diff --git a/.travis.yml b/.travis.yml
index b219b03e0eb..f74a3b205c4 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -14,6 +14,7 @@ addons:
     - valgrind
     - libboost-dev
     - libboost-filesystem-dev
+    - libboost-regex-dev
     - libboost-system-dev
     - libjemalloc-dev
     - gtk-doc-tools
diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index 9947a34e4e7..5852fe59da0 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -398,30 +398,36 @@ if (ARROW_BOOST_USE_SHARED)
     add_definitions(-DBOOST_ALL_DYN_LINK)
   endif()
 
-  find_package(Boost COMPONENTS system filesystem REQUIRED)
+  find_package(Boost COMPONENTS system filesystem regex REQUIRED)
   if ("${CMAKE_BUILD_TYPE}" STREQUAL "DEBUG")
     set(BOOST_SHARED_SYSTEM_LIBRARY ${Boost_SYSTEM_LIBRARY_DEBUG})
     set(BOOST_SHARED_FILESYSTEM_LIBRARY ${Boost_FILESYSTEM_LIBRARY_DEBUG})
+    set(BOOST_SHARED_REGEX_LIBRARY ${Boost_REGEX_LIBRARY_DEBUG})
   else()
     set(BOOST_SHARED_SYSTEM_LIBRARY ${Boost_SYSTEM_LIBRARY_RELEASE})
     set(BOOST_SHARED_FILESYSTEM_LIBRARY ${Boost_FILESYSTEM_LIBRARY_RELEASE})
+    set(BOOST_SHARED_REGEX_LIBRARY ${Boost_REGEX_LIBRARY_RELEASE})
   endif()
   set(BOOST_SYSTEM_LIBRARY boost_system_shared)
   set(BOOST_FILESYSTEM_LIBRARY boost_filesystem_shared)
+  set(BOOST_REGEX_LIBRARY boost_regex_shared)
 else()
   # Find static boost headers and libs
   # TODO Differentiate here between release and debug builds
   set(Boost_USE_STATIC_LIBS ON)
-  find_package(Boost COMPONENTS system filesystem REQUIRED)
+  find_package(Boost COMPONENTS system filesystem regex REQUIRED)
   if ("${CMAKE_BUILD_TYPE}" STREQUAL "DEBUG")
     set(BOOST_STATIC_SYSTEM_LIBRARY ${Boost_SYSTEM_LIBRARY_DEBUG})
     set(BOOST_STATIC_FILESYSTEM_LIBRARY ${Boost_FILESYSTEM_LIBRARY_DEBUG})
+    set(BOOST_STATIC_REGEX_LIBRARY ${Boost_REGEX_LIBRARY_DEBUG})
   else()
     set(BOOST_STATIC_SYSTEM_LIBRARY ${Boost_SYSTEM_LIBRARY_RELEASE})
     set(BOOST_STATIC_FILESYSTEM_LIBRARY ${Boost_FILESYSTEM_LIBRARY_RELEASE})
+    set(BOOST_STATIC_REGEX_LIBRARY ${Boost_REGEX_LIBRARY_RELEASE})
   endif()
   set(BOOST_SYSTEM_LIBRARY boost_system_static)
   set(BOOST_FILESYSTEM_LIBRARY boost_filesystem_static)
+  set(BOOST_REGEX_LIBRARY boost_regex_static)
 endif()
 
 message(STATUS "Boost include dir: " ${Boost_INCLUDE_DIRS})
@@ -435,7 +441,11 @@ ADD_THIRDPARTY_LIB(boost_filesystem
     STATIC_LIB "${BOOST_STATIC_FILESYSTEM_LIBRARY}"
     SHARED_LIB "${BOOST_SHARED_FILESYSTEM_LIBRARY}")
 
-SET(ARROW_BOOST_LIBS boost_system boost_filesystem)
+ADD_THIRDPARTY_LIB(boost_regex
+        STATIC_LIB "${BOOST_STATIC_REGEX_LIBRARY}"
+        SHARED_LIB "${BOOST_SHARED_REGEX_LIBRARY}")
+
+SET(ARROW_BOOST_LIBS boost_system boost_filesystem boost_regex)
 
 include_directories(SYSTEM ${Boost_INCLUDE_DIR})
 
@@ -695,14 +705,16 @@ endif()
 set(ARROW_MIN_TEST_LIBS
   arrow_static
   arrow_test_main
-  ${ARROW_BASE_LIBS})
+  ${ARROW_BASE_LIBS}
+  ${BOOST_REGEX_LIBRARY})
 
 set(ARROW_TEST_LINK_LIBS ${ARROW_MIN_TEST_LIBS})
 
 set(ARROW_BENCHMARK_LINK_LIBS
   arrow_static
   arrow_benchmark_main
-  ${ARROW_BASE_LIBS})
+  ${ARROW_BASE_LIBS}
+  ${BOOST_REGEX_LIBRARY})
 
 ############################################################
 # "make ctags" target
@@ -796,7 +808,7 @@ endif()
 ############################################################
 
 set(ARROW_LINK_LIBS
-)
+    ${BOOST_REGEX_LIBRARY})
 
 set(ARROW_PRIVATE_LINK_LIBS
 )
@@ -816,6 +828,7 @@ set(ARROW_SRCS
   src/arrow/visitor.cc
 
   src/arrow/util/bit-util.cc
+  src/arrow/util/decimal.cc
 )
 
 if(NOT APPLE AND NOT MSVC)
@@ -825,9 +838,11 @@ if(NOT APPLE AND NOT MSVC)
   set(ARROW_SHARED_LINK_FLAGS "-Wl,--version-script=${CMAKE_CURRENT_SOURCE_DIR}/src/arrow/symbols.map")
 endif()
 
+
 ADD_ARROW_LIB(arrow
   SOURCES ${ARROW_SRCS}
   SHARED_LINK_FLAGS ${ARROW_SHARED_LINK_FLAGS}
+  SHARED_LINK_LIBS ${ARROW_LINK_LIBS}
 )
 
 add_subdirectory(src/arrow)
diff --git a/cpp/cmake_modules/FindPythonLibsNew.cmake b/cpp/cmake_modules/FindPythonLibsNew.cmake
index dfe5661b015..d9cc4b39557 100644
--- a/cpp/cmake_modules/FindPythonLibsNew.cmake
+++ b/cpp/cmake_modules/FindPythonLibsNew.cmake
@@ -175,7 +175,8 @@ else()
     find_library(PYTHON_LIBRARY
         NAMES "python${PYTHON_LIBRARY_SUFFIX}"
         PATHS ${_PYTHON_LIBS_SEARCH}
-        NO_SYSTEM_ENVIRONMENT_PATH)
+        NO_SYSTEM_ENVIRONMENT_PATH
+        NO_CMAKE_SYSTEM_PATH)
     message(STATUS "Found Python lib ${PYTHON_LIBRARY}")
 endif()
 
diff --git a/cpp/src/arrow/array-decimal-test.cc b/cpp/src/arrow/array-decimal-test.cc
index b64023bbc6a..4c01f928a6f 100644
--- a/cpp/src/arrow/array-decimal-test.cc
+++ b/cpp/src/arrow/array-decimal-test.cc
@@ -15,13 +15,16 @@
 // specific language governing permissions and limitations
 // under the License.
 
+#include "arrow/type.h"
 #include "gtest/gtest.h"
 
-#include "arrow/type.h"
+#include "arrow/builder.h"
+#include "arrow/test-util.h"
+#include "arrow/util/decimal.h"
 
 namespace arrow {
 
-TEST(TypesTest, TestDecimalType) {
+TEST(TypesTest, TestDecimal32Type) {
   DecimalType t1(8, 4);
 
   ASSERT_EQ(t1.type, Type::DECIMAL);
@@ -29,6 +32,193 @@ TEST(TypesTest, TestDecimalType) {
   ASSERT_EQ(t1.scale, 4);
 
   ASSERT_EQ(t1.ToString(), std::string("decimal(8, 4)"));
+
+  // Test properties
+  ASSERT_EQ(t1.byte_width(), 4);
+  ASSERT_EQ(t1.bit_width(), 32);
 }
 
+TEST(TypesTest, TestDecimal64Type) {
+  DecimalType t1(12, 5);
+
+  ASSERT_EQ(t1.type, Type::DECIMAL);
+  ASSERT_EQ(t1.precision, 12);
+  ASSERT_EQ(t1.scale, 5);
+
+  ASSERT_EQ(t1.ToString(), std::string("decimal(12, 5)"));
+
+  // Test properties
+  ASSERT_EQ(t1.byte_width(), 8);
+  ASSERT_EQ(t1.bit_width(), 64);
+}
+
+TEST(TypesTest, TestDecimal128Type) {
+  DecimalType t1(27, 7);
+
+  ASSERT_EQ(t1.type, Type::DECIMAL);
+  ASSERT_EQ(t1.precision, 27);
+  ASSERT_EQ(t1.scale, 7);
+
+  ASSERT_EQ(t1.ToString(), std::string("decimal(27, 7)"));
+
+  // Test properties
+  ASSERT_EQ(t1.byte_width(), 16);
+  ASSERT_EQ(t1.bit_width(), 128);
+}
+
+template <typename T>
+class DecimalTestBase {
+ public:
+  virtual std::vector<uint8_t> data(
+      const std::vector<T>& input, size_t byte_width) const = 0;
+
+  void test(int precision, const std::vector<T>& draw,
+      const std::vector<uint8_t>& valid_bytes,
+      const std::vector<uint8_t>& sign_bitmap = {}, int64_t offset = 0) const {
+    auto type = std::make_shared<DecimalType>(precision, 4);
+    int byte_width = type->byte_width();
+    auto pool = default_memory_pool();
+    auto builder = std::make_shared<DecimalBuilder>(pool, type);
+    size_t null_count = 0;
+
+    size_t size = draw.size();
+    builder->Reserve(size);
+
+    for (size_t i = 0; i < size; ++i) {
+      if (valid_bytes[i]) {
+        builder->Append(draw[i]);
+      } else {
+        builder->AppendNull();
+        ++null_count;
+      }
+    }
+
+    std::shared_ptr<Buffer> expected_sign_bitmap;
+    if (!sign_bitmap.empty()) {
+      BitUtil::BytesToBits(sign_bitmap, &expected_sign_bitmap);
+    }
+
+    auto raw_bytes = data(draw, byte_width);
+    auto expected_data = std::make_shared<Buffer>(raw_bytes.data(), size * byte_width);
+    auto expected_null_bitmap = test::bytes_to_null_buffer(valid_bytes);
+    int64_t expected_null_count = test::null_count(valid_bytes);
+    auto expected = std::make_shared<DecimalArray>(type, size, expected_data,
+        expected_null_bitmap, expected_null_count, offset, expected_sign_bitmap);
+
+    std::shared_ptr<Array> out;
+    ASSERT_OK(builder->Finish(&out));
+    ASSERT_TRUE(out->Equals(*expected));
+  }
+};
+
+template <typename T>
+class DecimalTest : public DecimalTestBase<T> {
+ public:
+  std::vector<uint8_t> data(
+      const std::vector<T>& input, size_t byte_width) const override {
+    std::vector<uint8_t> result;
+    result.reserve(input.size() * byte_width);
+    // TODO(phillipc): There's probably a better way to do this
+    constexpr static const size_t bytes_per_element = sizeof(T);
+    for (size_t i = 0, j = 0; i < input.size(); ++i, j += bytes_per_element) {
+      *reinterpret_cast<typename T::value_type*>(&result[j]) = input[i].value;
+    }
+    return result;
+  }
+};
+
+template <>
+class DecimalTest<Decimal128> : public DecimalTestBase<Decimal128> {
+ public:
+  std::vector<uint8_t> data(
+      const std::vector<Decimal128>& input, size_t byte_width) const override {
+    std::vector<uint8_t> result;
+    result.reserve(input.size() * byte_width);
+    constexpr static const size_t bytes_per_element = 16;
+    for (size_t i = 0; i < input.size(); ++i) {
+      uint8_t stack_bytes[bytes_per_element] = {0};
+      uint8_t* bytes = stack_bytes;
+      bool is_negative;
+      ToBytes(input[i], &bytes, &is_negative);
+
+      for (size_t i = 0; i < bytes_per_element; ++i) {
+        result.push_back(bytes[i]);
+      }
+    }
+    return result;
+  }
+};
+
+class Decimal32BuilderTest : public ::testing::TestWithParam<int>,
+                             public DecimalTest<Decimal32> {};
+
+class Decimal64BuilderTest : public ::testing::TestWithParam<int>,
+                             public DecimalTest<Decimal64> {};
+
+class Decimal128BuilderTest : public ::testing::TestWithParam<int>,
+                              public DecimalTest<Decimal128> {};
+
+TEST_P(Decimal32BuilderTest, NoNulls) {
+  int precision = GetParam();
+  std::vector<Decimal32> draw = {
+      Decimal32(1), Decimal32(2), Decimal32(2389), Decimal32(4), Decimal32(-12348)};
+  std::vector<uint8_t> valid_bytes = {true, true, true, true, true};
+  this->test(precision, draw, valid_bytes);
+}
+
+TEST_P(Decimal64BuilderTest, NoNulls) {
+  int precision = GetParam();
+  std::vector<Decimal64> draw = {
+      Decimal64(1), Decimal64(2), Decimal64(2389), Decimal64(4), Decimal64(-12348)};
+  std::vector<uint8_t> valid_bytes = {true, true, true, true, true};
+  this->test(precision, draw, valid_bytes);
+}
+
+TEST_P(Decimal128BuilderTest, NoNulls) {
+  int precision = GetParam();
+  std::vector<Decimal128> draw = {
+      Decimal128(1), Decimal128(-2), Decimal128(2389), Decimal128(4), Decimal128(-12348)};
+  std::vector<uint8_t> valid_bytes = {true, true, true, true, true};
+  std::vector<uint8_t> sign_bitmap = {false, true, false, false, true};
+  this->test(precision, draw, valid_bytes, sign_bitmap);
+}
+
+TEST_P(Decimal32BuilderTest, WithNulls) {
+  int precision = GetParam();
+  std::vector<Decimal32> draw = {
+      Decimal32(1), Decimal32(2), Decimal32(-1), Decimal32(4), Decimal32(-1)};
+  std::vector<uint8_t> valid_bytes = {true, true, false, true, false};
+  this->test(precision, draw, valid_bytes);
+}
+
+TEST_P(Decimal64BuilderTest, WithNulls) {
+  int precision = GetParam();
+  std::vector<Decimal64> draw = {
+      Decimal64(-1), Decimal64(2), Decimal64(-1), Decimal64(4), Decimal64(-1)};
+  std::vector<uint8_t> valid_bytes = {true, true, false, true, false};
+  this->test(precision, draw, valid_bytes);
+}
+
+TEST_P(Decimal128BuilderTest, WithNulls) {
+  int precision = GetParam();
+  std::vector<Decimal128> draw = {Decimal128(1), Decimal128(2), Decimal128(-1),
+      Decimal128(4), Decimal128(-1), Decimal128(1), Decimal128(2),
+      Decimal128("230342903942.234234"), Decimal128("-23049302932.235234")};
+  std::vector<uint8_t> valid_bytes = {
+      true, true, false, true, false, true, true, true, true};
+  std::vector<uint8_t> sign_bitmap = {
+      false, false, false, false, false, false, false, false, true};
+  this->test(precision, draw, valid_bytes, sign_bitmap);
+}
+
+INSTANTIATE_TEST_CASE_P(Decimal32BuilderTest, Decimal32BuilderTest,
+    ::testing::Range(
+        DecimalPrecision<int32_t>::minimum, DecimalPrecision<int32_t>::maximum));
+INSTANTIATE_TEST_CASE_P(Decimal64BuilderTest, Decimal64BuilderTest,
+    ::testing::Range(
+        DecimalPrecision<int64_t>::minimum, DecimalPrecision<int64_t>::maximum));
+INSTANTIATE_TEST_CASE_P(Decimal128BuilderTest, Decimal128BuilderTest,
+    ::testing::Range(
+        DecimalPrecision<int128_t>::minimum, DecimalPrecision<int128_t>::maximum));
+
 }  // namespace arrow
diff --git a/cpp/src/arrow/array.cc b/cpp/src/arrow/array.cc
index bd20654bc87..4e73e7176fa 100644
--- a/cpp/src/arrow/array.cc
+++ b/cpp/src/arrow/array.cc
@@ -27,6 +27,7 @@
 #include "arrow/status.h"
 #include "arrow/type_traits.h"
 #include "arrow/util/bit-util.h"
+#include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 #include "arrow/visitor.h"
 #include "arrow/visitor_inline.h"
@@ -283,10 +284,8 @@ std::shared_ptr<Array> StringArray::Slice(int64_t offset, int64_t length) const
 FixedSizeBinaryArray::FixedSizeBinaryArray(const std::shared_ptr<DataType>& type,
     int64_t length, const std::shared_ptr<Buffer>& data,
     const std::shared_ptr<Buffer>& null_bitmap, int64_t null_count, int64_t offset)
-    : PrimitiveArray(type, length, data, null_bitmap, null_count, offset) {
-  DCHECK(type->type == Type::FIXED_SIZE_BINARY);
-  byte_width_ = static_cast<const FixedSizeBinaryType&>(*type).byte_width();
-}
+    : PrimitiveArray(type, length, data, null_bitmap, null_count, offset),
+      byte_width_(static_cast<const FixedSizeBinaryType&>(*type).byte_width()) {}
 
 std::shared_ptr<Array> FixedSizeBinaryArray::Slice(int64_t offset, int64_t length) const {
   ConformSliceParams(offset_, length_, &offset, &length);
@@ -294,6 +293,48 @@ std::shared_ptr<Array> FixedSizeBinaryArray::Slice(int64_t offset, int64_t lengt
       type_, length, data_, null_bitmap_, kUnknownNullCount, offset);
 }
 
+const uint8_t* FixedSizeBinaryArray::GetValue(int64_t i) const {
+  return raw_data_ + (i + offset_) * byte_width_;
+}
+
+// ----------------------------------------------------------------------
+// Decimal
+DecimalArray::DecimalArray(const std::shared_ptr<DataType>& type, int64_t length,
+    const std::shared_ptr<Buffer>& data, const std::shared_ptr<Buffer>& null_bitmap,
+    int64_t null_count, int64_t offset, const std::shared_ptr<Buffer>& sign_bitmap)
+    : FixedSizeBinaryArray(type, length, data, null_bitmap, null_count, offset),
+      sign_bitmap_(sign_bitmap),
+      sign_bitmap_data_(sign_bitmap != nullptr ? sign_bitmap->data() : nullptr) {}
+
+bool DecimalArray::IsNegative(int64_t i) const {
+  return sign_bitmap_data_ != nullptr ? BitUtil::GetBit(sign_bitmap_data_, i) : false;
+}
+
+template <typename T>
+ARROW_EXPORT Decimal<T> DecimalArray::Value(int64_t i) const {
+  Decimal<T> result;
+  FromBytes(GetValue(i), &result);
+  return result;
+}
+
+template ARROW_EXPORT Decimal32 DecimalArray::Value(int64_t i) const;
+template ARROW_EXPORT Decimal64 DecimalArray::Value(int64_t i) const;
+
+template <>
+ARROW_EXPORT Decimal128 DecimalArray::Value(int64_t i) const {
+  Decimal128 result;
+  FromBytes(GetValue(i), IsNegative(i), &result);
+  return result;
+}
+
+template ARROW_EXPORT Decimal128 DecimalArray::Value(int64_t i) const;
+
+std::shared_ptr<Array> DecimalArray::Slice(int64_t offset, int64_t length) const {
+  ConformSliceParams(offset_, length_, &offset, &length);
+  return std::make_shared<DecimalArray>(
+      type_, length, data_, null_bitmap_, kUnknownNullCount, offset, sign_bitmap_);
+}
+
 // ----------------------------------------------------------------------
 // Struct
 
diff --git a/cpp/src/arrow/array.h b/cpp/src/arrow/array.h
index 9f0e73914da..a4117facdef 100644
--- a/cpp/src/arrow/array.h
+++ b/cpp/src/arrow/array.h
@@ -39,6 +39,9 @@ class MemoryPool;
 class MutableBuffer;
 class Status;
 
+template <typename T>
+struct Decimal;
+
 /// Immutable data array with some logical type and some length.
 ///
 /// Any memory is owned by the respective Buffer instance (or its parents).
@@ -356,9 +359,7 @@ class ARROW_EXPORT FixedSizeBinaryArray : public PrimitiveArray {
       const std::shared_ptr<Buffer>& null_bitmap = nullptr, int64_t null_count = 0,
       int64_t offset = 0);
 
-  const uint8_t* GetValue(int64_t i) const {
-    return raw_data_ + (i + offset_) * byte_width_;
-  }
+  const uint8_t* GetValue(int64_t i) const;
 
   int32_t byte_width() const { return byte_width_; }
 
@@ -370,6 +371,30 @@ class ARROW_EXPORT FixedSizeBinaryArray : public PrimitiveArray {
   int32_t byte_width_;
 };
 
+// ----------------------------------------------------------------------
+// DecimalArray
+class ARROW_EXPORT DecimalArray : public FixedSizeBinaryArray {
+ public:
+  using TypeClass = Type;
+
+  DecimalArray(const std::shared_ptr<DataType>& type, int64_t length,
+      const std::shared_ptr<Buffer>& data,
+      const std::shared_ptr<Buffer>& null_bitmap = nullptr, int64_t null_count = 0,
+      int64_t offset = 0, const std::shared_ptr<Buffer>& sign_bitmap = nullptr);
+
+  bool IsNegative(int64_t i) const;
+
+  template <typename T>
+  ARROW_EXPORT Decimal<T> Value(int64_t i) const;
+
+  std::shared_ptr<Array> Slice(int64_t offset, int64_t length) const override;
+
+ private:
+  /// Only needed for 128 bit Decimals
+  std::shared_ptr<Buffer> sign_bitmap_;
+  const uint8_t* sign_bitmap_data_;
+};
+
 // ----------------------------------------------------------------------
 // Struct
 
diff --git a/cpp/src/arrow/builder.cc b/cpp/src/arrow/builder.cc
index 40b81cf015a..a3677eff686 100644
--- a/cpp/src/arrow/builder.cc
+++ b/cpp/src/arrow/builder.cc
@@ -27,6 +27,7 @@
 #include "arrow/type.h"
 #include "arrow/type_traits.h"
 #include "arrow/util/bit-util.h"
+#include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 
 namespace arrow {
@@ -323,6 +324,85 @@ Status BooleanBuilder::Append(
   return Status::OK();
 }
 
+// ----------------------------------------------------------------------
+// DecimalBuilder
+DecimalBuilder::DecimalBuilder(MemoryPool* pool, const std::shared_ptr<DataType>& type)
+    : FixedSizeBinaryBuilder(pool, type),
+      sign_bitmap_(nullptr),
+      sign_bitmap_data_(nullptr) {}
+
+template <typename T>
+ARROW_EXPORT Status DecimalBuilder::Append(const Decimal<T>& val) {
+  DCHECK_EQ(sign_bitmap_, nullptr) << "sign_bitmap_ is not null";
+  DCHECK_EQ(sign_bitmap_data_, nullptr) << "sign_bitmap_data_ is not null";
+
+  RETURN_NOT_OK(FixedSizeBinaryBuilder::Reserve(1));
+  return FixedSizeBinaryBuilder::Append(reinterpret_cast<const uint8_t*>(&val.value));
+}
+
+template ARROW_EXPORT Status DecimalBuilder::Append(const Decimal32& val);
+template ARROW_EXPORT Status DecimalBuilder::Append(const Decimal64& val);
+
+template <>
+ARROW_EXPORT Status DecimalBuilder::Append(const Decimal128& value) {
+  DCHECK_NE(sign_bitmap_, nullptr) << "sign_bitmap_ is null";
+  DCHECK_NE(sign_bitmap_data_, nullptr) << "sign_bitmap_data_ is null";
+
+  RETURN_NOT_OK(FixedSizeBinaryBuilder::Reserve(1));
+  uint8_t stack_bytes[16] = {0};
+  uint8_t* bytes = stack_bytes;
+  bool is_negative;
+  ToBytes(value, &bytes, &is_negative);
+  RETURN_NOT_OK(FixedSizeBinaryBuilder::Append(bytes));
+
+  // TODO(phillipc): calculate the proper storage size here (do we have a function to do
+  // this)?
+  // TODO(phillipc): Reserve number of elements
+  RETURN_NOT_OK(sign_bitmap_->Reserve(1));
+  BitUtil::SetBitTo(sign_bitmap_data_, length_ - 1, is_negative);
+  return Status::OK();
+}
+
+template ARROW_EXPORT Status DecimalBuilder::Append(const Decimal128& val);
+
+Status DecimalBuilder::Init(int64_t capacity) {
+  RETURN_NOT_OK(FixedSizeBinaryBuilder::Init(capacity));
+  if (byte_width_ == 16) {
+    AllocateResizableBuffer(pool_, null_bitmap_->size(), &sign_bitmap_);
+    sign_bitmap_data_ = sign_bitmap_->mutable_data();
+    memset(sign_bitmap_data_, 0, static_cast<size_t>(sign_bitmap_->capacity()));
+  }
+  return Status::OK();
+}
+
+Status DecimalBuilder::Resize(int64_t capacity) {
+  int64_t old_bytes = null_bitmap_ != nullptr ? null_bitmap_->size() : 0;
+  if (sign_bitmap_ == nullptr) { return Init(capacity); }
+  RETURN_NOT_OK(FixedSizeBinaryBuilder::Resize(capacity));
+
+  if (byte_width_ == 16) {
+    RETURN_NOT_OK(sign_bitmap_->Resize(null_bitmap_->size()));
+    int64_t new_bytes = sign_bitmap_->size();
+    sign_bitmap_data_ = sign_bitmap_->mutable_data();
+
+    // The buffer might be overpadded to deal with padding according to the spec
+    if (old_bytes < new_bytes) {
+      memset(sign_bitmap_data_ + old_bytes, 0,
+          static_cast<size_t>(sign_bitmap_->capacity() - old_bytes));
+    }
+  }
+  return Status::OK();
+}
+
+Status DecimalBuilder::Finish(std::shared_ptr<Array>* out) {
+  std::shared_ptr<Buffer> data = byte_builder_.Finish();
+
+  /// TODO(phillipc): not sure where to get the offset argument here
+  *out = std::make_shared<DecimalArray>(
+      type_, length_, data, null_bitmap_, null_count_, 0, sign_bitmap_);
+  return Status::OK();
+}
+
 // ----------------------------------------------------------------------
 // ListBuilder
 
@@ -440,10 +520,9 @@ Status StringBuilder::Finish(std::shared_ptr<Array>* out) {
 
 FixedSizeBinaryBuilder::FixedSizeBinaryBuilder(
     MemoryPool* pool, const std::shared_ptr<DataType>& type)
-    : ArrayBuilder(pool, type), byte_builder_(pool) {
-  DCHECK(type->type == Type::FIXED_SIZE_BINARY);
-  byte_width_ = static_cast<const FixedSizeBinaryType&>(*type).byte_width();
-}
+    : ArrayBuilder(pool, type),
+      byte_width_(static_cast<const FixedSizeBinaryType&>(*type).byte_width()),
+      byte_builder_(pool) {}
 
 Status FixedSizeBinaryBuilder::Append(const uint8_t* value) {
   RETURN_NOT_OK(Reserve(1));
@@ -543,6 +622,7 @@ Status MakeBuilder(MemoryPool* pool, const std::shared_ptr<DataType>& type,
     BUILDER_CASE(STRING, StringBuilder);
     BUILDER_CASE(BINARY, BinaryBuilder);
     BUILDER_CASE(FIXED_SIZE_BINARY, FixedSizeBinaryBuilder);
+    BUILDER_CASE(DECIMAL, DecimalBuilder);
     case Type::LIST: {
       std::shared_ptr<ArrayBuilder> value_builder;
       std::shared_ptr<DataType> value_type =
diff --git a/cpp/src/arrow/builder.h b/cpp/src/arrow/builder.h
index 60cdc4cb3a5..d42ab5b01d1 100644
--- a/cpp/src/arrow/builder.h
+++ b/cpp/src/arrow/builder.h
@@ -37,6 +37,9 @@ namespace arrow {
 
 class Array;
 
+template <typename T>
+struct Decimal;
+
 static constexpr int64_t kMinBuilderCapacity = 1 << 5;
 
 /// Base class for all data array builders.
@@ -76,12 +79,12 @@ class ARROW_EXPORT ArrayBuilder {
   Status SetNotNull(int64_t length);
 
   /// Allocates initial capacity requirements for the builder.  In most
-  /// cases subclasses should override and call there parent classes
+  /// cases subclasses should override and call their parent class's
   /// method as well.
   virtual Status Init(int64_t capacity);
 
   /// Resizes the null_bitmap array.  In most
-  /// cases subclasses should override and call there parent classes
+  /// cases subclasses should override and call their parent class's
   /// method as well.
   virtual Status Resize(int64_t new_bits);
 
@@ -275,9 +278,7 @@ class ARROW_EXPORT BooleanBuilder : public ArrayBuilder {
     return Status::OK();
   }
 
-  Status Append(uint8_t val) {
-    return Append(val != 0);
-  }
+  Status Append(uint8_t val) { return Append(val != 0); }
 
   /// Vector append
   ///
@@ -415,6 +416,24 @@ class ARROW_EXPORT FixedSizeBinaryBuilder : public ArrayBuilder {
   BufferBuilder byte_builder_;
 };
 
+class ARROW_EXPORT DecimalBuilder : public FixedSizeBinaryBuilder {
+ public:
+  explicit DecimalBuilder(MemoryPool* pool, const std::shared_ptr<DataType>& type);
+
+  template <typename T>
+  ARROW_EXPORT Status Append(const Decimal<T>& val);
+
+  Status Init(int64_t capacity) override;
+  Status Resize(int64_t capacity) override;
+  Status Finish(std::shared_ptr<Array>* out) override;
+
+ private:
+  /// We only need these for 128 bit decimals, because boost stores the sign
+  /// separate from the underlying bytes.
+  std::shared_ptr<ResizableBuffer> sign_bitmap_;
+  uint8_t* sign_bitmap_data_;
+};
+
 // ----------------------------------------------------------------------
 // Struct
 
diff --git a/cpp/src/arrow/compare.cc b/cpp/src/arrow/compare.cc
index 7451439a875..2297e4b206d 100644
--- a/cpp/src/arrow/compare.cc
+++ b/cpp/src/arrow/compare.cc
@@ -29,6 +29,7 @@
 #include "arrow/type.h"
 #include "arrow/type_traits.h"
 #include "arrow/util/bit-util.h"
+#include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 #include "arrow/visitor_inline.h"
 
@@ -232,6 +233,41 @@ class RangeEqualsVisitor {
     return Status::OK();
   }
 
+  Status Visit(const DecimalArray& left) {
+    const auto& right = static_cast<const DecimalArray&>(right_);
+
+    int32_t width = left.byte_width();
+
+    const uint8_t* left_data = nullptr;
+    const uint8_t* right_data = nullptr;
+
+    if (left.data()) { left_data = left.raw_data() + left.offset() * width; }
+
+    if (right.data()) { right_data = right.raw_data() + right.offset() * width; }
+
+    for (int64_t i = left_start_idx_, o_i = right_start_idx_; i < left_end_idx_;
+         ++i, ++o_i) {
+      if (left.IsNegative(i) != right.IsNegative(o_i)) {
+        result_ = false;
+        return Status::OK();
+      }
+
+      const bool is_null = left.IsNull(i);
+      if (is_null != right.IsNull(o_i)) {
+        result_ = false;
+        return Status::OK();
+      }
+      if (is_null) continue;
+
+      if (std::memcmp(left_data + width * i, right_data + width * o_i, width)) {
+        result_ = false;
+        return Status::OK();
+      }
+    }
+    result_ = true;
+    return Status::OK();
+  }
+
   Status Visit(const NullArray& left) {
     UNUSED(left);
     result_ = true;
@@ -244,10 +280,6 @@ class RangeEqualsVisitor {
     return CompareValues<T>(left);
   }
 
-  Status Visit(const DecimalArray& left) {
-    return Status::NotImplemented("Decimal type");
-  }
-
   Status Visit(const ListArray& left) {
     result_ = CompareLists(left);
     return Status::OK();
diff --git a/cpp/src/arrow/ipc/CMakeLists.txt b/cpp/src/arrow/ipc/CMakeLists.txt
index 57db03311c0..c6880c56e46 100644
--- a/cpp/src/arrow/ipc/CMakeLists.txt
+++ b/cpp/src/arrow/ipc/CMakeLists.txt
@@ -27,7 +27,8 @@ set(ARROW_IPC_SHARED_LINK_LIBS
 set(ARROW_IPC_TEST_LINK_LIBS
   arrow_ipc_static
   arrow_io_static
-  arrow_static)
+  arrow_static
+  ${BOOST_REGEX_LIBRARY})
 
 set(ARROW_IPC_SRCS
   feather.cc
@@ -161,7 +162,8 @@ if(MSVC)
     arrow_io_static
     arrow_static
     ${BOOST_FILESYSTEM_LIBRARY}
-    ${BOOST_SYSTEM_LIBRARY})
+    ${BOOST_SYSTEM_LIBRARY}
+    ${BOOST_REGEX_LIBRARY})
 else()
   set(UTIL_LINK_LIBS
     arrow_ipc_static
@@ -169,6 +171,7 @@ else()
     arrow_static
     ${BOOST_FILESYSTEM_LIBRARY}
     ${BOOST_SYSTEM_LIBRARY}
+    ${BOOST_REGEX_LIBRARY}
     dl)
 endif()
 
diff --git a/cpp/src/arrow/python/CMakeLists.txt b/cpp/src/arrow/python/CMakeLists.txt
index c69d976737f..604527f6304 100644
--- a/cpp/src/arrow/python/CMakeLists.txt
+++ b/cpp/src/arrow/python/CMakeLists.txt
@@ -37,7 +37,8 @@ set(ARROW_PYTHON_MIN_TEST_LIBS
   arrow_python_static
   arrow_ipc_static
   arrow_io_static
-  arrow_static)
+  arrow_static
+  ${BOOST_REGEX_LIBRARY})
 
 if(ARROW_BUILD_TESTS)
   ADD_THIRDPARTY_LIB(python
diff --git a/cpp/src/arrow/python/builtin_convert.cc b/cpp/src/arrow/python/builtin_convert.cc
index 25b32ee26a0..189ecee4fe0 100644
--- a/cpp/src/arrow/python/builtin_convert.cc
+++ b/cpp/src/arrow/python/builtin_convert.cc
@@ -17,12 +17,16 @@
 
 #include <Python.h>
 #include <datetime.h>
+
+#include <algorithm>
 #include <sstream>
+#include <string>
 
 #include "arrow/python/builtin_convert.h"
 
 #include "arrow/api.h"
 #include "arrow/status.h"
+#include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 
 #include "arrow/python/helpers.h"
@@ -109,7 +113,6 @@ class ScalarVisitor {
   int64_t float_count_;
   int64_t binary_count_;
   int64_t unicode_count_;
-
   // Place to accumulate errors
   // std::vector<Status> errors_;
 };
@@ -394,8 +397,7 @@ class BytesConverter : public TypedConverter<BinaryBuilder> {
       } else if (PyBytes_Check(item)) {
         bytes_obj = item;
       } else {
-        return Status::Invalid(
-            "Value that cannot be converted to bytes was encountered");
+        return Status::Invalid("Value that cannot be converted to bytes was encountered");
       }
       // No error checking
       length = PyBytes_GET_SIZE(bytes_obj);
@@ -429,8 +431,7 @@ class FixedWidthBytesConverter : public TypedConverter<FixedSizeBinaryBuilder> {
       } else if (PyBytes_Check(item)) {
         bytes_obj = item;
       } else {
-        return Status::Invalid(
-            "Value that cannot be converted to bytes was encountered");
+        return Status::Invalid("Value that cannot be converted to bytes was encountered");
       }
       // No error checking
       RETURN_NOT_OK(CheckPythonBytesAreFixedLength(bytes_obj, expected_length));
@@ -495,6 +496,54 @@ class ListConverter : public TypedConverter<ListBuilder> {
   std::shared_ptr<SeqConverter> value_converter_;
 };
 
+#define DECIMAL_CONVERT_CASE(bit_width, item, builder)        \
+  case bit_width: {                                           \
+    arrow::Decimal##bit_width out;                            \
+    RETURN_NOT_OK(PythonDecimalToArrowDecimal((item), &out)); \
+    RETURN_NOT_OK((builder)->Append(out));                    \
+    break;                                                    \
+  }
+
+class DecimalConverter : public TypedConverter<arrow::DecimalBuilder> {
+ public:
+  Status AppendData(PyObject* seq) override {
+    /// Ensure we've allocated enough space
+    Py_ssize_t size = PySequence_Size(seq);
+    RETURN_NOT_OK(typed_builder_->Reserve(size));
+
+    /// Can the compiler figure out that the case statement below isn't necessary
+    /// once we're running?
+    const int bit_width =
+        std::dynamic_pointer_cast<arrow::DecimalType>(typed_builder_->type())
+            ->bit_width();
+
+    OwnedRef ref;
+    PyObject* item = nullptr;
+    for (int64_t i = 0; i < size; ++i) {
+      ref.reset(PySequence_GetItem(seq, i));
+      item = ref.obj();
+
+      /// TODO(phillipc): Check for nan?
+      if (item != Py_None) {
+        switch (bit_width) {
+          DECIMAL_CONVERT_CASE(32, item, typed_builder_)
+          DECIMAL_CONVERT_CASE(64, item, typed_builder_)
+          DECIMAL_CONVERT_CASE(128, item, typed_builder_)
+          default:
+            break;
+        }
+        RETURN_IF_PYERROR();
+      } else {
+        RETURN_NOT_OK(typed_builder_->AppendNull());
+      }
+    }
+
+    return Status::OK();
+  }
+};
+
+#undef DECIMAL_CONVERT_CASE
+
 // Dynamic constructor for sequence converters
 std::shared_ptr<SeqConverter> GetConverter(const std::shared_ptr<DataType>& type) {
   switch (type->type) {
@@ -516,6 +565,9 @@ std::shared_ptr<SeqConverter> GetConverter(const std::shared_ptr<DataType>& type
       return std::make_shared<UTF8Converter>();
     case Type::LIST:
       return std::make_shared<ListConverter>();
+    case Type::DECIMAL: {
+      return std::make_shared<DecimalConverter>();
+    }
     case Type::STRUCT:
     default:
       return nullptr;
diff --git a/cpp/src/arrow/python/builtin_convert.h b/cpp/src/arrow/python/builtin_convert.h
index 00ff0fd8236..3c2e350269a 100644
--- a/cpp/src/arrow/python/builtin_convert.h
+++ b/cpp/src/arrow/python/builtin_convert.h
@@ -25,7 +25,7 @@
 
 #include <memory>
 
-#include <arrow/type.h>
+#include "arrow/type.h"
 
 #include "arrow/util/visibility.h"
 
diff --git a/cpp/src/arrow/python/common.h b/cpp/src/arrow/python/common.h
index 32bfa784acb..a6806ab95ab 100644
--- a/cpp/src/arrow/python/common.h
+++ b/cpp/src/arrow/python/common.h
@@ -57,12 +57,13 @@ class OwnedRef {
   }
 
   void reset(PyObject* obj) {
-    if (obj_ != nullptr) { Py_XDECREF(obj_); }
+    /// TODO(phillipc): Should we acquire the GIL here? It definitely needs to be
+    /// acquired,
+    /// but callers have probably already acquired it
+    Py_XDECREF(obj_);
     obj_ = obj;
   }
 
-  void release() { obj_ = nullptr; }
-
   PyObject* obj() const { return obj_; }
 
  private:
@@ -72,6 +73,7 @@ class OwnedRef {
 struct PyObjectStringify {
   OwnedRef tmp_obj;
   const char* bytes;
+  Py_ssize_t size;
 
   explicit PyObjectStringify(PyObject* obj) {
     PyObject* bytes_obj;
@@ -82,6 +84,7 @@ struct PyObjectStringify {
       bytes_obj = obj;
     }
     bytes = PyBytes_AsString(bytes_obj);
+    size = PyBytes_GET_SIZE(bytes_obj);
   }
 };
 
diff --git a/cpp/src/arrow/python/helpers.cc b/cpp/src/arrow/python/helpers.cc
index be5f412fbea..ffba7bbc21c 100644
--- a/cpp/src/arrow/python/helpers.cc
+++ b/cpp/src/arrow/python/helpers.cc
@@ -16,6 +16,8 @@
 // under the License.
 
 #include "arrow/python/helpers.h"
+#include "arrow/python/common.h"
+#include "arrow/util/decimal.h"
 
 #include <arrow/api.h>
 
@@ -52,5 +54,82 @@ std::shared_ptr<DataType> GetPrimitiveType(Type::type type) {
   }
 }
 
+Status ImportModule(const std::string& module_name, OwnedRef* ref) {
+  PyAcquireGIL lock;
+  PyObject* module = PyImport_ImportModule(module_name.c_str());
+  RETURN_IF_PYERROR();
+  ref->reset(module);
+  return Status::OK();
+}
+
+Status ImportFromModule(const OwnedRef& module, const std::string& name, OwnedRef* ref) {
+  /// Assumes that ImportModule was called first
+  DCHECK_NE(module.obj(), nullptr) << "Cannot import from nullptr Python module";
+
+  PyAcquireGIL lock;
+  PyObject* attr = PyObject_GetAttrString(module.obj(), name.c_str());
+  RETURN_IF_PYERROR();
+  ref->reset(attr);
+  return Status::OK();
+}
+
+template <typename T>
+Status PythonDecimalToArrowDecimal(PyObject* python_decimal, Decimal<T>* arrow_decimal) {
+  // Call Python's str(decimal_object)
+  OwnedRef str_obj(PyObject_Str(python_decimal));
+  RETURN_IF_PYERROR();
+
+  PyObjectStringify str(str_obj.obj());
+  RETURN_IF_PYERROR();
+
+  const char* bytes = str.bytes;
+  DCHECK_NE(bytes, nullptr);
+
+  Py_ssize_t size = str.size;
+
+  std::string c_string(bytes, size);
+  return FromString(c_string, arrow_decimal);
+}
+
+template Status PythonDecimalToArrowDecimal(
+    PyObject* python_decimal, Decimal32* arrow_decimal);
+template Status PythonDecimalToArrowDecimal(
+    PyObject* python_decimal, Decimal64* arrow_decimal);
+template Status PythonDecimalToArrowDecimal(
+    PyObject* python_decimal, Decimal128* arrow_decimal);
+
+Status InferDecimalPrecisionAndScale(
+    PyObject* python_decimal, int* precision, int* scale) {
+  // Call Python's str(decimal_object)
+  OwnedRef str_obj(PyObject_Str(python_decimal));
+  RETURN_IF_PYERROR();
+  PyObjectStringify str(str_obj.obj());
+
+  const char* bytes = str.bytes;
+  DCHECK_NE(bytes, nullptr);
+
+  auto size = str.size;
+
+  std::string c_string(bytes, size);
+  return FromString(c_string, static_cast<Decimal32*>(nullptr), precision, scale);
+}
+
+Status DecimalFromString(
+    PyObject* decimal_constructor, const std::string& decimal_string, PyObject** out) {
+  DCHECK_NE(decimal_constructor, nullptr);
+  DCHECK_NE(out, nullptr);
+
+  auto string_size = decimal_string.size();
+  DCHECK_GT(string_size, 0);
+
+  auto string_bytes = decimal_string.c_str();
+  DCHECK_NE(string_bytes, nullptr);
+
+  *out = PyObject_CallFunction(
+      decimal_constructor, const_cast<char*>("s#"), string_bytes, string_size);
+  RETURN_IF_PYERROR();
+  return Status::OK();
+}
+
 }  // namespace py
 }  // namespace arrow
diff --git a/cpp/src/arrow/python/helpers.h b/cpp/src/arrow/python/helpers.h
index 611e814b7d8..a19b25f7db8 100644
--- a/cpp/src/arrow/python/helpers.h
+++ b/cpp/src/arrow/python/helpers.h
@@ -18,16 +18,38 @@
 #ifndef PYARROW_HELPERS_H
 #define PYARROW_HELPERS_H
 
+#include <Python.h>
+
 #include <memory>
+#include <string>
+#include <utility>
 
 #include "arrow/type.h"
 #include "arrow/util/visibility.h"
 
 namespace arrow {
+
+template <typename T>
+struct Decimal;
+
 namespace py {
 
-ARROW_EXPORT
-std::shared_ptr<DataType> GetPrimitiveType(Type::type type);
+class OwnedRef;
+
+ARROW_EXPORT std::shared_ptr<DataType> GetPrimitiveType(Type::type type);
+
+Status ImportModule(const std::string& module_name, OwnedRef* ref);
+Status ImportFromModule(
+    const OwnedRef& module, const std::string& module_name, OwnedRef* ref);
+
+template <typename T>
+Status PythonDecimalToArrowDecimal(PyObject* python_decimal, Decimal<T>* arrow_decimal);
+
+Status InferDecimalPrecisionAndScale(
+    PyObject* python_decimal, int* precision = nullptr, int* scale = nullptr);
+
+Status DecimalFromString(
+    PyObject* decimal_constructor, const std::string& decimal_string, PyObject** out);
 
 }  // namespace py
 }  // namespace arrow
diff --git a/cpp/src/arrow/python/pandas_convert.cc b/cpp/src/arrow/python/pandas_convert.cc
index 48d3489bf90..f6e627e668e 100644
--- a/cpp/src/arrow/python/pandas_convert.cc
+++ b/cpp/src/arrow/python/pandas_convert.cc
@@ -41,12 +41,14 @@
 #include "arrow/type_fwd.h"
 #include "arrow/type_traits.h"
 #include "arrow/util/bit-util.h"
+#include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 #include "arrow/util/macros.h"
 
 #include "arrow/python/builtin_convert.h"
 #include "arrow/python/common.h"
 #include "arrow/python/config.h"
+#include "arrow/python/helpers.h"
 #include "arrow/python/numpy-internal.h"
 #include "arrow/python/numpy_convert.h"
 #include "arrow/python/type_traits.h"
@@ -375,6 +377,7 @@ class PandasConverter : public TypeVisitor {
   Status ConvertDates();
   Status ConvertLists(const std::shared_ptr<DataType>& type);
   Status ConvertObjects();
+  Status ConvertDecimals();
 
  protected:
   MemoryPool* pool_;
@@ -468,15 +471,14 @@ Status InvalidConversion(PyObject* obj, const std::string& expected_type_name) {
   RETURN_IF_PYERROR();
   DCHECK_NE(type_name.obj(), nullptr);
 
-  OwnedRef bytes_obj(PyUnicode_AsUTF8String(type_name.obj()));
+  PyObjectStringify bytestring(type_name.obj());
   RETURN_IF_PYERROR();
-  DCHECK_NE(bytes_obj.obj(), nullptr);
-
-  Py_ssize_t size = PyBytes_GET_SIZE(bytes_obj.obj());
-  const char* bytes = PyBytes_AS_STRING(bytes_obj.obj());
 
+  const char* bytes = bytestring.bytes;
   DCHECK_NE(bytes, nullptr) << "bytes from type(...).__name__ were null";
 
+  Py_ssize_t size = bytestring.size;
+
   std::string cpp_type_name(bytes, size);
 
   std::stringstream ss;
@@ -517,6 +519,59 @@ Status PandasConverter::ConvertDates() {
   return date_builder.Finish(&out_);
 }
 
+#define CONVERT_DECIMAL_CASE(bit_width, builder, object)      \
+  case bit_width: {                                           \
+    Decimal##bit_width d;                                     \
+    RETURN_NOT_OK(PythonDecimalToArrowDecimal((object), &d)); \
+    RETURN_NOT_OK((builder).Append(d));                       \
+    break;                                                    \
+  }
+
+Status PandasConverter::ConvertDecimals() {
+  PyAcquireGIL lock;
+
+  // Import the decimal module and Decimal class
+  OwnedRef decimal;
+  OwnedRef Decimal;
+  RETURN_NOT_OK(ImportModule("decimal", &decimal));
+  RETURN_NOT_OK(ImportFromModule(decimal, "Decimal", &Decimal));
+
+  PyObject** objects = reinterpret_cast<PyObject**>(PyArray_DATA(arr_));
+  PyObject* object = objects[0];
+
+  int precision;
+  int scale;
+
+  RETURN_NOT_OK(InferDecimalPrecisionAndScale(object, &precision, &scale));
+
+  type_ = std::make_shared<DecimalType>(precision, scale);
+
+  const int bit_width = std::dynamic_pointer_cast<DecimalType>(type_)->bit_width();
+  DecimalBuilder decimal_builder(pool_, type_);
+
+  RETURN_NOT_OK(decimal_builder.Resize(length_));
+
+  for (int64_t i = 0; i < length_; ++i) {
+    object = objects[i];
+    if (PyObject_IsInstance(object, Decimal.obj())) {
+      switch (bit_width) {
+        CONVERT_DECIMAL_CASE(32, decimal_builder, object)
+        CONVERT_DECIMAL_CASE(64, decimal_builder, object)
+        CONVERT_DECIMAL_CASE(128, decimal_builder, object)
+        default:
+          break;
+      }
+    } else if (PyObject_is_null(object)) {
+      decimal_builder.AppendNull();
+    } else {
+      return InvalidConversion(object, "decimal.Decimal");
+    }
+  }
+  return decimal_builder.Finish(&out_);
+}
+
+#undef CONVERT_DECIMAL_CASE
+
 Status PandasConverter::ConvertObjectStrings() {
   PyAcquireGIL lock;
 
@@ -554,6 +609,90 @@ Status PandasConverter::ConvertObjectFixedWidthBytes(
   return Status::OK();
 }
 
+template <typename T>
+Status validate_precision(int precision) {
+  constexpr static const int maximum_precision = DecimalPrecision<T>::maximum;
+  if (!(precision > 0 && precision <= maximum_precision)) {
+    std::stringstream ss;
+    ss << "Invalid precision: " << precision << ". Minimum is 1, maximum is "
+       << maximum_precision;
+    return Status::Invalid(ss.str());
+  }
+  return Status::OK();
+}
+
+template <typename T>
+Status RawDecimalToString(
+    const uint8_t* bytes, int precision, int scale, std::string* result) {
+  DCHECK_NE(bytes, nullptr);
+  DCHECK_NE(result, nullptr);
+  RETURN_NOT_OK(validate_precision<T>(precision));
+  Decimal<T> decimal;
+  FromBytes(bytes, &decimal);
+  *result = ToString(decimal, precision, scale);
+  return Status::OK();
+}
+
+template Status RawDecimalToString<int32_t>(
+    const uint8_t*, int, int, std::string* result);
+template Status RawDecimalToString<int64_t>(
+    const uint8_t*, int, int, std::string* result);
+
+Status RawDecimalToString(const uint8_t* bytes, int precision, int scale,
+    bool is_negative, std::string* result) {
+  DCHECK_NE(bytes, nullptr);
+  DCHECK_NE(result, nullptr);
+  RETURN_NOT_OK(validate_precision<int128_t>(precision));
+  Decimal128 decimal;
+  FromBytes(bytes, is_negative, &decimal);
+  *result = ToString(decimal, precision, scale);
+  return Status::OK();
+}
+
+static Status ConvertDecimals(const ChunkedArray& data, PyObject** out_values) {
+  PyAcquireGIL lock;
+  OwnedRef decimal_ref;
+  OwnedRef Decimal_ref;
+  RETURN_NOT_OK(ImportModule("decimal", &decimal_ref));
+  RETURN_NOT_OK(ImportFromModule(decimal_ref, "Decimal", &Decimal_ref));
+  PyObject* Decimal = Decimal_ref.obj();
+
+  for (int c = 0; c < data.num_chunks(); c++) {
+    auto* arr(static_cast<arrow::DecimalArray*>(data.chunk(c).get()));
+    auto type(std::dynamic_pointer_cast<arrow::DecimalType>(arr->type()));
+    const int precision = type->precision;
+    const int scale = type->scale;
+    const int bit_width = type->bit_width();
+
+    for (int64_t i = 0; i < arr->length(); ++i) {
+      if (arr->IsNull(i)) {
+        Py_INCREF(Py_None);
+        *out_values++ = Py_None;
+      } else {
+        const uint8_t* raw_value = arr->GetValue(i);
+        std::string s;
+        switch (bit_width) {
+          case 32:
+            RETURN_NOT_OK(RawDecimalToString<int32_t>(raw_value, precision, scale, &s));
+            break;
+          case 64:
+            RETURN_NOT_OK(RawDecimalToString<int64_t>(raw_value, precision, scale, &s));
+            break;
+          case 128:
+            RETURN_NOT_OK(
+                RawDecimalToString(raw_value, precision, scale, arr->IsNegative(i), &s));
+            break;
+          default:
+            break;
+        }
+        RETURN_NOT_OK(DecimalFromString(Decimal, s, out_values++));
+      }
+    }
+  }
+
+  return Status::OK();
+}
+
 Status PandasConverter::ConvertBooleans() {
   PyAcquireGIL lock;
 
@@ -598,6 +737,7 @@ Status PandasConverter::ConvertObjects() {
   //
   // * Strings
   // * Booleans with nulls
+  // * decimal.Decimals
   // * Mixed type (not supported at the moment by arrow format)
   //
   // Additionally, nulls may be encoded either as np.nan or None. So we have to
@@ -613,6 +753,7 @@ Status PandasConverter::ConvertObjects() {
     PyDateTime_IMPORT;
   }
 
+  // This means we received an explicit type from the user
   if (type_) {
     switch (type_->type) {
       case Type::STRING:
@@ -627,10 +768,17 @@ Status PandasConverter::ConvertObjects() {
         const auto& list_field = static_cast<const ListType&>(*type_);
         return ConvertLists(list_field.value_field()->type);
       }
+      case Type::DECIMAL:
+        return ConvertDecimals();
       default:
         return Status::TypeError("No known conversion to Arrow type");
     }
   } else {
+    OwnedRef decimal;
+    OwnedRef Decimal;
+    RETURN_NOT_OK(ImportModule("decimal", &decimal));
+    RETURN_NOT_OK(ImportFromModule(decimal, "Decimal", &Decimal));
+
     for (int64_t i = 0; i < length_; ++i) {
       if (PyObject_is_null(objects[i])) {
         continue;
@@ -640,6 +788,8 @@ Status PandasConverter::ConvertObjects() {
         return ConvertBooleans();
       } else if (PyDate_CheckExact(objects[i])) {
         return ConvertDates();
+      } else if (PyObject_IsInstance(const_cast<PyObject*>(objects[i]), Decimal.obj())) {
+        return ConvertDecimals();
       } else {
         return InvalidConversion(
             const_cast<PyObject*>(objects[i]), "string, bool, or date");
@@ -847,6 +997,7 @@ class PandasBlock {
     INT64,
     FLOAT,
     DOUBLE,
+    DECIMAL,
     BOOL,
     DATETIME,
     DATETIME_WITH_TZ,
@@ -1193,6 +1344,8 @@ class ObjectBlock : public PandasBlock {
       RETURN_NOT_OK(ConvertBinaryLike<StringArray>(data, out_buffer));
     } else if (type == Type::FIXED_SIZE_BINARY) {
       RETURN_NOT_OK(ConvertFixedSizeBinary(data, out_buffer));
+    } else if (type == Type::DECIMAL) {
+      RETURN_NOT_OK(ConvertDecimals(data, out_buffer));
     } else if (type == Type::LIST) {
       auto list_type = std::static_pointer_cast<ListType>(col->type());
       switch (list_type->value_type()->type) {
@@ -1519,6 +1672,7 @@ Status MakeBlock(PandasBlock::type type, int64_t num_rows, int num_columns,
     BLOCK_CASE(DOUBLE, Float64Block);
     BLOCK_CASE(BOOL, BoolBlock);
     BLOCK_CASE(DATETIME, DatetimeBlock);
+    BLOCK_CASE(DECIMAL, ObjectBlock);
     default:
       return Status::NotImplemented("Unsupported block type");
   }
@@ -1649,6 +1803,9 @@ class DataFrameBlockCreator {
         case Type::DICTIONARY:
           output_type = PandasBlock::CATEGORICAL;
           break;
+        case Type::DECIMAL:
+          output_type = PandasBlock::DECIMAL;
+          break;
         default:
           return Status::NotImplemented(col->type()->ToString());
       }
@@ -1892,6 +2049,7 @@ class ArrowDeserializer {
       CONVERT_CASE(TIMESTAMP);
       CONVERT_CASE(DICTIONARY);
       CONVERT_CASE(LIST);
+      CONVERT_CASE(DECIMAL);
       default: {
         std::stringstream ss;
         ss << "Arrow type reading not implemented for " << col_->type()->ToString();
@@ -1999,6 +2157,13 @@ class ArrowDeserializer {
     return ConvertFixedSizeBinary(data_, out_values);
   }
 
+  template <int TYPE>
+  inline typename std::enable_if<TYPE == Type::DECIMAL, Status>::type ConvertValues() {
+    RETURN_NOT_OK(AllocateOutput(NPY_OBJECT));
+    auto out_values = reinterpret_cast<PyObject**>(PyArray_DATA(arr_));
+    return ConvertDecimals(data_, out_values);
+  }
+
 #define CONVERTVALUES_LISTSLIKE_CASE(ArrowType, ArrowEnum) \
   case Type::ArrowEnum:                                    \
     return ConvertListsLike<ArrowType>(col_, out_values);
@@ -2021,6 +2186,7 @@ class ArrowDeserializer {
       CONVERTVALUES_LISTSLIKE_CASE(FloatType, FLOAT)
       CONVERTVALUES_LISTSLIKE_CASE(DoubleType, DOUBLE)
       CONVERTVALUES_LISTSLIKE_CASE(StringType, STRING)
+      CONVERTVALUES_LISTSLIKE_CASE(DecimalType, DECIMAL)
       default: {
         std::stringstream ss;
         ss << "Not implemented type for lists: " << list_type->value_type()->ToString();
diff --git a/cpp/src/arrow/python/python-test.cc b/cpp/src/arrow/python/python-test.cc
index f269ebfb642..b63d2ffb1cd 100644
--- a/cpp/src/arrow/python/python-test.cc
+++ b/cpp/src/arrow/python/python-test.cc
@@ -28,8 +28,11 @@
 
 #include "arrow/python/builtin_convert.h"
 #include "arrow/python/common.h"
+#include "arrow/python/helpers.h"
 #include "arrow/python/pandas_convert.h"
 
+#include "arrow/util/decimal.h"
+
 namespace arrow {
 namespace py {
 
@@ -37,6 +40,36 @@ TEST(PyBuffer, InvalidInputObject) {
   PyBuffer buffer(Py_None);
 }
 
+TEST(DecimalTest, TestPythonDecimalToArrowDecimal128) {
+  PyAcquireGIL lock;
+
+  OwnedRef decimal;
+  OwnedRef Decimal;
+  ASSERT_OK(ImportModule("decimal", &decimal));
+  ASSERT_NE(decimal.obj(), nullptr);
+
+  ASSERT_OK(ImportFromModule(decimal, "Decimal", &Decimal));
+  ASSERT_NE(Decimal.obj(), nullptr);
+
+  std::string decimal_string("-39402950693754869342983");
+  const char* format = "s#";
+  auto c_string = decimal_string.c_str();
+  ASSERT_NE(c_string, nullptr);
+
+  auto c_string_size = decimal_string.size();
+  ASSERT_GT(c_string_size, 0);
+  OwnedRef pydecimal(PyObject_CallFunction(
+      Decimal.obj(), const_cast<char*>(format), c_string, c_string_size));
+  ASSERT_NE(pydecimal.obj(), nullptr);
+  ASSERT_EQ(PyErr_Occurred(), nullptr);
+
+  Decimal128 arrow_decimal;
+  int128_t boost_decimal(decimal_string);
+  PyObject* obj = pydecimal.obj();
+  ASSERT_OK(PythonDecimalToArrowDecimal(obj, &arrow_decimal));
+  ASSERT_EQ(boost_decimal, arrow_decimal.value);
+}
+
 TEST(PandasConversionTest, TestObjectBlockWriteFails) {
   StringBuilder builder(default_memory_pool());
   const char value[] = {'\xf1', '\0'};
diff --git a/cpp/src/arrow/type.cc b/cpp/src/arrow/type.cc
index abbb626e0fc..df4590f18d7 100644
--- a/cpp/src/arrow/type.cc
+++ b/cpp/src/arrow/type.cc
@@ -17,6 +17,7 @@
 
 #include "arrow/type.h"
 
+#include <climits>
 #include <sstream>
 #include <string>
 
@@ -91,7 +92,7 @@ std::string BinaryType::ToString() const {
 }
 
 int FixedSizeBinaryType::bit_width() const {
-  return 8 * byte_width();
+  return CHAR_BIT * byte_width();
 }
 
 std::string FixedSizeBinaryType::ToString() const {
@@ -380,6 +381,10 @@ std::shared_ptr<Field> field(
   return std::make_shared<Field>(name, type, nullable);
 }
 
+std::shared_ptr<DataType> decimal(int precision, int scale) {
+  return std::make_shared<DecimalType>(precision, scale);
+}
+
 static const BufferDescr kValidityBuffer(BufferType::VALIDITY, 1);
 static const BufferDescr kOffsetBuffer(BufferType::OFFSET, 32);
 static const BufferDescr kTypeBuffer(BufferType::TYPE, 32);
@@ -402,7 +407,11 @@ std::vector<BufferDescr> BinaryType::GetBufferLayout() const {
 }
 
 std::vector<BufferDescr> FixedSizeBinaryType::GetBufferLayout() const {
-  return {kValidityBuffer, BufferDescr(BufferType::DATA, byte_width_ * 8)};
+  return {kValidityBuffer, BufferDescr(BufferType::DATA, bit_width())};
+}
+
+std::vector<BufferDescr> DecimalType::GetBufferLayout() const {
+  return {kValidityBuffer, kBooleanBuffer, BufferDescr(BufferType::DATA, bit_width())};
 }
 
 std::vector<BufferDescr> ListType::GetBufferLayout() const {
@@ -427,9 +436,4 @@ std::string DecimalType::ToString() const {
   return s.str();
 }
 
-std::vector<BufferDescr> DecimalType::GetBufferLayout() const {
-  // TODO(wesm)
-  return {};
-}
-
 }  // namespace arrow
diff --git a/cpp/src/arrow/type.h b/cpp/src/arrow/type.h
index 36ab9d8b2b9..3a35f563811 100644
--- a/cpp/src/arrow/type.h
+++ b/cpp/src/arrow/type.h
@@ -360,6 +360,8 @@ class ARROW_EXPORT FixedSizeBinaryType : public FixedWidthType {
 
   explicit FixedSizeBinaryType(int32_t byte_width)
       : FixedWidthType(Type::FIXED_SIZE_BINARY), byte_width_(byte_width) {}
+  explicit FixedSizeBinaryType(int32_t byte_width, Type::type type_id)
+      : FixedWidthType(type_id), byte_width_(byte_width) {}
 
   Status Accept(TypeVisitor* visitor) const override;
   std::string ToString() const override;
@@ -399,19 +401,31 @@ struct ARROW_EXPORT StructType : public NestedType {
   std::vector<BufferDescr> GetBufferLayout() const override;
 };
 
-struct ARROW_EXPORT DecimalType : public DataType {
+static inline int decimal_byte_width(int precision) {
+  if (precision >= 0 && precision < 10) {
+    return 4;
+  } else if (precision >= 10 && precision < 19) {
+    return 8;
+  } else {
+    // TODO(phillipc): validate that we can't construct > 128 bit types
+    return 16;
+  }
+}
+
+struct ARROW_EXPORT DecimalType : public FixedSizeBinaryType {
   static constexpr Type::type type_id = Type::DECIMAL;
 
   explicit DecimalType(int precision_, int scale_)
-      : DataType(Type::DECIMAL), precision(precision_), scale(scale_) {}
-  int precision;
-  int scale;
-
+      : FixedSizeBinaryType(decimal_byte_width(precision_), Type::DECIMAL),
+        precision(precision_),
+        scale(scale_) {}
+  std::vector<BufferDescr> GetBufferLayout() const override;
   Status Accept(TypeVisitor* visitor) const override;
   std::string ToString() const override;
   static std::string name() { return "decimal"; }
 
-  std::vector<BufferDescr> GetBufferLayout() const override;
+  int precision;
+  int scale;
 };
 
 enum class UnionMode : char { SPARSE, DENSE };
diff --git a/cpp/src/arrow/type_fwd.h b/cpp/src/arrow/type_fwd.h
index 2e27ce98589..acf12c3d9d1 100644
--- a/cpp/src/arrow/type_fwd.h
+++ b/cpp/src/arrow/type_fwd.h
@@ -69,6 +69,7 @@ class StructBuilder;
 
 struct DecimalType;
 class DecimalArray;
+class DecimalBuilder;
 
 struct UnionType;
 class UnionArray;
@@ -146,6 +147,7 @@ std::shared_ptr<DataType> ARROW_EXPORT binary();
 
 std::shared_ptr<DataType> ARROW_EXPORT date32();
 std::shared_ptr<DataType> ARROW_EXPORT date64();
+std::shared_ptr<DataType> ARROW_EXPORT decimal(int precision, int scale);
 
 }  // namespace arrow
 
diff --git a/cpp/src/arrow/type_traits.h b/cpp/src/arrow/type_traits.h
index 353b638fed8..3e8ea23432b 100644
--- a/cpp/src/arrow/type_traits.h
+++ b/cpp/src/arrow/type_traits.h
@@ -228,6 +228,13 @@ struct TypeTraits<DoubleType> {
   static inline std::shared_ptr<DataType> type_singleton() { return float64(); }
 };
 
+template <>
+struct TypeTraits<DecimalType> {
+  using ArrayType = DecimalArray;
+  using BuilderType = DecimalBuilder;
+  constexpr static bool is_parameter_free = false;
+};
+
 template <>
 struct TypeTraits<BooleanType> {
   using ArrayType = BooleanArray;
@@ -289,12 +296,6 @@ struct TypeTraits<DictionaryType> {
   constexpr static bool is_parameter_free = false;
 };
 
-template <>
-struct TypeTraits<DecimalType> {
-  // using ArrayType = DecimalArray;
-  constexpr static bool is_parameter_free = false;
-};
-
 // Not all type classes have a c_type
 template <typename T>
 struct as_void {
diff --git a/cpp/src/arrow/util/CMakeLists.txt b/cpp/src/arrow/util/CMakeLists.txt
index c1b6877a3e9..054f11055b6 100644
--- a/cpp/src/arrow/util/CMakeLists.txt
+++ b/cpp/src/arrow/util/CMakeLists.txt
@@ -22,6 +22,7 @@
 # Headers: top level
 install(FILES
   bit-util.h
+  decimal.h
   logging.h
   macros.h
   random.h
@@ -70,3 +71,4 @@ endif()
 
 ADD_ARROW_TEST(bit-util-test)
 ADD_ARROW_TEST(stl-util-test)
+ADD_ARROW_TEST(decimal-test)
diff --git a/cpp/src/arrow/util/bit-util.h b/cpp/src/arrow/util/bit-util.h
index 42afd0705f0..90a1c3eab92 100644
--- a/cpp/src/arrow/util/bit-util.h
+++ b/cpp/src/arrow/util/bit-util.h
@@ -149,7 +149,6 @@ int64_t ARROW_EXPORT CountSetBits(
 
 bool ARROW_EXPORT BitmapEquals(const uint8_t* left, int64_t left_offset,
     const uint8_t* right, int64_t right_offset, int64_t bit_length);
-
 }  // namespace arrow
 
 #endif  // ARROW_UTIL_BIT_UTIL_H
diff --git a/cpp/src/arrow/util/decimal-test.cc b/cpp/src/arrow/util/decimal-test.cc
new file mode 100644
index 00000000000..1e22643962d
--- /dev/null
+++ b/cpp/src/arrow/util/decimal-test.cc
@@ -0,0 +1,161 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+//
+
+#include "arrow/util/decimal.h"
+
+#include "gtest/gtest.h"
+
+#include "arrow/test-util.h"
+
+namespace arrow {
+
+template <typename T>
+class DecimalTest : public ::testing::Test {
+ public:
+  DecimalTest() : string_value("234.23445") { integer_value.value = 23423445; }
+  Decimal<T> integer_value;
+  std::string string_value;
+};
+
+typedef ::testing::Types<int32_t, int64_t, int128_t> DecimalTypes;
+TYPED_TEST_CASE(DecimalTest, DecimalTypes);
+
+TYPED_TEST(DecimalTest, TestToString) {
+  Decimal<TypeParam> decimal(this->integer_value);
+  int precision = 8;
+  int scale = 5;
+  std::string result = ToString(decimal, precision, scale);
+  ASSERT_EQ(result, this->string_value);
+}
+
+TYPED_TEST(DecimalTest, TestFromString) {
+  Decimal<TypeParam> expected(this->integer_value);
+  Decimal<TypeParam> result;
+  int precision, scale;
+  ASSERT_OK(FromString(this->string_value, &result, &precision, &scale));
+  ASSERT_EQ(result.value, expected.value);
+  ASSERT_EQ(precision, 8);
+  ASSERT_EQ(scale, 5);
+}
+
+TEST(DecimalTest, TestStringToInt32) {
+  int32_t value = 0;
+  StringToInteger("123", "456", 1, &value);
+  ASSERT_EQ(value, 123456);
+}
+
+TEST(DecimalTest, TestStringToInt64) {
+  int64_t value = 0;
+  StringToInteger("123456789", "456", -1, &value);
+  ASSERT_EQ(value, -123456789456);
+}
+
+TEST(DecimalTest, TestStringToInt128) {
+  int128_t value = 0;
+  StringToInteger("123456789", "456789123", 1, &value);
+  ASSERT_EQ(value, 123456789456789123);
+}
+
+TEST(DecimalTest, TestFromString128) {
+  static const std::string string_value("-23049223942343532412");
+  Decimal<int128_t> result(string_value);
+  int128_t expected = -230492239423435324;
+  ASSERT_EQ(result.value, expected * 100 - 12);
+
+  // Sanity check that our number is actually using more than 64 bits
+  ASSERT_NE(result.value, static_cast<int64_t>(result.value));
+}
+
+TEST(DecimalTest, TestFromDecimalString128) {
+  static const std::string string_value("-23049223942343.532412");
+  Decimal<int128_t> result(string_value);
+  int128_t expected = -230492239423435324;
+  ASSERT_EQ(result.value, expected * 100 - 12);
+
+  // Sanity check that our number is actually using more than 64 bits
+  ASSERT_NE(result.value, static_cast<int64_t>(result.value));
+}
+
+TEST(DecimalTest, TestDecimal32Precision) {
+  auto min_precision = DecimalPrecision<int32_t>::minimum;
+  auto max_precision = DecimalPrecision<int32_t>::maximum;
+  ASSERT_EQ(min_precision, 1);
+  ASSERT_EQ(max_precision, 9);
+}
+
+TEST(DecimalTest, TestDecimal64Precision) {
+  auto min_precision = DecimalPrecision<int64_t>::minimum;
+  auto max_precision = DecimalPrecision<int64_t>::maximum;
+  ASSERT_EQ(min_precision, 10);
+  ASSERT_EQ(max_precision, 18);
+}
+
+TEST(DecimalTest, TestDecimal128Precision) {
+  auto min_precision = DecimalPrecision<int128_t>::minimum;
+  auto max_precision = DecimalPrecision<int128_t>::maximum;
+  ASSERT_EQ(min_precision, 19);
+  ASSERT_EQ(max_precision, 38);
+}
+
+TEST(DecimalTest, TestDecimal32SignedRoundTrip) {
+  Decimal32 expected(std::string("-3402692"));
+
+  uint8_t stack_bytes[4] = {0};
+  uint8_t* bytes = stack_bytes;
+  ToBytes(expected, &bytes);
+
+  Decimal32 result;
+  FromBytes(bytes, &result);
+  ASSERT_EQ(expected.value, result.value);
+}
+
+TEST(DecimalTest, TestDecimal64SignedRoundTrip) {
+  Decimal64 expected(std::string("-34034293045.921"));
+
+  uint8_t stack_bytes[8] = {0};
+  uint8_t* bytes = stack_bytes;
+  ToBytes(expected, &bytes);
+
+  Decimal64 result;
+  FromBytes(bytes, &result);
+
+  ASSERT_EQ(expected.value, result.value);
+}
+
+TEST(DecimalTest, TestDecimal128StringAndBytesRoundTrip) {
+  std::string string_value("-340282366920938463463374607431.711455");
+  Decimal128 expected(string_value);
+
+  std::string expected_string_value("-340282366920938463463374607431711455");
+  int128_t expected_underlying_value(expected_string_value);
+
+  ASSERT_EQ(expected.value, expected_underlying_value);
+
+  uint8_t stack_bytes[16] = {0};
+  uint8_t* bytes = stack_bytes;
+  bool is_negative;
+  ToBytes(expected, &bytes, &is_negative);
+
+  ASSERT_TRUE(is_negative);
+
+  Decimal128 result;
+  FromBytes(bytes, is_negative, &result);
+
+  ASSERT_EQ(expected.value, result.value);
+}
+}  // namespace arrow
diff --git a/cpp/src/arrow/util/decimal.cc b/cpp/src/arrow/util/decimal.cc
new file mode 100644
index 00000000000..1ac347180fe
--- /dev/null
+++ b/cpp/src/arrow/util/decimal.cc
@@ -0,0 +1,141 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+#include "arrow/util/decimal.h"
+
+#include <boost/regex.hpp>
+
+namespace arrow {
+
+static const boost::regex DECIMAL_PATTERN("(\\+?|-?)((0*)(\\d*))(\\.(\\d+))?");
+
+template <typename T>
+ARROW_EXPORT Status FromString(
+    const std::string& s, Decimal<T>* out, int* precision, int* scale) {
+  if (s.empty()) {
+    return Status::Invalid("Empty string cannot be converted to decimal");
+  }
+  boost::smatch match;
+  if (!boost::regex_match(s, match, DECIMAL_PATTERN)) {
+    std::stringstream ss;
+    ss << "String " << s << " is not a valid decimal string";
+    return Status::Invalid(ss.str());
+  }
+  const int8_t sign = match[1].str() == "-" ? -1 : 1;
+  std::string whole_part = match[4].str();
+  std::string fractional_part = match[6].str();
+  if (scale != nullptr) { *scale = static_cast<int>(fractional_part.size()); }
+  if (precision != nullptr) {
+    *precision =
+        static_cast<int>(whole_part.size()) + static_cast<int>(fractional_part.size());
+  }
+  if (out != nullptr) { StringToInteger(whole_part, fractional_part, sign, &out->value); }
+  return Status::OK();
+}
+
+template ARROW_EXPORT Status FromString(
+    const std::string& s, Decimal32* out, int* precision, int* scale);
+template ARROW_EXPORT Status FromString(
+    const std::string& s, Decimal64* out, int* precision, int* scale);
+template ARROW_EXPORT Status FromString(
+    const std::string& s, Decimal128* out, int* precision, int* scale);
+
+void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int32_t* out) {
+  DCHECK(sign == -1 || sign == 1);
+  DCHECK_NE(out, nullptr);
+  DCHECK(!whole.empty() || !fractional.empty());
+  if (!whole.empty()) {
+    *out = std::stoi(whole, nullptr, 10) *
+           static_cast<int32_t>(pow(10.0, static_cast<double>(fractional.size())));
+  }
+  if (!fractional.empty()) { *out += std::stoi(fractional, nullptr, 10); }
+  *out *= sign;
+}
+
+void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int64_t* out) {
+  DCHECK(sign == -1 || sign == 1);
+  DCHECK_NE(out, nullptr);
+  DCHECK(!whole.empty() || !fractional.empty());
+  if (!whole.empty()) {
+    *out = static_cast<int64_t>(std::stoll(whole, nullptr, 10)) *
+           static_cast<int64_t>(pow(10.0, static_cast<double>(fractional.size())));
+  }
+  if (!fractional.empty()) { *out += std::stoll(fractional, nullptr, 10); }
+  *out *= sign;
+}
+
+void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int128_t* out) {
+  DCHECK(sign == -1 || sign == 1);
+  DCHECK_NE(out, nullptr);
+  DCHECK(!whole.empty() || !fractional.empty());
+  *out = int128_t(whole + fractional) * sign;
+}
+
+void FromBytes(const uint8_t* bytes, Decimal32* decimal) {
+  DCHECK_NE(bytes, nullptr);
+  DCHECK_NE(decimal, nullptr);
+  decimal->value = *reinterpret_cast<const int32_t*>(bytes);
+}
+
+void FromBytes(const uint8_t* bytes, Decimal64* decimal) {
+  DCHECK_NE(bytes, nullptr);
+  DCHECK_NE(decimal, nullptr);
+  decimal->value = *reinterpret_cast<const int64_t*>(bytes);
+}
+
+constexpr static const size_t BYTES_IN_128_BITS = 128 / CHAR_BIT;
+constexpr static const size_t LIMB_SIZE =
+    sizeof(std::remove_pointer<int128_t::backend_type::limb_pointer>::type);
+constexpr static const size_t BYTES_PER_LIMB = BYTES_IN_128_BITS / LIMB_SIZE;
+
+void FromBytes(const uint8_t* bytes, bool is_negative, Decimal128* decimal) {
+  DCHECK_NE(bytes, nullptr);
+  DCHECK_NE(decimal, nullptr);
+
+  auto& decimal_value(decimal->value);
+  int128_t::backend_type& backend(decimal_value.backend());
+  backend.resize(BYTES_PER_LIMB, BYTES_PER_LIMB);
+  std::memcpy(backend.limbs(), bytes, BYTES_IN_128_BITS);
+  if (is_negative) { decimal->value = -decimal->value; }
+}
+
+void ToBytes(const Decimal32& value, uint8_t** bytes) {
+  DCHECK_NE(*bytes, nullptr);
+  *reinterpret_cast<int32_t*>(*bytes) = value.value;
+}
+
+void ToBytes(const Decimal64& value, uint8_t** bytes) {
+  DCHECK_NE(*bytes, nullptr);
+  *reinterpret_cast<int64_t*>(*bytes) = value.value;
+}
+
+void ToBytes(const Decimal128& decimal, uint8_t** bytes, bool* is_negative) {
+  DCHECK_NE(*bytes, nullptr);
+  DCHECK_NE(is_negative, nullptr);
+
+  /// TODO(phillipc): boost multiprecision is unreliable here, int128_t can't be
+  /// roundtripped
+  const auto& backend(decimal.value.backend());
+  auto boost_bytes = reinterpret_cast<const uint8_t*>(backend.limbs());
+  std::memcpy(*bytes, boost_bytes, BYTES_IN_128_BITS);
+  *is_negative = backend.isneg();
+}
+
+}  // namespace arrow
diff --git a/cpp/src/arrow/util/decimal.h b/cpp/src/arrow/util/decimal.h
new file mode 100644
index 00000000000..46883e3de93
--- /dev/null
+++ b/cpp/src/arrow/util/decimal.h
@@ -0,0 +1,144 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+#ifndef ARROW_DECIMAL_H
+#define ARROW_DECIMAL_H
+
+#include <cmath>
+#include <cstdlib>
+#include <iterator>
+#include <regex>
+#include <string>
+
+#include "arrow/status.h"
+#include "arrow/util/bit-util.h"
+#include "arrow/util/logging.h"
+
+#include <boost/multiprecision/cpp_int.hpp>
+
+namespace arrow {
+
+using boost::multiprecision::int128_t;
+
+template <typename T>
+struct ARROW_EXPORT Decimal;
+
+ARROW_EXPORT void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int32_t* out);
+ARROW_EXPORT void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int64_t* out);
+ARROW_EXPORT void StringToInteger(
+    const std::string& whole, const std::string& fractional, int8_t sign, int128_t* out);
+
+template <typename T>
+ARROW_EXPORT Status FromString(const std::string& s, Decimal<T>* out,
+    int* precision = nullptr, int* scale = nullptr);
+
+template <typename T>
+struct ARROW_EXPORT Decimal {
+  Decimal() : value() {}
+  explicit Decimal(const std::string& s) : value() { FromString(s, this); }
+  explicit Decimal(const char* s) : Decimal(std::string(s)) {}
+  explicit Decimal(const T& value) : value(value) {}
+
+  using value_type = T;
+  value_type value;
+};
+
+using Decimal32 = Decimal<int32_t>;
+using Decimal64 = Decimal<int64_t>;
+using Decimal128 = Decimal<int128_t>;
+
+template <typename T>
+struct ARROW_EXPORT DecimalPrecision {};
+
+template <>
+struct ARROW_EXPORT DecimalPrecision<int32_t> {
+  constexpr static const int minimum = 1;
+  constexpr static const int maximum = 9;
+};
+
+template <>
+struct ARROW_EXPORT DecimalPrecision<int64_t> {
+  constexpr static const int minimum = 10;
+  constexpr static const int maximum = 18;
+};
+
+template <>
+struct ARROW_EXPORT DecimalPrecision<int128_t> {
+  constexpr static const int minimum = 19;
+  constexpr static const int maximum = 38;
+};
+
+template <typename T>
+ARROW_EXPORT std::string ToString(
+    const Decimal<T>& decimal_value, int precision, int scale) {
+  T value = decimal_value.value;
+
+  // Decimal values are sent to clients as strings so in the interest of
+  // speed the string will be created without the using stringstream with the
+  // whole/fractional_part().
+  size_t last_char_idx = precision + (scale > 0)  // Add a space for decimal place
+                         + (scale == precision)   // Add a space for leading 0
+                         + (value < 0);           // Add a space for negative sign
+  std::string str = std::string(last_char_idx, '0');
+  // Start filling in the values in reverse order by taking the last digit
+  // of the value. Use a positive value and worry about the sign later. At this
+  // point the last_char_idx points to the string terminator.
+  T remaining_value = value;
+  size_t first_digit_idx = 0;
+  if (value < 0) {
+    remaining_value = -value;
+    first_digit_idx = 1;
+  }
+  if (scale > 0) {
+    int remaining_scale = scale;
+    do {
+      str[--last_char_idx] = static_cast<char>(
+          (remaining_value % 10) + static_cast<T>('0'));  // Ascii offset
+      remaining_value /= 10;
+    } while (--remaining_scale > 0);
+    str[--last_char_idx] = '.';
+    DCHECK_GT(last_char_idx, first_digit_idx) << "Not enough space remaining";
+  }
+  do {
+    str[--last_char_idx] =
+        static_cast<char>((remaining_value % 10) + static_cast<T>('0'));  // Ascii offset
+    remaining_value /= 10;
+    if (remaining_value == 0) {
+      // Trim any extra leading 0's.
+      if (last_char_idx > first_digit_idx) str.erase(0, last_char_idx - first_digit_idx);
+      break;
+    }
+    // For safety, enforce string length independent of remaining_value.
+  } while (last_char_idx > first_digit_idx);
+  if (value < 0) str[0] = '-';
+  return str;
+}
+
+/// Conversion from raw bytes to a Decimal value
+ARROW_EXPORT void FromBytes(const uint8_t* bytes, Decimal32* value);
+ARROW_EXPORT void FromBytes(const uint8_t* bytes, Decimal64* value);
+ARROW_EXPORT void FromBytes(const uint8_t* bytes, bool is_negative, Decimal128* decimal);
+
+/// Conversion from a Decimal value to raw bytes
+ARROW_EXPORT void ToBytes(const Decimal32& value, uint8_t** bytes);
+ARROW_EXPORT void ToBytes(const Decimal64& value, uint8_t** bytes);
+ARROW_EXPORT void ToBytes(const Decimal128& decimal, uint8_t** bytes, bool* is_negative);
+
+}  // namespace arrow
+#endif  // ARROW_DECIMAL_H
diff --git a/cpp/src/arrow/visitor_inline.h b/cpp/src/arrow/visitor_inline.h
index c61c9f59f7a..29b3db60cad 100644
--- a/cpp/src/arrow/visitor_inline.h
+++ b/cpp/src/arrow/visitor_inline.h
@@ -93,7 +93,7 @@ inline Status VisitArrayInline(const Array& array, VISITOR* visitor) {
     ARRAY_VISIT_INLINE(TimestampType);
     ARRAY_VISIT_INLINE(Time32Type);
     ARRAY_VISIT_INLINE(Time64Type);
-    // ARRAY_VISIT_INLINE(DecimalType);
+    ARRAY_VISIT_INLINE(DecimalType);
     ARRAY_VISIT_INLINE(ListType);
     ARRAY_VISIT_INLINE(StructType);
     ARRAY_VISIT_INLINE(UnionType);
diff --git a/format/Schema.fbs b/format/Schema.fbs
index ca9c8e6c3e7..badc7ea8bef 100644
--- a/format/Schema.fbs
+++ b/format/Schema.fbs
@@ -77,7 +77,9 @@ table Bool {
 }
 
 table Decimal {
+  /// Total number of decimal digits
   precision: int;
+  /// Number of digits after the decimal point "."
   scale: int;
 }
 
diff --git a/python/pyarrow/__init__.py b/python/pyarrow/__init__.py
index 8c520748cf3..7b23cf66c6f 100644
--- a/python/pyarrow/__init__.py
+++ b/python/pyarrow/__init__.py
@@ -71,7 +71,7 @@
                             uint8, uint16, uint32, uint64,
                             timestamp, date32, date64,
                             float16, float32, float64,
-                            binary, string,
+                            binary, string, decimal,
                             list_, struct, dictionary, field,
                             DataType, FixedSizeBinaryType,
                             Field, Schema, schema)
diff --git a/python/pyarrow/array.pxd b/python/pyarrow/array.pxd
index f6aaea2582e..3ba48718265 100644
--- a/python/pyarrow/array.pxd
+++ b/python/pyarrow/array.pxd
@@ -116,6 +116,10 @@ cdef class FixedSizeBinaryArray(Array):
     pass
 
 
+cdef class DecimalArray(FixedSizeBinaryArray):
+    pass
+
+
 cdef class ListArray(Array):
     pass
 
diff --git a/python/pyarrow/array.pyx b/python/pyarrow/array.pyx
index 9f302e02cdb..ee500e68129 100644
--- a/python/pyarrow/array.pyx
+++ b/python/pyarrow/array.pyx
@@ -481,6 +481,10 @@ cdef class FixedSizeBinaryArray(Array):
     pass
 
 
+cdef class DecimalArray(FixedSizeBinaryArray):
+    pass
+
+
 cdef class ListArray(Array):
     pass
 
@@ -602,6 +606,7 @@ cdef dict _array_classes = {
     Type_STRING: StringArray,
     Type_DICTIONARY: DictionaryArray,
     Type_FIXED_SIZE_BINARY: FixedSizeBinaryArray,
+    Type_DECIMAL: DecimalArray,
 }
 
 cdef object box_array(const shared_ptr[CArray]& sp_array):
diff --git a/python/pyarrow/includes/common.pxd b/python/pyarrow/includes/common.pxd
index ab38ff3084f..4860334a921 100644
--- a/python/pyarrow/includes/common.pxd
+++ b/python/pyarrow/includes/common.pxd
@@ -51,6 +51,11 @@ cdef extern from "arrow/api.h" namespace "arrow" nogil:
         c_bool IsTypeError()
 
 
+cdef extern from "arrow/util/decimal.h" namespace "arrow" nogil:
+    cdef cppclass int128_t:
+        pass
+
+
 cdef inline object PyObject_to_object(PyObject* o):
     # Cast to "object" increments reference count
     cdef object result = <object> o
diff --git a/python/pyarrow/includes/libarrow.pxd b/python/pyarrow/includes/libarrow.pxd
index 2a0488f3a01..73d96b25f52 100644
--- a/python/pyarrow/includes/libarrow.pxd
+++ b/python/pyarrow/includes/libarrow.pxd
@@ -39,6 +39,8 @@ cdef extern from "arrow/api.h" namespace "arrow" nogil:
         Type_FLOAT" arrow::Type::FLOAT"
         Type_DOUBLE" arrow::Type::DOUBLE"
 
+        Type_DECIMAL" arrow::Type::DECIMAL"
+
         Type_DATE32" arrow::Type::DATE32"
         Type_DATE64" arrow::Type::DATE64"
         Type_TIMESTAMP" arrow::Type::TIMESTAMP"
@@ -58,6 +60,11 @@ cdef extern from "arrow/api.h" namespace "arrow" nogil:
         TimeUnit_MICRO" arrow::TimeUnit::MICRO"
         TimeUnit_NANO" arrow::TimeUnit::NANO"
 
+    cdef cppclass Decimal[T]:
+        Decimal(const T&)
+
+    cdef c_string ToString[T](const Decimal[T]&, int, int)
+
     cdef cppclass CDataType" arrow::DataType":
         Type type
 
@@ -144,6 +151,12 @@ cdef extern from "arrow/api.h" namespace "arrow" nogil:
     cdef cppclass CFixedSizeBinaryType" arrow::FixedSizeBinaryType"(CFixedWidthType):
         CFixedSizeBinaryType(int byte_width)
         int byte_width()
+        int bit_width()
+
+    cdef cppclass CDecimalType" arrow::DecimalType"(CFixedSizeBinaryType):
+        int precision
+        int scale
+        CDecimalType(int precision, int scale)
 
     cdef cppclass CField" arrow::Field":
         c_string name
@@ -212,6 +225,9 @@ cdef extern from "arrow/api.h" namespace "arrow" nogil:
     cdef cppclass CFixedSizeBinaryArray" arrow::FixedSizeBinaryArray"(CArray):
         const uint8_t* GetValue(int i)
 
+    cdef cppclass CDecimalArray" arrow::DecimalArray"(CFixedSizeBinaryArray):
+        Decimal[T] Value[T](int i)
+
     cdef cppclass CListArray" arrow::ListArray"(CArray):
         const int32_t* raw_value_offsets()
         int32_t value_offset(int i)
diff --git a/python/pyarrow/scalar.pxd b/python/pyarrow/scalar.pxd
index d6c3b35160c..62a5664e57e 100644
--- a/python/pyarrow/scalar.pxd
+++ b/python/pyarrow/scalar.pxd
@@ -20,6 +20,7 @@ from pyarrow.includes.libarrow cimport *
 
 from pyarrow.schema cimport DataType
 
+
 cdef class Scalar:
     cdef readonly:
         DataType type
diff --git a/python/pyarrow/scalar.pyx b/python/pyarrow/scalar.pyx
index 1c0790a4fdc..f3d93213269 100644
--- a/python/pyarrow/scalar.pyx
+++ b/python/pyarrow/scalar.pyx
@@ -17,9 +17,10 @@
 
 from pyarrow.schema cimport DataType, box_data_type
 
+from pyarrow.includes.common cimport int128_t
 from pyarrow.compat import frombytes
 import pyarrow.schema as schema
-
+import decimal
 import datetime
 
 cimport cpython as cp
@@ -64,7 +65,7 @@ cdef class ArrayValue(Scalar):
         if hasattr(self, 'as_py'):
             return repr(self.as_py())
         else:
-            return Scalar.__repr__(self)
+            return super(Scalar, self).__repr__()
 
 
 cdef class BooleanValue(ArrayValue):
@@ -199,6 +200,25 @@ cdef class DoubleValue(ArrayValue):
         return ap.Value(self.index)
 
 
+cdef class DecimalValue(ArrayValue):
+
+    def as_py(self):
+        cdef:
+            CDecimalArray* ap = <CDecimalArray*> self.sp_array.get()
+            CDecimalType* t = <CDecimalType*> ap.type().get()
+            int bit_width = t.bit_width()
+            int precision = t.precision
+            int scale = t.scale
+            c_string s
+        if bit_width == 32:
+            s = ToString[int32_t](ap.Value[int32_t](self.index), precision, scale)
+        elif bit_width == 64:
+            s = ToString[int64_t](ap.Value[int64_t](self.index), precision, scale)
+        elif bit_width == 128:
+            s = ToString[int128_t](ap.Value[int128_t](self.index), precision, scale)
+        return decimal.Decimal(s.decode('utf8'))
+
+
 cdef class StringValue(ArrayValue):
 
     def as_py(self):
@@ -286,6 +306,7 @@ cdef dict _scalar_classes = {
     Type_BINARY: BinaryValue,
     Type_STRING: StringValue,
     Type_FIXED_SIZE_BINARY: FixedSizeBinaryValue,
+    Type_DECIMAL: DecimalValue,
 }
 
 cdef object box_scalar(DataType type, const shared_ptr[CArray]& sp_array,
diff --git a/python/pyarrow/schema.pxd b/python/pyarrow/schema.pxd
index 94d65bfc157..eceedbad0ba 100644
--- a/python/pyarrow/schema.pxd
+++ b/python/pyarrow/schema.pxd
@@ -20,6 +20,7 @@ from pyarrow.includes.libarrow cimport (CDataType,
                                         CDictionaryType,
                                         CTimestampType,
                                         CFixedSizeBinaryType,
+                                        CDecimalType,
                                         CField, CSchema)
 
 cdef class DataType:
@@ -27,7 +28,7 @@ cdef class DataType:
         shared_ptr[CDataType] sp_type
         CDataType* type
 
-    cdef init(self, const shared_ptr[CDataType]& type)
+    cdef void init(self, const shared_ptr[CDataType]& type)
 
 
 cdef class DictionaryType(DataType):
@@ -45,6 +46,11 @@ cdef class FixedSizeBinaryType(DataType):
         const CFixedSizeBinaryType* fixed_size_binary_type
 
 
+cdef class DecimalType(FixedSizeBinaryType):
+    cdef:
+        const CDecimalType* decimal_type
+
+
 cdef class Field:
     cdef:
         shared_ptr[CField] sp_field
@@ -55,6 +61,7 @@ cdef class Field:
 
     cdef init(self, const shared_ptr[CField]& field)
 
+
 cdef class Schema:
     cdef:
         shared_ptr[CSchema] sp_schema
@@ -63,6 +70,7 @@ cdef class Schema:
     cdef init(self, const vector[shared_ptr[CField]]& fields)
     cdef init_schema(self, const shared_ptr[CSchema]& schema)
 
+
 cdef DataType box_data_type(const shared_ptr[CDataType]& type)
 cdef Field box_field(const shared_ptr[CField]& field)
 cdef Schema box_schema(const shared_ptr[CSchema]& schema)
diff --git a/python/pyarrow/schema.pyx b/python/pyarrow/schema.pyx
index 253be4590b5..4b931bf4522 100644
--- a/python/pyarrow/schema.pyx
+++ b/python/pyarrow/schema.pyx
@@ -29,6 +29,7 @@ from pyarrow.array cimport Array
 from pyarrow.error cimport check_status
 from pyarrow.includes.libarrow cimport (CDataType, CStructType, CListType,
                                         CFixedSizeBinaryType,
+                                        CDecimalType,
                                         TimeUnit_SECOND, TimeUnit_MILLI,
                                         TimeUnit_MICRO, TimeUnit_NANO,
                                         Type, TimeUnit)
@@ -45,7 +46,7 @@ cdef class DataType:
     def __cinit__(self):
         pass
 
-    cdef init(self, const shared_ptr[CDataType]& type):
+    cdef void init(self, const shared_ptr[CDataType]& type):
         self.sp_type = type
         self.type = type.get()
 
@@ -66,14 +67,14 @@ cdef class DataType:
 
 cdef class DictionaryType(DataType):
 
-    cdef init(self, const shared_ptr[CDataType]& type):
+    cdef void init(self, const shared_ptr[CDataType]& type):
         DataType.init(self, type)
         self.dict_type = <const CDictionaryType*> type.get()
 
 
 cdef class TimestampType(DataType):
 
-    cdef init(self, const shared_ptr[CDataType]& type):
+    cdef void init(self, const shared_ptr[CDataType]& type):
         DataType.init(self, type)
         self.ts_type = <const CTimestampType*> type.get()
 
@@ -93,7 +94,7 @@ cdef class TimestampType(DataType):
 
 cdef class FixedSizeBinaryType(DataType):
 
-    cdef init(self, const shared_ptr[CDataType]& type):
+    cdef void init(self, const shared_ptr[CDataType]& type):
         DataType.init(self, type)
         self.fixed_size_binary_type = <const CFixedSizeBinaryType*> type.get()
 
@@ -103,6 +104,13 @@ cdef class FixedSizeBinaryType(DataType):
             return self.fixed_size_binary_type.byte_width()
 
 
+cdef class DecimalType(FixedSizeBinaryType):
+
+    cdef void init(self, const shared_ptr[CDataType]& type):
+        DataType.init(self, type)
+        self.decimal_type = <const CDecimalType*> type.get()
+
+
 cdef class Field:
 
     def __cinit__(self):
@@ -354,6 +362,12 @@ def float64():
     return primitive_type(la.Type_DOUBLE)
 
 
+cpdef DataType decimal(int precision, int scale=0):
+    cdef shared_ptr[CDataType] decimal_type
+    decimal_type.reset(new CDecimalType(precision, scale))
+    return box_data_type(decimal_type)
+
+
 def string():
     """
     UTF8 string
@@ -374,11 +388,9 @@ def binary(int length=-1):
     if length == -1:
         return primitive_type(la.Type_BINARY)
 
-    cdef FixedSizeBinaryType out = FixedSizeBinaryType()
     cdef shared_ptr[CDataType] fixed_size_binary_type
     fixed_size_binary_type.reset(new CFixedSizeBinaryType(length))
-    out.init(fixed_size_binary_type)
-    return out
+    return box_data_type(fixed_size_binary_type)
 
 
 def list_(DataType value_type):
@@ -436,6 +448,8 @@ cdef DataType box_data_type(const shared_ptr[CDataType]& type):
         out = TimestampType()
     elif type.get().type == la.Type_FIXED_SIZE_BINARY:
         out = FixedSizeBinaryType()
+    elif type.get().type == la.Type_DECIMAL:
+        out = DecimalType()
     else:
         out = DataType()
 
diff --git a/python/pyarrow/tests/test_convert_builtin.py b/python/pyarrow/tests/test_convert_builtin.py
index e2b03d85ecd..d89a8e0c54c 100644
--- a/python/pyarrow/tests/test_convert_builtin.py
+++ b/python/pyarrow/tests/test_convert_builtin.py
@@ -20,6 +20,7 @@
 import pyarrow as pa
 
 import datetime
+import decimal
 
 
 class TestConvertList(unittest.TestCase):
@@ -162,3 +163,42 @@ def test_mixed_types_fails(self):
         data = ['a', 1, 2.0]
         with self.assertRaises(pa.ArrowException):
             pa.from_pylist(data)
+
+    def test_decimal(self):
+        data = [decimal.Decimal('1234.183'), decimal.Decimal('8094.234')]
+        type = pa.decimal(precision=7, scale=3)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
+
+    def test_decimal_different_precisions(self):
+        data = [
+            decimal.Decimal('1234234983.183'), decimal.Decimal('80943244.234')
+        ]
+        type = pa.decimal(precision=13, scale=3)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
+
+    def test_decimal_no_scale(self):
+        data = [decimal.Decimal('1234234983'), decimal.Decimal('8094324')]
+        type = pa.decimal(precision=10)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
+
+    def test_decimal_negative(self):
+        data = [decimal.Decimal('-1234.234983'), decimal.Decimal('-8.094324')]
+        type = pa.decimal(precision=10, scale=6)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
+
+    def test_decimal_no_whole_part(self):
+        data = [decimal.Decimal('-.4234983'), decimal.Decimal('.0103943')]
+        type = pa.decimal(precision=7, scale=7)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
+
+    def test_decimal_large_integer(self):
+        data = [decimal.Decimal('-394029506937548693.42983'),
+                decimal.Decimal('32358695912932.01033')]
+        type = pa.decimal(precision=23, scale=5)
+        arr = pa.from_pylist(data, type=type)
+        assert arr.to_pylist() == data
diff --git a/python/pyarrow/tests/test_convert_pandas.py b/python/pyarrow/tests/test_convert_pandas.py
index 87c9c03d7da..0504e1ddb4f 100644
--- a/python/pyarrow/tests/test_convert_pandas.py
+++ b/python/pyarrow/tests/test_convert_pandas.py
@@ -20,6 +20,7 @@
 
 import datetime
 import unittest
+import decimal
 
 import numpy as np
 
@@ -451,3 +452,72 @@ def test_strided_data_import(self):
             self._check_pandas_roundtrip(df)
             self._check_array_roundtrip(col)
             self._check_array_roundtrip(col, mask=strided_mask)
+
+    def test_decimal_32_from_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('-1234.123'),
+                decimal.Decimal('1234.439'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        field = A.Field.from_py('decimals', A.decimal(7, 3))
+        schema = A.Schema.from_fields([field])
+        assert converted.schema.equals(schema)
+
+    def test_decimal_32_to_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('-1234.123'),
+                decimal.Decimal('1234.439'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        df = converted.to_pandas()
+        tm.assert_frame_equal(df, expected)
+
+    def test_decimal_64_from_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('-129934.123331'),
+                decimal.Decimal('129534.123731'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        field = A.Field.from_py('decimals', A.decimal(12, 6))
+        schema = A.Schema.from_fields([field])
+        assert converted.schema.equals(schema)
+
+    def test_decimal_64_to_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('-129934.123331'),
+                decimal.Decimal('129534.123731'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        df = converted.to_pandas()
+        tm.assert_frame_equal(df, expected)
+
+    def test_decimal_128_from_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('394092382910493.12341234678'),
+                -decimal.Decimal('314292388910493.12343437128'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        field = A.Field.from_py('decimals', A.decimal(26, 11))
+        schema = A.Schema.from_fields([field])
+        assert converted.schema.equals(schema)
+
+    def test_decimal_128_to_pandas(self):
+        expected = pd.DataFrame({
+            'decimals': [
+                decimal.Decimal('394092382910493.12341234678'),
+                -decimal.Decimal('314292388910493.12343437128'),
+            ]
+        })
+        converted = A.Table.from_pandas(expected)
+        df = converted.to_pandas()
+        tm.assert_frame_equal(df, expected)