diff --git a/stdlib/SparseArrays/src/sparsematrix.jl b/stdlib/SparseArrays/src/sparsematrix.jl
index 6a4f7a36c2214..31a18b2a5edd4 100644
--- a/stdlib/SparseArrays/src/sparsematrix.jl
+++ b/stdlib/SparseArrays/src/sparsematrix.jl
@@ -327,17 +327,19 @@ function _sparsesimilar(S::SparseMatrixCSC, ::Type{TvNew}, ::Type{TiNew}) where
     newrowval = copyto!(similar(S.rowval, TiNew), S.rowval)
     return SparseMatrixCSC(S.m, S.n, newcolptr, newrowval, similar(S.nzval, TvNew))
 end
-# parent methods for similar that preserves only storage space (for when new and old dims differ)
+# parent methods for similar that allocates an empty matrix (for when new dims are
+# two-dimensional).
 _sparsesimilar(S::SparseMatrixCSC, ::Type{TvNew}, ::Type{TiNew}, dims::Dims{2}) where {TvNew,TiNew} =
-    SparseMatrixCSC(dims..., fill(one(TiNew), last(dims)+1), similar(S.rowval, TiNew), similar(S.nzval, TvNew))
+    SparseMatrixCSC(dims..., fill(one(TiNew), last(dims)+1), similar(S.rowval, TiNew, 0), similar(S.nzval, TvNew, 0))
+
 # parent method for similar that allocates an empty sparse vector (when new dims are single)
 _sparsesimilar(S::SparseMatrixCSC, ::Type{TvNew}, ::Type{TiNew}, dims::Dims{1}) where {TvNew,TiNew} =
     SparseVector(dims..., similar(S.rowval, TiNew, 0), similar(S.nzval, TvNew, 0))
 #
 # The following methods hook into the AbstractArray similar hierarchy. The first method
 # covers similar(A[, Tv]) calls, which preserve stored-entry structure, and the latter
-# methods cover similar(A[, Tv], shape...) calls, which preserve storage space when the shape
-# calls for a two-dimensional result.
+# methods cover similar(A[, Tv], shape...) calls, with partially preserved structure
+# when the shape calls for a two-dimensional result.
 similar(S::SparseMatrixCSC{<:Any,Ti}, ::Type{TvNew}) where {Ti,TvNew} = _sparsesimilar(S, TvNew, Ti)
 similar(S::SparseMatrixCSC{<:Any,Ti}, ::Type{TvNew}, dims::Union{Dims{1},Dims{2}}) where {Ti,TvNew} =
     _sparsesimilar(S, TvNew, Ti, dims)
diff --git a/stdlib/SparseArrays/src/sparsevector.jl b/stdlib/SparseArrays/src/sparsevector.jl
index 5daaa8d42195c..39baa61ba2697 100644
--- a/stdlib/SparseArrays/src/sparsevector.jl
+++ b/stdlib/SparseArrays/src/sparsevector.jl
@@ -67,22 +67,20 @@ _sparsesimilar(S::SparseVector, ::Type{TvNew}, ::Type{TiNew}) where {TvNew,TiNew
 # parent method for similar that preserves nothing (for when old and new dims differ, and new is 1d)
 _sparsesimilar(S::SparseVector, ::Type{TvNew}, ::Type{TiNew}, dims::Dims{1}) where {TvNew,TiNew} =
     SparseVector(dims..., similar(S.nzind, TiNew, 0), similar(S.nzval, TvNew, 0))
-# parent method for similar that preserves storage space (for old and new dims differ, and new is 2d)
+# parent method for similar with emtpy storage space (for old and new dims differ, and new is 2d)
 _sparsesimilar(S::SparseVector, ::Type{TvNew}, ::Type{TiNew}, dims::Dims{2}) where {TvNew,TiNew} =
-    SparseMatrixCSC(dims..., fill(one(TiNew), last(dims)+1), similar(S.nzind, TiNew), similar(S.nzval, TvNew))
+    SparseMatrixCSC(dims..., fill(one(TiNew), last(dims)+1), similar(S.nzind, TiNew, 0), similar(S.nzval, TvNew, 0))
 # The following methods hook into the AbstractArray similar hierarchy. The first method
 # covers similar(A[, Tv]) calls, which preserve stored-entry structure, and the latter
 # methods cover similar(A[, Tv], shape...) calls, which preserve nothing if the dims
-# specify a SparseVector result and storage space if the dims specify a SparseMatrixCSC result.
+# specify a SparseVector or a SparseMatrixCSC result.
 similar(S::SparseVector{<:Any,Ti}, ::Type{TvNew}) where {Ti,TvNew} =
     _sparsesimilar(S, TvNew, Ti)
 similar(S::SparseVector{<:Any,Ti}, ::Type{TvNew}, dims::Union{Dims{1},Dims{2}}) where {Ti,TvNew} =
     _sparsesimilar(S, TvNew, Ti, dims)
 # The following methods cover similar(A, Tv, Ti[, shape...]) calls, which specify the
 # result's index type in addition to its entry type, and aren't covered by the hooks above.
-# The calls without shape again preserve stored-entry structure, whereas those with
-# one-dimensional shape preserve nothing, and those with two-dimensional shape
-# preserve storage space.
+# The calls without shape again preserve stored-entry structure but no storage space.
 similar(S::SparseVector, ::Type{TvNew}, ::Type{TiNew}) where{TvNew,TiNew} =
     _sparsesimilar(S, TvNew, TiNew)
 similar(S::SparseVector, ::Type{TvNew}, ::Type{TiNew}, dims::Union{Dims{1},Dims{2}}) where {TvNew,TiNew} =
diff --git a/stdlib/SparseArrays/test/sparse.jl b/stdlib/SparseArrays/test/sparse.jl
index 3ef2cef966ae2..d02e5d75b1c5e 100644
--- a/stdlib/SparseArrays/test/sparse.jl
+++ b/stdlib/SparseArrays/test/sparse.jl
@@ -2159,27 +2159,27 @@ end
     @test simA.colptr == A.colptr
     @test simA.rowval == A.rowval
     @test length(simA.nzval) == length(A.nzval)
-    # test similar with Dims{2} specification (preserves storage space only, not stored-entry structure)
+    # test similar with Dims{2} specification (empty storage space, not stored-entry structure)
     simA = similar(A, (6,6))
     @test typeof(simA) == typeof(A)
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.rowval)
-    @test length(simA.nzval) == length(A.nzval)
-    # test similar with entry type and Dims{2} specification (preserves storage space only)
+    @test length(simA.rowval) == 0
+    @test length(simA.nzval) == 0
+    # test similar with entry type and Dims{2} specification (empty storage space)
     simA = similar(A, Float32, (6,6))
     @test typeof(simA) == SparseMatrixCSC{Float32,eltype(A.colptr)}
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.rowval)
-    @test length(simA.nzval) == length(A.nzval)
+    @test length(simA.rowval) == 0
+    @test length(simA.nzval) == 0
     # test similar with entry type, index type, and Dims{2} specification (preserves storage space only)
     simA = similar(A, Float32, Int8, (6,6))
     @test typeof(simA) == SparseMatrixCSC{Float32, Int8}
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.rowval)
-    @test length(simA.nzval) == length(A.nzval)
+    @test length(simA.rowval) == 0
+    @test length(simA.nzval) == 0
     # test similar with Dims{1} specification (preserves nothing)
     simA = similar(A, (6,))
     @test typeof(simA) == SparseVector{eltype(A.nzval),eltype(A.colptr)}
diff --git a/stdlib/SparseArrays/test/sparsevector.jl b/stdlib/SparseArrays/test/sparsevector.jl
index d8ada3963e2c5..a3e5d06efa4f4 100644
--- a/stdlib/SparseArrays/test/sparsevector.jl
+++ b/stdlib/SparseArrays/test/sparsevector.jl
@@ -1229,22 +1229,22 @@ end
     @test typeof(simA) == SparseMatrixCSC{eltype(A.nzval),eltype(A.nzind)}
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.nzind)
-    @test length(simA.nzval) == length(A.nzval)
+    @test length(simA.rowval) == 0 # length(A.nzind)
+    @test length(simA.nzval) == 0 # length(A.nzval)
     # test similar with entry type and Dims{2} specification (preserves storage space only)
     simA = similar(A, Float32, (6,6))
     @test typeof(simA) == SparseMatrixCSC{Float32,eltype(A.nzind)}
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.nzind)
-    @test length(simA.nzval) == length(A.nzval)
+    @test length(simA.rowval) == 0 # length(A.nzind)
+    @test length(simA.nzval) == 0 # length(A.nzval)
     # test similar with entry type, index type, and Dims{2} specification (preserves storage space only)
     simA = similar(A, Float32, Int8, (6,6))
     @test typeof(simA) == SparseMatrixCSC{Float32, Int8}
     @test size(simA) == (6,6)
     @test simA.colptr == fill(1, 6+1)
-    @test length(simA.rowval) == length(A.nzind)
-    @test length(simA.nzval) == length(A.nzval)
+    @test length(simA.rowval) == 0 # length(A.nzind)
+    @test length(simA.nzval) == 0 # length(A.nzval)
 end
 
 @testset "Fast operations on full column views" begin