Several related computational complexity classes have similar properties to EFA:
*One can omit the binary function symbol exp from the language, by taking Robinson arithmetic together with induction for all formulas with bounded quantifiers and an axiom stating roughly that exponentiation is a function defined everywhere. This is similar to EFA and has the same proof theoretic strength, but is more cumbersome to work with.
*There are weak fragments of second-order arithmetic called <math>\mathsf{RCA{{su|p=*|b=0}}_0^*</math> and WKL{<math>\mathsf{su|p=WKL}_0^*|b=</math> that are conservative over EFA for <math>\Pi_2^0</math> sentences (i.e. any <math>\Pi_2^0</math> sentences proven by <math>\mathsf{RCA}_0^*</math> or <math>\mathsf{WKL}_0^*</math> thatare havealready theproven sameby consistencyEFA.)<ref>S. strengthG. asSimpson, EFAR. L. Smith, "[https://www.sciencedirect.com/science/article/pii/0168007286900746 Factorization of polynomials and are<math>\Sigma_1^0</math>-induction]" conservative(1986). overAnnals itof forPure Π{{su|and Applied Logic, vol. 31 (p=0|b=2}}.305)</ref> sentences{{explain|date=NovemberIn 2017}}particular, whichthey are conservative for consistency statements. These fragments are sometimes studied in [[reverse mathematics]] {{harv|Simpson|2009}}.
*'''Elementary recursive arithmetic''' ('''ERA''') is a subsystem of [[primitive recursive arithmetic]] (PRA) in which recursion is restricted to [[ELEMENTARY#Definition|bounded sums and products]]. This also has the same Π{{su|p=<math>\Pi_2^0|b=2}}</math> sentences as EFA, in the sense that whenever EFA proves ∀x∃y ''P''(''x'',''y''), with ''P'' quantifier-free, ERA proves the open formula ''P''(''x'',''T''(''x'')), with ''T'' a term definable in ERA. Like PRA, ERA can be defined in an entirely logic-free{{clarify|date=November 2017}} manner, with just the rules of substitution and induction, and defining equations for all elementary recursive functions. Unlike PRA, however, the elementary recursive functions can be characterized by the closure under composition and projection of a ''finite'' number of basis functions, and thus only a finite number of defining equations are needed.
