Estimation of the price of natural gas in Mexico using a decision tree model

Mexico liberalized natural gas prices in 2017 by establishing the National Wholesale Natural Gas Price Index (IPGN). This institutional change created observable conditions where prices could deviate from efficiency. We classified these deviations into three regimes: stability, overestimation and underestimation. To do this, we use a multinomial decision tree with monthly data from 2000 to 2025. The main findings are: First, time validation achieves an accuracy of 73.40 %. In this case, Henry Hub futures prices explain 73.7 % of the regime variance, confirming that price formation in Mexico responds primarily to North American market expectations, rather than domestic fundamentals. Second, markets operate in stability 41 % of the time, being correctly identified in 82.75 % of cases, while overestimation and underestimation represent temporary deviations driven by expectations, rather than a chronic dysfunction. In terms of energy policy, a drop in Henry Hub futures below −6.3 % could trigger regime transitions, thus warranting increased monitoring. The proposal provides high accuracy in identifying stability allows regulators to distinguish crises requiring intervention from normal volatility, thereby avoiding disproportionate policy reactions. Finally, it is worth noting that the interpretability of the decision tree provides essential monitoring thresholds for infrastructure investors that commit resources over the long term (up to ten years). • First multinomial decision tree model classifying Mexican natural gas price regimes. • Model achieves 73.40 % accuracy identifying stability, overestimation, underestimation. • Henry Hub futures dominate Mexican pricing with 73.7 % explanatory importance. • Critical threshold at −6.3 % Henry Hub futures signals regime transitions. • Tool provides early warnings for energy policy intervention in emerging markets.