as the mimetic behaviour is revealed only in countries facing weaker fiscal reputation.

KEYWORDS:

• Fiscal rules
• policy diffusion
• Bayesian SAR probit model
• imitation
• spatial auto-correlation
• strategic complements

JEL CLASSIFICATION:

• E62
• P48
• D78
• H87

Acknowledgment

I am very grateful to two anonymous referees, as well as Jean-Louis Combes and Alexandru Minea for their valuable comments that allowed for the paper to be significantly improved.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Davoodi et al. (2022).

² See https://blogs.imf.org/2022/01/27/pandemic-tests-resilience-and-credibility-of-fiscal-rules.

³ See https://www.ft.com/content/ecbdd1ad-fcb0–4908-a29a-5a3e14185966.

⁴ It is worth noting the homogeneity of the results in this part of the literature.

⁵ Countries without rating in 2001 are supposed to be part of the lower grade group (BBB+ or lower). This assumption is in line with the shadow ratings determined by Ratha, De, and Mohapatra (2011). The list of classified countries is presented in Table A1.

⁶ See https://gavinr.com/open-data/world-countries-centroids..

⁷ The bilateral trade weighting is also particularly useful to overcome the issue faced by the geographic proximity weighting due to relatively isolated countries after sub-sampling.

⁸ The same procedure is used by Hall, Lacombe, and Tackett (2020) in a Spatial Durbin Model..

⁹ Detailed information on the simulation approach proposed by 2009) adapted to the panel context is given in Caruso, Pontarollo, and Ricciuti (2020).

¹⁰ 3-year period dummies are used instead of 1-year fixed effects to avoid too little variation in the spatial lag across countries that could confound the imitation effect as stated by Davies and Chaitanya Vadlamannati (2013). The inclusion of year dummies would also raise the question of incidental parameter problem inherent to probit models.

¹¹ The coefficients presented in the tables do not correspond to the marginal effects. Indeed, the purpose of this paper is essentially to investigate the presence of strategic complentarity rather than interpreting the amplitude of the coefficients. In addition”,standard spatial autoregressive (SAR) probit models restrict the direct, indirect, and total effects to the same sign” as stated by Hall, Lacombe, and Tackett (2020). Following Caruso, Pontarollo, and Ricciuti (2020), the stationarity of the continuous variables was previously verified using the Levin-Lin-Chu Unit-Root test, the results are available upon request.

¹² These results are in line with those observed in the uncontrolled setting reported in table 6. Naturally, the log-likelihoods are considerably diminished showing the importance of controls previously highlighted in the literature.

¹³ Revenue rules are not considered here as their spatial diffusion is strongly limited with only 11 countries in the sample adopted this type of rule by the end of 2015.

¹⁴ I thank an anonymous referee for suggesting these two additional robustness checks based on the weighting and the lagged value of the spatial lag.

¹⁵ I would like to thank an anonymous referee for this suggestion.

¹⁶ The results of this extended specification of the model are available upon request. The coefficient associated to the dummy variable for EU candidacy was found to be negative and significant. This could be explained by EU countries generally adopting fiscal rules after having joined the club.

¹⁷ Kopits (2001) concludes that countries already benefiting from a strong reputation of fiscal prudence do not need fiscal rules.

¹⁸ These results are available upon demand.

¹⁹ It is interesting to observe that the size of the coefficients is close comparing columns (2) and (4) of Table 2 while they are different comparing columns (2) and (4) of table 10. One might therefore think that mimicking intensity depends on the level of development rather than rating. Nevertheless, the level of development was chosen to proxy fiscal credibility. In this respect, the two countries that moved upward (from lower grade to high-income) are Bahrain and South Korea. Both countries had rapidly increasing ratings over the period 2001–2006 to the extent that their ratings were both superior to the rating of Hungary in 2007 and at least equal to the rating of Botswana in 2008 (the two countries moving downward). Thus, the hypothesis that mimicking intensity depends on the rating could not be ruled out.

²⁰ Results available upon demand.