By Ece Toksabay and Birsen Altayli

ANKARA (Reuters) -The man challenging Tayyip Erdogan in Turkey's presidential elections this weekend received a boost on Thursday when a small party candidate abruptly withdrew and a closely-watched poll gave him a more than five percentage point advantage.

The withdrawal by Muharrem Ince, one of four contesting Sunday's presidential vote, could reshape the last days of campaigning in what is seen as Erdogan's biggest test in his two-decade reign.

Turkey's main stock index jumped 6% after Ince made the shock announcement in front of his party's headquarters in Ankara. Financial markets are on edge given Kemal Kilicdaroglu has pledged to reverse Erdogan's unorthodox economic policies.

The survey by pollster Konda put support for Erdogan on 43.7% and Kilicdaroglu on 49.3%, leaving him short of the majority needed to win in the first round and suggesting the election would go to a run-off between the two men on May 28.

The survey was carried out May 6-7, before Ince's announcement.

Nonetheless, the findings were largely in line with some other polls that put Kilicdaroglu ahead. He was named candidate of a six-party opposition alliance and also heads the Republican People's Party (CHP), the banner under which Ince unsuccessfully ran for president in 2018.

Erdogan's re-election bid has been complicated by a cost-of-living crisis, triggered by a lira slump and soaring inflation, and a devastating earthquake in February which killed more than 50,000 people in Turkey and left millions homeless.

Ince, who had 2.2% support in the Konda poll, said he had been the target of a smear campaign including fake videos and documents being circulated on social media, and he accused journalists and prosecutors of not doing their jobs.

"I am not afraid of plots, fake video edits, invoices and non-existent jeeps. I have been resisting these for 45 days already," he said, adding that some people had accused him of working with Erdogan to prevent Kilicdaroglu's election.

"I am withdrawing from the candidacy. I am doing this for my country," he said.

He did not back another candidate and asked people to vote for his party in the parliamentary elections.

"There is no doubt that Erdogan is facing a majority that wants change - and that includes younger people," said Asli Aydintasbas, a Brookings Institution visiting fellow. "The only question is whether folks believe Kilicdaroglu is that agent of change."

"Whether he barely wins or not, I feel like the Erdogan era is over," she added. "Turkish society is ready to move on. And sadly President Erdogan is not leaving behind an institutional governance model."

DOWN TO THE WIRE

The Konda survey put support for the fourth presidential candidate, Sinan Ogan, at 4.8%. Konda had said the majority of Ogan's and Ince's voters were leaning towards voting for Kilicdaroglu in a second round.

A Metropoll survey also showed the vote going to a second round, with Kilicdaroglu getting 49.1% and Erdogan 46.9%. In a run-off, it showed Kilicdaroglu winning with 51.3%.

Hakan Akbas, managing director of Strategic Advisory Services, an Istanbul-based political advisory, said Erdogan aims to get to a second round vote against Kilicdaroglu.

"Given the earthquakes and economic crisis, this would still be a success for him. What matters now even more is parliamentary results," he said.

"If it is a hung parliament, Erdogan will appeal to voters to select stability over chaos likely following coalition of six opposition parties."

The Konda survey put support for Erdogan's ruling alliance on 44.0% in the parliamentary vote, ahead of the main opposition alliance on 39.9%. The pro-Kurdish HDP party, which is backing Kilicdaroglu, is expected to play a 'kingmaker' role.

Konda said the HDP, running under another party's emblem due to the threat of a court ban, and its leftist allies are seen winning 12.3% of support in the parliamentary vote. That would leave Erdogan and his allies in the minority.

Konda, which publicly releases only one poll ahead of votes, undertook face-to-face interviews with 3,480 people across 35 provincial centres. The survey has a margin of error of +/- 2.2% at 99% confidence level, Konda said.

(Additional reporting by Jonathan Spicer, Ezgi Erkoyun and Ali Kucukgocmen; Writing by Daren Butler;Editing by Jonathan Spicer, Frank Jack Daniel, Alexandra Hudson)