diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/xml/XmlInferSchema.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/xml/XmlInferSchema.scala
index be5a29d299a3..b9342c53d020 100644
--- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/xml/XmlInferSchema.scala
+++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/xml/XmlInferSchema.scala
@@ -195,7 +195,9 @@ class XmlInferSchema(options: XmlOptions, caseSensitive: Boolean)
 
   private def inferField(parser: XMLEventReader): DataType = {
     parser.peek match {
-      case _: EndElement => NullType
+      case _: EndElement =>
+        parser.nextEvent()
+        NullType
       case _: StartElement => inferObject(parser)
       case _: Characters =>
         val structType = inferObject(parser).asInstanceOf[StructType]
@@ -450,7 +452,7 @@ class XmlInferSchema(options: XmlOptions, caseSensitive: Boolean)
     oldTypeOpt match {
       // If the field name already exists,
       // merge the type and infer the combined field as an array type if necessary
-      case Some(oldType) if !oldType.isInstanceOf[ArrayType] && !newType.isInstanceOf[NullType] =>
+      case Some(oldType) if !oldType.isInstanceOf[ArrayType] =>
         ArrayType(compatibleType(caseSensitive, options.valueTag)(oldType, newType))
       case Some(oldType) =>
         compatibleType(caseSensitive, options.valueTag)(oldType, newType)