Yes, you can, but the results won't be good - from the FAQ:

I have used a GTF file generated by the Table Browser function of the UCSC Genome Browser, and most reads are counted as ambiguous. Why?

In these files, the gene_id attribute incorrectly contains the same value as the transcript_id attribute and hence a different value for each transcript of the same gene. Hence, if a read maps to an exon shared by several transcripts of the same gene, this will appear to htseq-count as and overlap with several genes. Therefore, these GTF files cannot be used as is. Either correct the incorrect gene_id attributes with a suitable script, or use a GTF file from a different source.

edit: Try the following solution, taken from the excellent Introduction to differential gene expression analysis using RNA-seq:

# first , download a table  for "Genes  and  Gene  Predictions" from  the  UCSC  Table 
# Browser  indicating  as the  output  format: "all  fields  from  selected  table"
# NOTE: this  may not  work  for all GTF  files  downloaded  from  UCSC! genePredToGtf
# is very  finicky  and  every  organism's annotation  may  have  been  generated  and
# deposited  by a different  person)

head -n1  allfields_hg19.txt

# remove  first  column  and  first line , feed  that  into  genePredToGtf

cut -f 2- allfields_hg19.txt | sed '1d' | \
genePredToGtf file stdin hg19_RefSeq.gtf

head -n1  hg19_RefSeq.gtf